repo
stringlengths
2
99
file
stringlengths
13
225
code
stringlengths
0
18.3M
file_length
int64
0
18.3M
avg_line_length
float64
0
1.36M
max_line_length
int64
0
4.26M
extension_type
stringclasses
1 value
Few-shot-WSI
Few-shot-WSI-master/openselfsup/hooks/__init__.py
from .builder import build_hook from .byol_hook import BYOLHook from .deepcluster_hook import DeepClusterHook from .odc_hook import ODCHook from .optimizer_hook import DistOptimizerHook from .extractor import Extractor from .validate_hook import ValidateHook from .registry import HOOKS
287
31
45
py
Few-shot-WSI
Few-shot-WSI-master/openselfsup/hooks/builder.py
from openselfsup.utils import build_from_cfg from .registry import HOOKS def build_hook(cfg, default_args=None): return build_from_cfg(cfg, HOOKS, default_args)
168
20.125
51
py
Few-shot-WSI
Few-shot-WSI-master/openselfsup/hooks/deepcluster_hook.py
import numpy as np from mmcv.runner import Hook import torch import torch.distributed as dist from openselfsup.third_party import clustering as _clustering from openselfsup.utils import print_log from .registry import HOOKS from .extractor import Extractor @HOOKS.register_module class DeepClusterHook(Hook): """Hook for DeepCluster. Args: extractor (dict): Config dict for feature extraction. clustering (dict): Config dict that specifies the clustering algorithm. unif_sampling (bool): Whether to apply uniform sampling. reweight (bool): Whether to apply loss re-weighting. reweight_pow (float): The power of re-weighting. init_memory (bool): Whether to initialize memory banks for ODC. Default: False. initial (bool): Whether to call the hook initially. Default: True. interval (int): Frequency of epochs to call the hook. Default: 1. dist_mode (bool): Use distributed training or not. Default: True. data_loaders (DataLoader): A PyTorch dataloader. Default: None. """ def __init__( self, extractor, clustering, unif_sampling, reweight, reweight_pow, init_memory=False, # for ODC initial=True, interval=1, dist_mode=True, data_loaders=None): self.extractor = Extractor(dist_mode=dist_mode, **extractor) self.clustering_type = clustering.pop('type') self.clustering_cfg = clustering self.unif_sampling = unif_sampling self.reweight = reweight self.reweight_pow = reweight_pow self.init_memory = init_memory self.initial = initial self.interval = interval self.dist_mode = dist_mode self.data_loaders = data_loaders def before_run(self, runner): if self.initial: self.deepcluster(runner) def after_train_epoch(self, runner): if not self.every_n_epochs(runner, self.interval): return self.deepcluster(runner) def deepcluster(self, runner): # step 1: get features runner.model.eval() features = self.extractor(runner) runner.model.train() # step 2: get labels if not self.dist_mode or (self.dist_mode and runner.rank == 0): clustering_algo = _clustering.__dict__[self.clustering_type]( **self.clustering_cfg) # Features are normalized during clustering clustering_algo.cluster(features, verbose=True) assert isinstance(clustering_algo.labels, np.ndarray) new_labels = clustering_algo.labels.astype(np.int64) np.save( "{}/cluster_epoch_{}.npy".format(runner.work_dir, runner.epoch), new_labels) self.evaluate(runner, new_labels) else: new_labels = np.zeros((len(self.data_loaders[0].dataset), ), dtype=np.int64) if self.dist_mode: new_labels_tensor = torch.from_numpy(new_labels).cuda() dist.broadcast(new_labels_tensor, 0) new_labels = new_labels_tensor.cpu().numpy() new_labels_list = list(new_labels) # step 3: assign new labels self.data_loaders[0].dataset.assign_labels(new_labels_list) # step 4 (a): set uniform sampler if self.unif_sampling: self.data_loaders[0].sampler.set_uniform_indices( new_labels_list, self.clustering_cfg.k) # step 4 (b): set loss reweight if self.reweight: runner.model.module.set_reweight(new_labels, self.reweight_pow) # step 5: randomize classifier runner.model.module.head.init_weights(init_linear='normal') if self.dist_mode: for p in runner.model.module.head.state_dict().values(): dist.broadcast(p, 0) # step 6: init memory for ODC if self.init_memory: runner.model.module.memory_bank.init_memory(features, new_labels) def evaluate(self, runner, new_labels): hist = np.bincount(new_labels, minlength=self.clustering_cfg.k) empty_cls = (hist == 0).sum() minimal_cls_size, maximal_cls_size = hist.min(), hist.max() if runner.rank == 0: print_log( "empty_num: {}\tmin_cluster: {}\tmax_cluster:{}".format( empty_cls.item(), minimal_cls_size.item(), maximal_cls_size.item()), logger='root')
4,637
36.104
79
py
Few-shot-WSI
Few-shot-WSI-master/openselfsup/utils/contextmanagers.py
# coding: utf-8 import asyncio import contextlib import logging import os import time from typing import List import torch logger = logging.getLogger(__name__) DEBUG_COMPLETED_TIME = bool(os.environ.get('DEBUG_COMPLETED_TIME', False)) @contextlib.asynccontextmanager async def completed(trace_name='', name='', sleep_interval=0.05, streams: List[torch.cuda.Stream] = None): """Async context manager that waits for work to complete on given CUDA streams.""" if not torch.cuda.is_available(): yield return stream_before_context_switch = torch.cuda.current_stream() if not streams: streams = [stream_before_context_switch] else: streams = [s if s else stream_before_context_switch for s in streams] end_events = [ torch.cuda.Event(enable_timing=DEBUG_COMPLETED_TIME) for _ in streams ] if DEBUG_COMPLETED_TIME: start = torch.cuda.Event(enable_timing=True) stream_before_context_switch.record_event(start) cpu_start = time.monotonic() logger.debug('%s %s starting, streams: %s', trace_name, name, streams) grad_enabled_before = torch.is_grad_enabled() try: yield finally: current_stream = torch.cuda.current_stream() assert current_stream == stream_before_context_switch if DEBUG_COMPLETED_TIME: cpu_end = time.monotonic() for i, stream in enumerate(streams): event = end_events[i] stream.record_event(event) grad_enabled_after = torch.is_grad_enabled() # observed change of torch.is_grad_enabled() during concurrent run of # async_test_bboxes code assert (grad_enabled_before == grad_enabled_after ), 'Unexpected is_grad_enabled() value change' are_done = [e.query() for e in end_events] logger.debug('%s %s completed: %s streams: %s', trace_name, name, are_done, streams) with torch.cuda.stream(stream_before_context_switch): while not all(are_done): await asyncio.sleep(sleep_interval) are_done = [e.query() for e in end_events] logger.debug( '%s %s completed: %s streams: %s', trace_name, name, are_done, streams, ) current_stream = torch.cuda.current_stream() assert current_stream == stream_before_context_switch if DEBUG_COMPLETED_TIME: cpu_time = (cpu_end - cpu_start) * 1000 stream_times_ms = '' for i, stream in enumerate(streams): elapsed_time = start.elapsed_time(end_events[i]) stream_times_ms += ' {} {:.2f} ms'.format(stream, elapsed_time) logger.info('%s %s %.2f ms %s', trace_name, name, cpu_time, stream_times_ms) @contextlib.asynccontextmanager async def concurrent(streamqueue: asyncio.Queue, trace_name='concurrent', name='stream'): """Run code concurrently in different streams. :param streamqueue: asyncio.Queue instance. Queue tasks define the pool of streams used for concurrent execution. """ if not torch.cuda.is_available(): yield return initial_stream = torch.cuda.current_stream() with torch.cuda.stream(initial_stream): stream = await streamqueue.get() assert isinstance(stream, torch.cuda.Stream) try: with torch.cuda.stream(stream): logger.debug('%s %s is starting, stream: %s', trace_name, name, stream) yield current = torch.cuda.current_stream() assert current == stream logger.debug('%s %s has finished, stream: %s', trace_name, name, stream) finally: streamqueue.task_done() streamqueue.put_nowait(stream)
4,103
32.365854
79
py
Few-shot-WSI
Few-shot-WSI-master/openselfsup/utils/registry.py
import inspect from functools import partial import mmcv class Registry(object): def __init__(self, name): self._name = name self._module_dict = dict() def __repr__(self): format_str = self.__class__.__name__ + '(name={}, items={})'.format( self._name, list(self._module_dict.keys())) return format_str @property def name(self): return self._name @property def module_dict(self): return self._module_dict def get(self, key): return self._module_dict.get(key, None) def _register_module(self, module_class, force=False): """Register a module. Args: module (:obj:`nn.Module`): Module to be registered. """ if not inspect.isclass(module_class): raise TypeError('module must be a class, but got {}'.format( type(module_class))) module_name = module_class.__name__ if not force and module_name in self._module_dict: raise KeyError('{} is already registered in {}'.format( module_name, self.name)) self._module_dict[module_name] = module_class def register_module(self, cls=None, force=False): if cls is None: return partial(self.register_module, force=force) self._register_module(cls, force=force) return cls def build_from_cfg(cfg, registry, default_args=None): """Build a module from config dict. Args: cfg (dict): Config dict. It should at least contain the key "type". registry (:obj:`Registry`): The registry to search the type from. default_args (dict, optional): Default initialization arguments. Returns: obj: The constructed object. """ assert isinstance(cfg, dict) and 'type' in cfg assert isinstance(default_args, dict) or default_args is None args = cfg.copy() obj_type = args.pop('type') if mmcv.is_str(obj_type): obj_cls = registry.get(obj_type) if obj_cls is None: raise KeyError('{} is not in the {} registry'.format( obj_type, registry.name)) elif inspect.isclass(obj_type): obj_cls = obj_type else: raise TypeError('type must be a str or valid type, but got {}'.format( type(obj_type))) if default_args is not None: for name, value in default_args.items(): args.setdefault(name, value) return obj_cls(**args)
2,478
29.9875
78
py
Few-shot-WSI
Few-shot-WSI-master/openselfsup/utils/optimizers.py
import torch from torch.optim.optimizer import Optimizer, required from torch.optim import * class LARS(Optimizer): r"""Implements layer-wise adaptive rate scaling for SGD. Args: params (iterable): iterable of parameters to optimize or dicts defining parameter groups lr (float): base learning rate (\gamma_0) momentum (float, optional): momentum factor (default: 0) ("m") weight_decay (float, optional): weight decay (L2 penalty) (default: 0) ("\beta") dampening (float, optional): dampening for momentum (default: 0) eta (float, optional): LARS coefficient nesterov (bool, optional): enables Nesterov momentum (default: False) Based on Algorithm 1 of the following paper by You, Gitman, and Ginsburg. Large Batch Training of Convolutional Networks: https://arxiv.org/abs/1708.03888 Example: >>> optimizer = LARS(model.parameters(), lr=0.1, momentum=0.9, >>> weight_decay=1e-4, eta=1e-3) >>> optimizer.zero_grad() >>> loss_fn(model(input), target).backward() >>> optimizer.step() """ def __init__(self, params, lr=required, momentum=0, dampening=0, weight_decay=0, eta=0.001, nesterov=False): if lr is not required and lr < 0.0: raise ValueError("Invalid learning rate: {}".format(lr)) if momentum < 0.0: raise ValueError("Invalid momentum value: {}".format(momentum)) if weight_decay < 0.0: raise ValueError( "Invalid weight_decay value: {}".format(weight_decay)) if eta < 0.0: raise ValueError("Invalid LARS coefficient value: {}".format(eta)) defaults = dict( lr=lr, momentum=momentum, dampening=dampening, weight_decay=weight_decay, nesterov=nesterov, eta=eta) if nesterov and (momentum <= 0 or dampening != 0): raise ValueError("Nesterov momentum requires a momentum and zero dampening") super(LARS, self).__init__(params, defaults) def __setstate__(self, state): super(LARS, self).__setstate__(state) for group in self.param_groups: group.setdefault('nesterov', False) @torch.no_grad() def step(self, closure=None): """Performs a single optimization step. Args: closure (callable, optional): A closure that reevaluates the model and returns the loss. """ loss = None if closure is not None: with torch.enable_grad(): loss = closure() for group in self.param_groups: weight_decay = group['weight_decay'] momentum = group['momentum'] dampening = group['dampening'] eta = group['eta'] nesterov = group['nesterov'] lr = group['lr'] lars_exclude = group.get('lars_exclude', False) for p in group['params']: if p.grad is None: continue d_p = p.grad if lars_exclude: local_lr = 1. else: weight_norm = torch.norm(p).item() grad_norm = torch.norm(d_p).item() # Compute local learning rate for this layer local_lr = eta * weight_norm / \ (grad_norm + weight_decay * weight_norm) actual_lr = local_lr * lr d_p = d_p.add(p, alpha=weight_decay).mul(actual_lr) if momentum != 0: param_state = self.state[p] if 'momentum_buffer' not in param_state: buf = param_state['momentum_buffer'] = \ torch.clone(d_p).detach() else: buf = param_state['momentum_buffer'] buf.mul_(momentum).add_(d_p, alpha=1 - dampening) if nesterov: d_p = d_p.add(buf, alpha=momentum) else: d_p = buf p.add_(-d_p) return loss
4,327
35.991453
88
py
Few-shot-WSI
Few-shot-WSI-master/openselfsup/utils/profiling.py
import contextlib import sys import time import torch if sys.version_info >= (3, 7): @contextlib.contextmanager def profile_time(trace_name, name, enabled=True, stream=None, end_stream=None): """Print time spent by CPU and GPU. Useful as a temporary context manager to find sweet spots of code suitable for async implementation. """ if (not enabled) or not torch.cuda.is_available(): yield return stream = stream if stream else torch.cuda.current_stream() end_stream = end_stream if end_stream else stream start = torch.cuda.Event(enable_timing=True) end = torch.cuda.Event(enable_timing=True) stream.record_event(start) try: cpu_start = time.monotonic() yield finally: cpu_end = time.monotonic() end_stream.record_event(end) end.synchronize() cpu_time = (cpu_end - cpu_start) * 1000 gpu_time = start.elapsed_time(end) msg = "{} {} cpu_time {:.2f} ms ".format(trace_name, name, cpu_time) msg += "gpu_time {:.2f} ms stream {}".format(gpu_time, stream) print(msg, end_stream)
1,363
32.268293
74
py
Few-shot-WSI
Few-shot-WSI-master/openselfsup/utils/collect.py
import numpy as np import mmcv import torch from .gather import gather_tensors_batch def nondist_forward_collect(func, data_loader, length): """Forward and collect network outputs. This function performs forward propagation and collects outputs. It can be used to collect results, features, losses, etc. Args: func (function): The function to process data. The output must be a dictionary of CPU tensors. length (int): Expected length of output arrays. Returns: results_all (dict(np.ndarray)): The concatenated outputs. """ results = [] prog_bar = mmcv.ProgressBar(len(data_loader)) for i, data in enumerate(data_loader): with torch.no_grad(): result = func(**data) results.append(result) prog_bar.update() results_all = {} for k in results[0].keys(): results_all[k] = np.concatenate( [batch[k].numpy() for batch in results], axis=0) assert results_all[k].shape[0] == length return results_all def dist_forward_collect(func, data_loader, rank, length, ret_rank=-1): """Forward and collect network outputs in a distributed manner. This function performs forward propagation and collects outputs. It can be used to collect results, features, losses, etc. Args: func (function): The function to process data. The output must be a dictionary of CPU tensors. rank (int): This process id. length (int): Expected length of output arrays. ret_rank (int): The process that returns. Other processes will return None. Returns: results_all (dict(np.ndarray)): The concatenated outputs. """ results = [] if rank == 0: prog_bar = mmcv.ProgressBar(len(data_loader)) for idx, data in enumerate(data_loader): with torch.no_grad(): result = func(**data) # dict{key: tensor} results.append(result) if rank == 0: prog_bar.update() results_all = {} for k in results[0].keys(): results_cat = np.concatenate([batch[k].numpy() for batch in results], axis=0) if ret_rank == -1: results_gathered = gather_tensors_batch(results_cat, part_size=20) results_strip = np.concatenate(results_gathered, axis=0)[:length] else: results_gathered = gather_tensors_batch( results_cat, part_size=20, ret_rank=ret_rank) if rank == ret_rank: results_strip = np.concatenate( results_gathered, axis=0)[:length] else: results_strip = None results_all[k] = results_strip return results_all
2,773
32.02381
78
py
Few-shot-WSI
Few-shot-WSI-master/openselfsup/utils/misc.py
from functools import partial import mmcv import numpy as np from six.moves import map, zip def tensor2imgs(tensor, mean=(0, 0, 0), std=(1, 1, 1), to_rgb=True): num_imgs = tensor.size(0) mean = np.array(mean, dtype=np.float32) std = np.array(std, dtype=np.float32) imgs = [] for img_id in range(num_imgs): img = tensor[img_id, ...].cpu().numpy().transpose(1, 2, 0) img = mmcv.imdenormalize( img, mean, std, to_bgr=to_rgb).astype(np.uint8) imgs.append(np.ascontiguousarray(img)) return imgs def multi_apply(func, *args, **kwargs): pfunc = partial(func, **kwargs) if kwargs else func map_results = map(pfunc, *args) return tuple(map(list, zip(*map_results))) def unmap(data, count, inds, fill=0): """Unmap a subset of item (data) back to the original set of items (of size count).""" if data.dim() == 1: ret = data.new_full((count, ), fill) ret[inds] = data else: new_size = (count, ) + data.size()[1:] ret = data.new_full(new_size, fill) ret[inds, :] = data return ret
1,107
28.157895
74
py
Few-shot-WSI
Few-shot-WSI-master/openselfsup/utils/logger.py
import logging from mmcv.runner import get_dist_info def get_root_logger(log_file=None, log_level=logging.INFO): """Get the root logger. The logger will be initialized if it has not been initialized. By default a StreamHandler will be added. If `log_file` is specified, a FileHandler will also be added. The name of the root logger is the top-level package name, e.g., "openselfsup". Args: log_file (str | None): The log filename. If specified, a FileHandler will be added to the root logger. log_level (int): The root logger level. Note that only the process of rank 0 is affected, while other processes will set the level to "Error" and be silent most of the time. Returns: logging.Logger: The root logger. """ logger = logging.getLogger(__name__.split('.')[0]) # i.e., openselfsup # if the logger has been initialized, just return it if logger.hasHandlers(): return logger format_str = '%(asctime)s - %(name)s - %(levelname)s - %(message)s' logging.basicConfig(format=format_str, level=log_level) rank, _ = get_dist_info() if rank != 0: logger.setLevel('ERROR') elif log_file is not None: file_handler = logging.FileHandler(log_file, 'w') file_handler.setFormatter(logging.Formatter(format_str)) file_handler.setLevel(log_level) logger.addHandler(file_handler) return logger def print_log(msg, logger=None, level=logging.INFO): """Print a log message. Args: msg (str): The message to be logged. logger (logging.Logger | str | None): The logger to be used. Some special loggers are: - "root": the root logger obtained with `get_root_logger()`. - "silent": no message will be printed. - None: The `print()` method will be used to print log messages. level (int): Logging level. Only available when `logger` is a Logger object or "root". """ if logger is None: print(msg) elif logger == 'root': _logger = get_root_logger() _logger.log(level, msg) elif isinstance(logger, logging.Logger): logger.log(level, msg) elif logger != 'silent': raise TypeError( 'logger should be either a logging.Logger object, "root", ' '"silent" or None, but got {}'.format(logger))
2,424
35.19403
79
py
Few-shot-WSI
Few-shot-WSI-master/openselfsup/utils/alias_multinomial.py
import torch import numpy as np class AliasMethod(object): """The alias method for sampling. From: https://hips.seas.harvard.edu/blog/2013/03/03/the-alias-method-efficient-sampling-with-many-discrete-outcomes/ Args: probs (Tensor): Sampling probabilities. """ def __init__(self, probs): if probs.sum() > 1: probs.div_(probs.sum()) K = len(probs) self.prob = torch.zeros(K) self.alias = torch.LongTensor([0] * K) # Sort the data into the outcomes with probabilities # that are larger and smaller than 1/K. smaller = [] larger = [] for kk, prob in enumerate(probs): self.prob[kk] = K * prob if self.prob[kk] < 1.0: smaller.append(kk) else: larger.append(kk) # Loop though and create little binary mixtures that # appropriately allocate the larger outcomes over the # overall uniform mixture. while len(smaller) > 0 and len(larger) > 0: small = smaller.pop() large = larger.pop() self.alias[small] = large self.prob[large] = (self.prob[large] - 1.0) + self.prob[small] if self.prob[large] < 1.0: smaller.append(large) else: larger.append(large) for last_one in smaller + larger: self.prob[last_one] = 1 def cuda(self): self.prob = self.prob.cuda() self.alias = self.alias.cuda() def draw(self, N): """Draw N samples from multinomial. Args: N (int): Number of samples. Returns: Tensor: Samples. """ K = self.alias.size(0) kk = torch.zeros( N, dtype=torch.long, device=self.prob.device).random_(0, K) prob = self.prob.index_select(0, kk) alias = self.alias.index_select(0, kk) # b is whether a random number is greater than q b = torch.bernoulli(prob) oq = kk.mul(b.long()) oj = alias.mul((1 - b).long()) return oq + oj
2,132
27.065789
120
py
Few-shot-WSI
Few-shot-WSI-master/openselfsup/utils/gather.py
import numpy as np import torch import torch.distributed as dist def gather_tensors(input_array): world_size = dist.get_world_size() ## gather shapes first myshape = input_array.shape mycount = input_array.size shape_tensor = torch.Tensor(np.array(myshape)).cuda() all_shape = [ torch.Tensor(np.array(myshape)).cuda() for i in range(world_size) ] dist.all_gather(all_shape, shape_tensor) ## compute largest shapes all_shape = [x.cpu().numpy() for x in all_shape] all_count = [int(x.prod()) for x in all_shape] all_shape = [list(map(int, x)) for x in all_shape] max_count = max(all_count) ## padding tensors and gather them output_tensors = [ torch.Tensor(max_count).cuda() for i in range(world_size) ] padded_input_array = np.zeros(max_count) padded_input_array[:mycount] = input_array.reshape(-1) input_tensor = torch.Tensor(padded_input_array).cuda() dist.all_gather(output_tensors, input_tensor) ## unpadding gathered tensors padded_output = [x.cpu().numpy() for x in output_tensors] output = [ x[:all_count[i]].reshape(all_shape[i]) for i, x in enumerate(padded_output) ] return output def gather_tensors_batch(input_array, part_size=100, ret_rank=-1): # batch-wize gathering to avoid CUDA out of memory rank = dist.get_rank() all_features = [] part_num = input_array.shape[0] // part_size + 1 if input_array.shape[ 0] % part_size != 0 else input_array.shape[0] // part_size for i in range(part_num): part_feat = input_array[i * part_size:min((i + 1) * part_size, input_array.shape[0]), ...] assert part_feat.shape[ 0] > 0, "rank: {}, length of part features should > 0".format(rank) #print("rank: {}, gather part: {}/{}, length: {}".format(rank, i, part_num, len(part_feat))) gather_part_feat = gather_tensors(part_feat) all_features.append(gather_part_feat) if ret_rank == -1: all_features = [ np.concatenate([all_features[i][j] for i in range(part_num)], axis=0) for j in range(len(all_features[0])) ] return all_features else: if rank == ret_rank: all_features = [ np.concatenate([all_features[i][j] for i in range(part_num)], axis=0) for j in range(len(all_features[0])) ] return all_features else: return None
2,629
36.571429
100
py
Few-shot-WSI
Few-shot-WSI-master/openselfsup/utils/collect_env.py
import os.path as osp import subprocess import sys from collections import defaultdict import cv2 import mmcv import torch import torchvision import openselfsup def collect_env(): """Collect the information of the running environments.""" env_info = {} env_info['sys.platform'] = sys.platform env_info['Python'] = sys.version.replace('\n', '') cuda_available = torch.cuda.is_available() env_info['CUDA available'] = cuda_available if cuda_available: from torch.utils.cpp_extension import CUDA_HOME env_info['CUDA_HOME'] = CUDA_HOME if CUDA_HOME is not None and osp.isdir(CUDA_HOME): try: nvcc = osp.join(CUDA_HOME, 'bin/nvcc') nvcc = subprocess.check_output( '"{}" -V | tail -n1'.format(nvcc), shell=True) nvcc = nvcc.decode('utf-8').strip() except subprocess.SubprocessError: nvcc = 'Not Available' env_info['NVCC'] = nvcc devices = defaultdict(list) for k in range(torch.cuda.device_count()): devices[torch.cuda.get_device_name(k)].append(str(k)) for name, devids in devices.items(): env_info['GPU ' + ','.join(devids)] = name gcc = subprocess.check_output('gcc --version | head -n1', shell=True) gcc = gcc.decode('utf-8').strip() env_info['GCC'] = gcc env_info['PyTorch'] = torch.__version__ env_info['PyTorch compiling details'] = torch.__config__.show() env_info['TorchVision'] = torchvision.__version__ env_info['OpenCV'] = cv2.__version__ env_info['MMCV'] = mmcv.__version__ env_info['OpenSelfSup'] = openselfsup.__version__ #from openselfsup.ops import get_compiler_version, get_compiling_cuda_version #env_info['OpenSelfSup Compiler'] = get_compiler_version() #env_info['OpenSelfSup CUDA Compiler'] = get_compiling_cuda_version() return env_info if __name__ == "__main__": for name, val in collect_env().items(): print('{}: {}'.format(name, val))
2,055
30.630769
81
py
Few-shot-WSI
Few-shot-WSI-master/openselfsup/utils/flops_counter.py
# Modified from flops-counter.pytorch by Vladislav Sovrasov # original repo: https://github.com/sovrasov/flops-counter.pytorch # MIT License # Copyright (c) 2018 Vladislav Sovrasov # Permission is hereby granted, free of charge, to any person obtaining a copy # of this software and associated documentation files (the "Software"), to deal # in the Software without restriction, including without limitation the rights # to use, copy, modify, merge, publish, distribute, sublicense, and/or sell # copies of the Software, and to permit persons to whom the Software is # furnished to do so, subject to the following conditions: # The above copyright notice and this permission notice shall be included in # all copies or substantial portions of the Software. # THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR # IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, # FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE # AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER # LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, # OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE # SOFTWARE. import sys import numpy as np import torch import torch.nn as nn from torch.nn.modules.batchnorm import _BatchNorm from torch.nn.modules.conv import _ConvNd, _ConvTransposeMixin from torch.nn.modules.pooling import (_AdaptiveAvgPoolNd, _AdaptiveMaxPoolNd, _AvgPoolNd, _MaxPoolNd) def get_model_complexity_info(model, input_res, print_per_layer_stat=True, as_strings=True, input_constructor=None, ost=sys.stdout): assert type(input_res) is tuple assert len(input_res) >= 2 flops_model = add_flops_counting_methods(model) flops_model.eval().start_flops_count() if input_constructor: input = input_constructor(input_res) _ = flops_model(**input) else: batch = torch.ones(()).new_empty( (1, *input_res), dtype=next(flops_model.parameters()).dtype, device=next(flops_model.parameters()).device) flops_model(batch) if print_per_layer_stat: print_model_with_flops(flops_model, ost=ost) flops_count = flops_model.compute_average_flops_cost() params_count = get_model_parameters_number(flops_model) flops_model.stop_flops_count() if as_strings: return flops_to_string(flops_count), params_to_string(params_count) return flops_count, params_count def flops_to_string(flops, units='GMac', precision=2): if units is None: if flops // 10**9 > 0: return str(round(flops / 10.**9, precision)) + ' GMac' elif flops // 10**6 > 0: return str(round(flops / 10.**6, precision)) + ' MMac' elif flops // 10**3 > 0: return str(round(flops / 10.**3, precision)) + ' KMac' else: return str(flops) + ' Mac' else: if units == 'GMac': return str(round(flops / 10.**9, precision)) + ' ' + units elif units == 'MMac': return str(round(flops / 10.**6, precision)) + ' ' + units elif units == 'KMac': return str(round(flops / 10.**3, precision)) + ' ' + units else: return str(flops) + ' Mac' def params_to_string(params_num): """converting number to string :param float params_num: number :returns str: number >>> params_to_string(1e9) '1000.0 M' >>> params_to_string(2e5) '200.0 k' >>> params_to_string(3e-9) '3e-09' """ if params_num // 10**6 > 0: return str(round(params_num / 10**6, 2)) + ' M' elif params_num // 10**3: return str(round(params_num / 10**3, 2)) + ' k' else: return str(params_num) def print_model_with_flops(model, units='GMac', precision=3, ost=sys.stdout): total_flops = model.compute_average_flops_cost() def accumulate_flops(self): if is_supported_instance(self): return self.__flops__ / model.__batch_counter__ else: sum = 0 for m in self.children(): sum += m.accumulate_flops() return sum def flops_repr(self): accumulated_flops_cost = self.accumulate_flops() return ', '.join([ flops_to_string( accumulated_flops_cost, units=units, precision=precision), '{:.3%} MACs'.format(accumulated_flops_cost / total_flops), self.original_extra_repr() ]) def add_extra_repr(m): m.accumulate_flops = accumulate_flops.__get__(m) flops_extra_repr = flops_repr.__get__(m) if m.extra_repr != flops_extra_repr: m.original_extra_repr = m.extra_repr m.extra_repr = flops_extra_repr assert m.extra_repr != m.original_extra_repr def del_extra_repr(m): if hasattr(m, 'original_extra_repr'): m.extra_repr = m.original_extra_repr del m.original_extra_repr if hasattr(m, 'accumulate_flops'): del m.accumulate_flops model.apply(add_extra_repr) print(model, file=ost) model.apply(del_extra_repr) def get_model_parameters_number(model): params_num = sum(p.numel() for p in model.parameters() if p.requires_grad) return params_num def add_flops_counting_methods(net_main_module): # adding additional methods to the existing module object, # this is done this way so that each function has access to self object net_main_module.start_flops_count = start_flops_count.__get__( net_main_module) net_main_module.stop_flops_count = stop_flops_count.__get__( net_main_module) net_main_module.reset_flops_count = reset_flops_count.__get__( net_main_module) net_main_module.compute_average_flops_cost = \ compute_average_flops_cost.__get__(net_main_module) net_main_module.reset_flops_count() # Adding variables necessary for masked flops computation net_main_module.apply(add_flops_mask_variable_or_reset) return net_main_module def compute_average_flops_cost(self): """ A method that will be available after add_flops_counting_methods() is called on a desired net object. Returns current mean flops consumption per image. """ batches_count = self.__batch_counter__ flops_sum = 0 for module in self.modules(): if is_supported_instance(module): flops_sum += module.__flops__ return flops_sum / batches_count def start_flops_count(self): """ A method that will be available after add_flops_counting_methods() is called on a desired net object. Activates the computation of mean flops consumption per image. Call it before you run the network. """ add_batch_counter_hook_function(self) self.apply(add_flops_counter_hook_function) def stop_flops_count(self): """ A method that will be available after add_flops_counting_methods() is called on a desired net object. Stops computing the mean flops consumption per image. Call whenever you want to pause the computation. """ remove_batch_counter_hook_function(self) self.apply(remove_flops_counter_hook_function) def reset_flops_count(self): """ A method that will be available after add_flops_counting_methods() is called on a desired net object. Resets statistics computed so far. """ add_batch_counter_variables_or_reset(self) self.apply(add_flops_counter_variable_or_reset) def add_flops_mask(module, mask): def add_flops_mask_func(module): if isinstance(module, torch.nn.Conv2d): module.__mask__ = mask module.apply(add_flops_mask_func) def remove_flops_mask(module): module.apply(add_flops_mask_variable_or_reset) def is_supported_instance(module): for mod in hook_mapping: if issubclass(type(module), mod): return True return False def empty_flops_counter_hook(module, input, output): module.__flops__ += 0 def upsample_flops_counter_hook(module, input, output): output_size = output[0] batch_size = output_size.shape[0] output_elements_count = batch_size for val in output_size.shape[1:]: output_elements_count *= val module.__flops__ += int(output_elements_count) def relu_flops_counter_hook(module, input, output): active_elements_count = output.numel() module.__flops__ += int(active_elements_count) def linear_flops_counter_hook(module, input, output): input = input[0] batch_size = input.shape[0] module.__flops__ += int(batch_size * input.shape[1] * output.shape[1]) def pool_flops_counter_hook(module, input, output): input = input[0] module.__flops__ += int(np.prod(input.shape)) def bn_flops_counter_hook(module, input, output): input = input[0] batch_flops = np.prod(input.shape) if module.affine: batch_flops *= 2 module.__flops__ += int(batch_flops) def gn_flops_counter_hook(module, input, output): elems = np.prod(input[0].shape) # there is no precise FLOPs estimation of computing mean and variance, # and we just set it 2 * elems: half muladds for computing # means and half for computing vars batch_flops = 3 * elems if module.affine: batch_flops += elems module.__flops__ += int(batch_flops) def deconv_flops_counter_hook(conv_module, input, output): # Can have multiple inputs, getting the first one input = input[0] batch_size = input.shape[0] input_height, input_width = input.shape[2:] kernel_height, kernel_width = conv_module.kernel_size in_channels = conv_module.in_channels out_channels = conv_module.out_channels groups = conv_module.groups filters_per_channel = out_channels // groups conv_per_position_flops = ( kernel_height * kernel_width * in_channels * filters_per_channel) active_elements_count = batch_size * input_height * input_width overall_conv_flops = conv_per_position_flops * active_elements_count bias_flops = 0 if conv_module.bias is not None: output_height, output_width = output.shape[2:] bias_flops = out_channels * batch_size * output_height * output_height overall_flops = overall_conv_flops + bias_flops conv_module.__flops__ += int(overall_flops) def conv_flops_counter_hook(conv_module, input, output): # Can have multiple inputs, getting the first one input = input[0] batch_size = input.shape[0] output_dims = list(output.shape[2:]) kernel_dims = list(conv_module.kernel_size) in_channels = conv_module.in_channels out_channels = conv_module.out_channels groups = conv_module.groups filters_per_channel = out_channels // groups conv_per_position_flops = np.prod( kernel_dims) * in_channels * filters_per_channel active_elements_count = batch_size * np.prod(output_dims) if conv_module.__mask__ is not None: # (b, 1, h, w) output_height, output_width = output.shape[2:] flops_mask = conv_module.__mask__.expand(batch_size, 1, output_height, output_width) active_elements_count = flops_mask.sum() overall_conv_flops = conv_per_position_flops * active_elements_count bias_flops = 0 if conv_module.bias is not None: bias_flops = out_channels * active_elements_count overall_flops = overall_conv_flops + bias_flops conv_module.__flops__ += int(overall_flops) hook_mapping = { # conv _ConvNd: conv_flops_counter_hook, # deconv _ConvTransposeMixin: deconv_flops_counter_hook, # fc nn.Linear: linear_flops_counter_hook, # pooling _AvgPoolNd: pool_flops_counter_hook, _MaxPoolNd: pool_flops_counter_hook, _AdaptiveAvgPoolNd: pool_flops_counter_hook, _AdaptiveMaxPoolNd: pool_flops_counter_hook, # activation nn.ReLU: relu_flops_counter_hook, nn.PReLU: relu_flops_counter_hook, nn.ELU: relu_flops_counter_hook, nn.LeakyReLU: relu_flops_counter_hook, nn.ReLU6: relu_flops_counter_hook, # normalization _BatchNorm: bn_flops_counter_hook, nn.GroupNorm: gn_flops_counter_hook, # upsample nn.Upsample: upsample_flops_counter_hook, } def batch_counter_hook(module, input, output): batch_size = 1 if len(input) > 0: # Can have multiple inputs, getting the first one input = input[0] batch_size = len(input) else: print('Warning! No positional inputs found for a module, ' 'assuming batch size is 1.') module.__batch_counter__ += batch_size def add_batch_counter_variables_or_reset(module): module.__batch_counter__ = 0 def add_batch_counter_hook_function(module): if hasattr(module, '__batch_counter_handle__'): return handle = module.register_forward_hook(batch_counter_hook) module.__batch_counter_handle__ = handle def remove_batch_counter_hook_function(module): if hasattr(module, '__batch_counter_handle__'): module.__batch_counter_handle__.remove() del module.__batch_counter_handle__ def add_flops_counter_variable_or_reset(module): if is_supported_instance(module): module.__flops__ = 0 def add_flops_counter_hook_function(module): if is_supported_instance(module): if hasattr(module, '__flops_handle__'): return for mod_type, counter_hook in hook_mapping.items(): if issubclass(type(module), mod_type): handle = module.register_forward_hook(counter_hook) break module.__flops_handle__ = handle def remove_flops_counter_hook_function(module): if is_supported_instance(module): if hasattr(module, '__flops_handle__'): module.__flops_handle__.remove() del module.__flops_handle__ # --- Masked flops counting # Also being run in the initialization def add_flops_mask_variable_or_reset(module): if is_supported_instance(module): module.__mask__ = None
14,304
31.146067
79
py
Few-shot-WSI
Few-shot-WSI-master/openselfsup/utils/__init__.py
from .alias_multinomial import AliasMethod from .collect import nondist_forward_collect, dist_forward_collect from .collect_env import collect_env from .config_tools import traverse_replace from .flops_counter import get_model_complexity_info from .logger import get_root_logger, print_log from .registry import Registry, build_from_cfg from . import optimizers
362
39.333333
66
py
Few-shot-WSI
Few-shot-WSI-master/openselfsup/utils/config_tools.py
from mmcv import Config def traverse_replace(d, key, value): if isinstance(d, (dict, Config)): for k, v in d.items(): if k == key: d[k] = value else: traverse_replace(v, key, value) elif isinstance(d, (list, tuple, set)): for v in d: traverse_replace(v, key, value)
359
26.692308
47
py
rivuletpy
rivuletpy-master/build.py
""" Build C extensions. Adapted from: https://github.com/zoj613/htnorm/blob/main/build.py """ import os from distutils.core import Extension import numpy as np source_files = [ "rivuletpy/msfm/msfmmodule.c", "rivuletpy/msfm/_msfm.c", ] # get environmental variables to determine the flow of the build process BUILD_WHEELS = os.getenv("BUILD_WHEELS", None) LIBS_DIR = os.getenv("LIBS_DIR", "/usr/lib") libraries = ["m"] # when building manylinux2014 wheels for pypi use different directories as # required by CentOS, else allow the user to specify them when building from # source distribution if BUILD_WHEELS: library_dirs = ["/usr/lib64"] libraries.append("openblas") else: library_dirs = [LIBS_DIR] libraries.extend(["blas", "lapack"]) extensions = [ Extension( "msfm", source_files, include_dirs=[np.get_include(), "rivuletpy/msfm"], library_dirs=library_dirs, libraries=libraries, define_macros=[("NPY_NO_DEPRECATED_API", 0)], extra_compile_args=["-std=c99"], ) ] def build(setup_kwargs): """Build extension modules.""" kwargs = {"ext_modules": extensions, "zip_safe": False} setup_kwargs.update(kwargs)
1,216
24.354167
76
py
rivuletpy
rivuletpy-master/tests/test_riveal.py
from filtering.riveal import riveal from rivuletpy.utils.io import * from filtering.thresholding import rescale img = loadimg('tests/data/test.tif') dtype = img.dtype swc = loadswc('tests/data/test.swc') img = riveal(img, swc, nsample=5e4, epoch=30) img = rescale(img) try: from skimage import filters except ImportError: from skimage import filter as filters threshold = filters.threshold_otsu(img) img[img<=threshold] = 0 writetiff3d('dt.tif', img.astype(dtype)) my_env = os.environ.copy() v3dcmd = "%s/vaa3d" % my_env['V3DPATH'] v3dcmd += ' -v -i dt.tif' os.system(v3dcmd)
585
26.904762
45
py
rivuletpy
rivuletpy-master/tests/testbgrsp.py
from filtering.anisotropic import * from rivuletpy.utils.io import * import matplotlib.pyplot as plt from scipy import io as sio try: from skimage import filters except ImportError: from skimage import filter as filters from scipy.ndimage.filters import gaussian_filter mat = sio.loadmat('tests/data/very-small-oof.mat', ) img = mat['img'] ostu_img = 0. radii = np.arange(0.2, 2, 0.5) rho = 0.5 oof_matlab = mat['oof'] ostu_matlaboof = filters.threshold_otsu(oof_matlab) rps, _ = response(img.astype('float'), rsptype='bg', radii=radii, rho=rho) thr = 1 from scipy import ndimage as ndi from skimage import feature canny = feature.canny(rps, sigma=3) smoothed_rps = gaussian_filter(rps, 0.5) # ostu_smooth = filters.threshold_otsu(smoothed_rps) ostu_smooth = 1 plotidx = 1 plt.subplot(4, 4, plotidx) plt.imshow(rps.max(axis=0)) plt.title('OOF Python MEM_SAVE YZ') plotidx += 1 plt.subplot(4, 4, plotidx) plt.imshow(rps.max(axis=1)) plt.title('OOF Python MEM_SAVE XZ') plotidx += 1 plt.subplot(4, 4, plotidx) plt.imshow(rps.max(axis=2)) plt.title('OOF Python MEM_SAVE XY') plotidx += 1 plt.subplot(4, 4, plotidx) plt.imshow((rps > thr).max(axis=2)) plt.title('OOF Python MEM_SAVE Otsu XY') plotidx += 1 plt.subplot(4, 4, plotidx) plt.imshow(smoothed_rps.max(axis=0)) plt.title('Smooth YZ') plotidx += 1 plt.subplot(4, 4, plotidx) plt.imshow(smoothed_rps.max(axis=1)) plt.title('Smooth XZ') plotidx += 1 plt.subplot(4, 4, plotidx) plt.imshow(smoothed_rps.max(axis=2)) plt.title('Smooth XY') plotidx += 1 plt.subplot(4, 4, plotidx) plt.imshow((smoothed_rps > ostu_smooth).max(axis=2)) plt.title('Smooth XY') plotidx +=1 plt.subplot(4, 4, plotidx) plt.imshow(canny.max(axis=0)) plt.title('OOF Matlab YZ') plotidx += 1 plt.subplot(4, 4, plotidx) plt.imshow(canny.max(axis=1)) plt.title('OOF Matlab XZ') plotidx += 1 plt.subplot(4, 4, plotidx) plt.imshow(canny.max(axis=2)) plt.title('OOF Matlab XY') plotidx += 1 plt.subplot(4, 4, plotidx) plt.imshow((canny > ostu_matlaboof).max(axis=2)) plt.title('OOF Matlab Otsu XY') plotidx += 1 plt.subplot(4, 4, plotidx) plt.imshow(img.max(axis=0)) plt.title('Original YZ') plotidx += 1 plt.subplot(4, 4, plotidx) plt.imshow(img.max(axis=1)) plt.title('Original XZ') plotidx += 1 plt.subplot(4, 4, plotidx) plt.imshow(img.max(axis=2)) plt.title('Original XY') plotidx += 1 plt.subplot(4, 4, plotidx) plt.imshow((img > ostu_img).max(axis=2)) plt.title('Original Otsu XY') plt.show()
2,448
20.866071
74
py
rivuletpy
rivuletpy-master/tests/testmsfm.py
import msfm from rivuletpy.utils.io import * import skfmm import os from matplotlib import pyplot as plt dir_path = os.path.dirname(os.path.realpath(__file__)) img = loadimg(os.path.join(dir_path, 'data/test.tif')) dt = skfmm.distance(img > 0, dx=1) # Boundary DT somaradius = dt.max() somapos = np.asarray(np.unravel_index(dt.argmax(), dt.shape)) print('somapos in python', somapos, somapos.dtype) print('dt[soma pos] in python: %f' % dt[somapos[0], somapos[1], somapos[2]]) print('Running MSFM...') T = msfm.run(dt, somapos, False, True) plt.imshow(np.squeeze(T.min(axis=-1))) plt.show()
592
30.210526
76
py
rivuletpy
rivuletpy-master/tests/test_fuzzy_threshold.py
from os import path from rivuletpy.utils.io import * from filtering.thresholding import fuzzy import matplotlib.pyplot as plt img = loadimg(path.join('tests', 'data', 'test.tif')) thr = fuzzy(img, render=True)
213
22.777778
53
py
rivuletpy
rivuletpy-master/tests/testssm.py
from filtering.morphology import ssm from rivuletpy.utils.io import * import matplotlib.pyplot as plt import skfmm ITER = 30 img = loadimg('/home/siqi/ncidata/rivuletpy/tests/data/test-crop.tif') bimg = (img > 0).astype('int') dt = skfmm.distance(bimg, dx=1) sdt = ssm(dt, anisotropic=True, iterations=ITER) try: from skimage import filters except ImportError: from skimage import filter as filters s_seg = s > filters.threshold_otsu(s) plt.figure() plt.title('DT') plt.imshow(dt.max(-1)) plt.figure() plt.title('img > 0') plt.imshow((img > 0).max(-1)) plt.figure() plt.title('SSM-DT') plt.imshow(sdt.max(-1))
624
20.551724
70
py
rivuletpy
rivuletpy-master/tests/testmetrics.py
from rivuletpy.utils.metrics import * from rivuletpy.utils.io import * from os.path import join datapath = 'tests/data' swc1 = loadswc(join(datapath, 'test-output.swc')) swc2 = loadswc(join(datapath, 'test-expected.swc')) prf, swc_compare = precision_recall(swc1, swc2) print('Precision: %.2f\tRecall: %.2f\tF1: %.2f\t' % prf) M1, M2 = gaussian_distance(swc1, swc2, 3.0) print('M1 MEAN: %.2f\tM2 MEAN: %.2f' % (M1.mean(), M2.mean())) midx1, midx2 = connectivity_distance(swc1, swc2) for i in midx1: swc1[i, 1] = 2 swc1[i, 5] = 4 saveswc(join(datapath, 'test.connect1.swc'), swc1) for i in midx2: swc2[i, 1] = 2 swc2[i, 5] = 4 saveswc(join(datapath, 'test.connect2.swc'), swc2)
700
27.04
62
py
rivuletpy
rivuletpy-master/tests/testoof.py
from filtering.anisotropic import * from rivuletpy.utils.io import * import matplotlib.pyplot as plt from scipy import io as sio try: from skimage import filters except ImportError: from skimage import filter as filters mat = sio.loadmat('tests/data/very-small-oof.mat', ) img = mat['img'] ostu_img = filters.threshold_otsu(img) radii = np.arange(1, 1.5, 0.1) oof_matlab = mat['oof'] ostu_matlaboof = filters.threshold_otsu(oof_matlab) oofrps_memsave, _ = response(img.astype('float'), rsptype='oof', radii=np.asarray(radii), memory_save=True) otsu_memsave = filters.threshold_otsu(oofrps_memsave) oofrps_highmem, _ = response(img.astype('float'), rsptype='oof', radii=np.asarray(radii), memory_save=False) otsu_highmem = filters.threshold_otsu(oofrps_highmem) plotidx = 1 plt.subplot(4, 4, plotidx) plt.imshow(oofrps_memsave.max(axis=0)) plt.title('OOF Python MEM_SAVE YZ') plotidx += 1 plt.subplot(4, 4, plotidx) plt.imshow(oofrps_memsave.max(axis=1)) plt.title('OOF Python MEM_SAVE XZ') plotidx += 1 plt.subplot(4, 4, plotidx) plt.imshow(oofrps_memsave.max(axis=2)) plt.title('OOF Python MEM_SAVE XY') plotidx += 1 plt.subplot(4, 4, plotidx) plt.imshow((oofrps_memsave > otsu_memsave).max(axis=2)) plt.title('OOF Python MEM_SAVE Otsu XY') plotidx += 1 plt.subplot(4, 4, plotidx) plt.imshow(oofrps_highmem.max(axis=0)) plt.title('OOF Python HIGHMEM YZ') plotidx += 1 plt.subplot(4, 4, plotidx) plt.imshow(oofrps_highmem.max(axis=1)) plt.title('OOF Python HIGHMEM XZ') plotidx += 1 plt.subplot(4, 4, plotidx) plt.imshow(oofrps_highmem.max(axis=2)) plt.title('OOF Python HIGHMEM XY') plotidx += 1 plt.subplot(4, 4, plotidx) plt.imshow((oofrps_highmem > otsu_highmem).max(axis=2)) plt.title('OOF Python HIGHMEM Otsu XY') plotidx += 1 plt.subplot(4, 4, plotidx) plt.imshow(oof_matlab.max(axis=0)) plt.title('OOF Matlab YZ') plotidx += 1 plt.subplot(4, 4, plotidx) plt.imshow(oof_matlab.max(axis=1)) plt.title('OOF Matlab XZ') plotidx += 1 plt.subplot(4, 4, plotidx) plt.imshow(oof_matlab.max(axis=2)) plt.title('OOF Matlab XY') plotidx += 1 plt.subplot(4, 4, plotidx) plt.imshow((oof_matlab > ostu_matlaboof).max(axis=2)) plt.title('OOF Matlab Otsu XY') plotidx += 1 plt.subplot(4, 4, plotidx) plt.imshow(img.max(axis=0)) plt.title('Original YZ') plotidx += 1 plt.subplot(4, 4, plotidx) plt.imshow(img.max(axis=1)) plt.title('Original XZ') plotidx += 1 plt.subplot(4, 4, plotidx) plt.imshow(img.max(axis=2)) plt.title('Original XY') plotidx += 1 plt.subplot(4, 4, plotidx) plt.imshow((img > ostu_img).max(axis=2)) plt.title('Original Otsu XY') plt.show()
2,584
23.619048
108
py
rivuletpy
rivuletpy-master/tests/test_node_push.py
# Load swc import SimpleITK as sitk from rivuletpy.utils.io import loadswc, loadimg from rivuletpy.swc import SWC from rivuletpy.utils.io import swc2world, swc2vtk swc_mat = loadswc( '/home/z003s24h/Desktop/zhoubing_vessel_example/mask/Anonymous EJRH_16.r2.swc') s = SWC() s._data = swc_mat # Load image and binarize img = loadimg( '/home/z003s24h/Desktop/zhoubing_vessel_example/mask/Anonymous EJRH_16.mhd', 1) imgdtype = img.dtype imgshape = img.shape bimg = img > 0 s._data[:, 2] *= .7 / 1. s._data[:, 3] *= 0.363281 / 1. s._data[:, 4] *= 0.363281 / 1. s.push_nodes_with_binary(bimg) # s.view() print('Converting to world space...') mhd = sitk.ReadImage( '/home/z003s24h/Desktop/zhoubing_vessel_example/mask/Anonymous EJRH_16.mhd') swc = swc2world(s.get_array(), mhd.GetOrigin(), [1.] * 3) print('Saving to VTK format...') swc2vtk(swc, '/home/z003s24h/Desktop/zhoubing_vessel_example/mask/Anonymous EJRH_16.r2.pushed.vtk')
974
28.545455
99
py
rivuletpy
rivuletpy-master/tests/testbg.py
from filtering.anisotropic import * from rivuletpy.utils.io import * import matplotlib.pyplot as plt from scipy import io as sio try: from skimage import filters except ImportError: from skimage import filter as filters # plot the gaussian kernel nsig = 5 nmu = 5 kerlen = 101 kr = (kerlen - 1) / 2 X, Y, Z = np.meshgrid(np.arange(-kr, kr+1), np.arange(-kr, kr+1), np.arange(-kr, kr+1)) indstack = np.stack((X, Y, Z)) dist = np.linalg.norm(indstack, axis=0) plt.title('Gaussian') for i in range(nsig): for j in range(nmu): mu = float(j*10) sigma = float(i+4) k = gkern3(dist, mu, sigma) ax = plt.subplot(nsig, nmu, (i) * nsig + (j + 1)) ax.set_title('mu=%.2f, sigma=%.2f' % (mu, sigma)) imgplot = plt.imshow(k[:, :, int((kerlen-1)/2)]) plt.colorbar() plt.figure(2) plt.title('Bi-Gaussian') nsig = 5 nrho = 5 kerlen = 101 for i in range(nsig): for j in range(nrho): sigma = float(i * 10)+1 rho = (j+1) * 0.1 k = bgkern3(kerlen, 0, sigma, rho) ax = plt.subplot(nsig, nrho, (i) * nsig + (j + 1)) ax.set_title('sigma=%.2f, rho=%.2f' % (sigma, rho)) imgplot = plt.imshow(k[:, :, int((kerlen-1)/2) ]) plt.colorbar() plt.show()
1,287
24.76
59
py
rivuletpy
rivuletpy-master/tests/test_viewer.py
from rivuletpy.utils.io import loadswc from rivuletpy.swc import SWC swc_mat = loadswc('test_data/test.tif.r2.swc') s = SWC() s._data = swc_mat s.view() input("Press any key to continue...")
191
23
46
py
rivuletpy
rivuletpy-master/rivuletpy/soma.py
# -*- coding: utf-8 -*- """ somasnakes =========== Original package is adjusted for soma detection by donghaozhang and siqiliu. This soma submodule can be used for soma detection only, but this submodule is currently embedded in rivuletpy. The soma mask can be generate by setting its corresponding argument. Soma detection requires an initial soma centroid, estimated somatic radius and grayscale neuron image. Soma growth is based on the Morphological Active Contours without Edges algorithm. The original paper is named as A morphological approach to curvature-based evolution of curves and surfaces.The following papers are Rivulet papers. The soma growth algorithm can converge by applying the sliding window. Journal Rivulet Paper : Rivulet: 3D Neuron Morphology Tracing with Iterative Back-Tracking Conference Rivulet Paper : Reconstruction of 3D neuron morphology using Rivulet back-tracking soma is a submodule of rivuletpy """ __author__ = "Donghao Zhang <zdhpeter1991@gmail.com>, Siqi Liu <lsqshr@gmail.com>" from itertools import cycle import math import numpy as np from scipy.ndimage import binary_dilation, binary_erosion from scipy.ndimage import gaussian_filter, gaussian_gradient_magnitude from scipy.ndimage.measurements import center_of_mass from scipy.ndimage.morphology import generate_binary_structure from rivuletpy.utils.io import writetiff3d import skfmm class Soma(object): def __init__(self): self.centroid = None self.radius = 0 self.mask = None def simple_mask(self, bimg): ''' Make soma binary mask with the original binary image and its radius and position ''' # Make a ball like mask with 2 X somaradius ballvolume = np.zeros(bimg.shape) ballvolume[self.centroid[0], self.centroid[1], self.centroid[2]] = 1 stt = generate_binary_structure(3, 1) for i in range(math.ceil(self.radius * 2.5)): ballvolume = binary_dilation(ballvolume, structure=stt) # Make the soma mask with the intersection # between the ball area and the original binary self.mask = np.logical_and(ballvolume, bimg) # Shift the centroid according to the cropped region def crop_centroid(self, crop_region): self.centroid[0] = self.centroid[0] - crop_region[0, 0] self.centroid[1] = self.centroid[1] - crop_region[1, 0] self.centroid[2] = self.centroid[2] - crop_region[2, 0] def detect(self, bimg, simple=False, silent=False): """ Automatic detection of soma volume unless the iterations are given. """ # Smooth iterations smoothing = 1 # A float number controls the weight of internal energy lambda1 = 1 # A float number controls the weight of external energy lambda2 = 1.5 # Manually set the number of iterations required for the soma # The type of iterations is int iterations = -1 bimg = bimg.astype('int') # Segment dt = skfmm.distance(bimg, dx=1.1) # Boundary DT # somaradius : the approximate value of # soma radius estimated from distance transform # the type of somaradius is float64 # somaradius is just a float number somaradius = dt.max() # somapos : the coordinate of estimated soma centroid # the type of somapos is int64 # the shape of somapos is (3,) # somapos is array-like somapos = np.asarray(np.unravel_index(dt.argmax(), dt.shape)) # Soma detection is required if not simple: if not silent: print('Reconstructing Soma with SRET') ratioxz = bimg.shape[0] / bimg.shape[2] ratioyz = bimg.shape[1] / bimg.shape[2] sqrval = (somaradius**0.5 * max(ratioxz, ratioyz)) sqrval = np.floor(min(max(sqrval, 3), (somaradius**0.5) * 6)) startpt = somapos - 3 * sqrval endpt = somapos + 3 * sqrval # # To constrain the soma growth region inside the cubic region # # Python index start from 0 startpt[0] = min(max(0, startpt[0]), bimg.shape[0] - 1) startpt[1] = min(max(0, startpt[1]), bimg.shape[1] - 1) startpt[2] = min(max(0, startpt[2]), bimg.shape[2] - 1) endpt[0] = min(max(0, endpt[0]), bimg.shape[0] - 1) endpt[1] = min(max(0, endpt[1]), bimg.shape[1] - 1) endpt[2] = min(max(0, endpt[2]), bimg.shape[2] - 1) startpt = startpt.astype(int) # Convert type to int for indexing endpt = endpt.astype(int) # # Extract soma region for fast soma detection somaimg = bimg[startpt[0]:endpt[0], startpt[1]:endpt[1], startpt[2]: endpt[2]] centerpt = np.zeros(3) centerpt[0] = somaimg.shape[0] / 2 centerpt[1] = somaimg.shape[1] / 2 centerpt[2] = somaimg.shape[2] / 2 centerpt = np.floor(centerpt) # Morphological ACWE. Initialization of the level-set. macwe = MorphACWE(somaimg, startpt, endpt, smoothing, lambda1, lambda2) macwe.levelset = circle_levelset(somaimg.shape, np.floor(centerpt), sqrval) # -1 means the automatic detection # Positive integers means the number of iterations if iterations == -1: macwe.autoconvg() # automatic soma detection else: # Input the iteration number manually for i in range(iterations): macwe.step() # The following achieves the automatic somtic box extension # The maximum somatic region extension iteration # It is set to 10 avoid infinite loops for i in range(1, 11): # if not silent: # print('The somatic region extension iteration is', i) if macwe.enlrspt is None: break # Copy the values to new variables for the safe purpose startpt = macwe.enlrspt.copy() endpt = macwe.enlrept.copy() startpt[0] = min(max(0, startpt[0]), bimg.shape[0]) startpt[1] = min(max(0, startpt[1]), bimg.shape[1]) startpt[2] = min(max(0, startpt[2]), bimg.shape[2]) endpt[0] = min(max(0, endpt[0]), bimg.shape[0]) endpt[1] = min(max(0, endpt[1]), bimg.shape[1]) endpt[2] = min(max(0, endpt[2]), bimg.shape[2]) somaimg = bimg[startpt[0]:endpt[0], startpt[1]:endpt[1], startpt[2]: endpt[2]] full_soma_mask = np.zeros( (bimg.shape[0], bimg.shape[1], bimg.shape[2])) # Put the detected somas into the whole image # It is either true or false full_soma_mask[macwe.startpoint[0]:macwe.endpoint[ 0], macwe.startpoint[1]:macwe.endpoint[1], macwe.startpoint[2]: macwe.endpoint[2]] = macwe._u # The newlevelset is the initial soma volume from previous iteration #(the automatic converge operation) newlevelset = full_soma_mask[startpt[0]:endpt[0], startpt[1]:endpt[1], startpt[2]:endpt[2]] # The previous macwe class is released # To avoid the conflicts with the new initialisation of the # macwe class del macwe # Initialisation for the new class macwe = MorphACWE(somaimg, startpt, endpt, smoothing, lambda1, lambda2) del somaimg, full_soma_mask, startpt, endpt # Reuse the soma volume from previous iteration macwe.set_levelset(newlevelset) # Release memory to avoid conflicts with previous newlevelset del newlevelset macwe.autoconvg() # The automatic smoothing operation to remove the interferes with # dendrites macwe.autosmooth() # Initialise soma mask image full_soma_mask = np.zeros( (bimg.shape[0], bimg.shape[1], bimg.shape[2])) # There are two possible scenarios # The first scenrio is that the automatic box extension is not # necessary if macwe.enlrspt is None: startpt = macwe.startpoint.copy() endpt = macwe.endpoint.copy() # The second scenrio is that the automatic box extension operations # has been performed else: startpt = macwe.enlrspt.copy() endpt = macwe.enlrept.copy() startpt[0] = min(max(0, startpt[0]), bimg.shape[0]) startpt[1] = min(max(0, startpt[1]), bimg.shape[1]) startpt[2] = min(max(0, startpt[2]), bimg.shape[2]) endpt[0] = min(max(0, endpt[0]), bimg.shape[0]) endpt[1] = min(max(0, endpt[1]), bimg.shape[1]) endpt[2] = min(max(0, endpt[2]), bimg.shape[2]) # The soma mask image contains only two possible values # Each element is either 0 or 40 # Value 40 is assigned for the visualisation purpose. full_soma_mask[startpt[0]:endpt[0], startpt[1]:endpt[1], startpt[2]:endpt[ 2]] = macwe._u > 0 # Calculate the new centroid using the soma volume newsomapos = center_of_mass(full_soma_mask) # Round the float coordinates into integers newsomapos = [math.floor(p) for p in newsomapos] self.centroid = newsomapos self.radius = somaradius self.mask = full_soma_mask else: if not silent: print('Reconstructing Soma with Simple Mask') self.centroid = somapos self.radius = somaradius self.simple_mask(bimg) def pad(self, crop_region, original_shape): xmin = crop_region[0, 0] ymin = crop_region[1, 0] zmin = crop_region[2, 0] xmax = crop_region[0, 1] ymax = crop_region[1, 1] zmax = crop_region[2, 1] self.mask = np.pad(self.mask, ((xmin, original_shape[0] - xmax), (ymin, original_shape[1] - ymax), (zmin, original_shape[2] - zmax)), mode='constant', constant_values=0) def save(self, fname): writetiff3d(fname, self.mask * 255) class Fcycle(object): def __init__(self, iterable): """Call functions from the iterable each time it is called.""" self.funcs = cycle(iterable) def __call__(self, *args, **kwargs): f = next(self.funcs) return f(*args, **kwargs) # SI and IS operators for 2D and 3D. _P2 = [ np.eye(3), np.array([[0, 1, 0]] * 3), np.flipud(np.eye(3)), np.rot90([[0, 1, 0]] * 3) ] _P3 = [np.zeros((3, 3, 3)) for i in range(9)] _P3[0][:, :, 1] = 1 _P3[1][:, 1, :] = 1 _P3[2][1, :, :] = 1 _P3[3][:, [0, 1, 2], [0, 1, 2]] = 1 _P3[4][:, [0, 1, 2], [2, 1, 0]] = 1 _P3[5][[0, 1, 2], :, [0, 1, 2]] = 1 _P3[6][[0, 1, 2], :, [2, 1, 0]] = 1 _P3[7][[0, 1, 2], [0, 1, 2], :] = 1 _P3[8][[0, 1, 2], [2, 1, 0], :] = 1 _aux = np.zeros((0)) def SI(u): """SI operator.""" # print('SI operator has been called') global _aux if np.ndim(u) == 2: P = _P2 elif np.ndim(u) == 3: P = _P3 else: raise ValueError( "u has an invalid number of dimensions (should be 2 or 3)") if u.shape != _aux.shape[1:]: _aux = np.zeros((len(P), ) + u.shape) for i in range(len(P)): _aux[i] = binary_erosion(u, P[i]) return _aux.max(0) def circle_levelset(shape, center, sqradius, scalerow=1.0): """Build a binary function with a circle as the 0.5-levelset.""" grid = np.mgrid[list(map(slice, shape))].T - center phi = sqradius - np.sqrt(np.sum((grid.T)**2, 0)) u = np.float_(phi > 0) return u def IS(u): """IS operator.""" global _aux if np.ndim(u) == 2: P = _P2 elif np.ndim(u) == 3: P = _P3 else: raise ValueError( "u has an invalid number of dimensions (should be 2 or 3)") if u.shape != _aux.shape[1:]: _aux = np.zeros((len(P), ) + u.shape) for i in range(len(P)): _aux[i] = binary_dilation(u, P[i]) return _aux.min(0) # SIoIS operator. SIoIS = lambda u: SI(IS(u)) ISoSI = lambda u: IS(SI(u)) curvop = Fcycle([SIoIS, ISoSI]) # Stopping factors (function g(I) in the paper). def gborders(img, alpha=1.0, sigma=1.0): """Stopping criterion for image borders.""" # The norm of the gradient. gradnorm = gaussian_gradient_magnitude(img, sigma, mode='constant') return 1.0 / np.sqrt(1.0 + alpha * gradnorm) def glines(img, sigma=1.0): """Stopping criterion for image black lines.""" return gaussian_filter(img, sigma) class MorphACWE(object): """Morphological ACWE based on the Chan-Vese energy functional.""" def __init__(self, data, startpoint, endpoint, imgshape, smoothing=1, lambda1=1, lambda2=1.5): """Create a Morphological ACWE solver. Parameters ---------- data : ndarray The image data. smoothing : scalar The number of repetitions of the smoothing step (the curv operator) in each iteration. In other terms, this is the strength of the smoothing. This is the parameter µ. lambda1, lambda2 : scalars Relative importance of the inside pixels (lambda1) against the outside pixels (lambda2). startpt, endpt : numpy int array startpt is the initial starting point of the somatic region endpt is the initial ending point of the somatic region """ self._u = None self.smoothing = smoothing self.lambda1 = lambda1 self.lambda2 = lambda2 self.imgshape = imgshape self.data = data self.startpoint = startpoint self.endpoint = endpoint self.enlrspt = None self.enlrept = None def set_levelset(self, u): self._u = np.double(u) self._u[u > 0] = 1 self._u[u <= 0] = 0 levelset = property( lambda self: self._u, set_levelset, doc="The level set embedding function (u).") def step(self): """Perform a single step of the morphological Chan-Vese evolution.""" # Assign attributes to local variables for convenience. u = self._u if u is None: raise ValueError( "the levelset function is not set (use set_levelset)") data = self.data # Determine c0 and c1. inside = u > 0 outside = u <= 0 c0 = data[outside].sum() / float(outside.sum()) c1 = data[inside].sum() / float(inside.sum()) # Image attachment. dres = np.array(np.gradient(u)) abs_dres = np.abs(dres).sum(0) #aux = abs_dres * (c0 - c1) * (c0 + c1 - 2*data) aux = abs_dres * (self.lambda1 * (data - c1)**2 - self.lambda2 * (data - c0)**2) res = np.copy(u) res[aux < 0] = 1 res[aux > 0] = 0 res = IS(res) # Smoothing. for i in range(self.smoothing): res = curvop(res) self._u = res def step_sm(self): """A smoothing step of the morphological Chan-Vese evolution.""" # Assign attributes to local variables for convenience. u = self._u if u is None: raise ValueError( "the levelset function is not set (use set_levelset)") res = np.copy(u) # Smoothing. res = curvop(res) self._u = res def run(self, iterations): """Run several iterations of the morphological Chan-Vese method.""" for i in range(iterations): self.step() def autoconvg(self): """Soma detection converges by itself.""" # Autoconvg is the abbreviation of automatic convergence iterations = 200 # The following vector is the number of foreground voxels foreground_num = np.zeros(iterations) # The following vector is initialised for storing forward difference forward_diff_store = np.zeros(iterations) # This is the initilization of automatic converge for i in range(iterations): self.step() u = self._u volu = sum(u[u > 0]) foreground_num[i] = volu if i > 0: # The variable diff_step is the current first order difference diff_step = foreground_num[i] - foreground_num[i - 1] forward_diff_store[i - 1] = diff_step if i > 6: # The variable cur_slider_diff is the sum of sliding window # The size of sliding window is 6 cur_slider_diff = np.sum(forward_diff_store[i - 6:i - 1]) volu_thres = 0.05 * foreground_num[i] convg_one = np.absolute(cur_slider_diff) < 20 convg_two = np.absolute(cur_slider_diff) < volu_thres convg_criteria = np.logical_or(convg_one, convg_two) if convg_criteria: break A = self._u > 0.5 slicevalarray = np.zeros(6) # Front face along dimension 1 somaslice = A[0, :, :] slicearray = np.sum(somaslice, axis=0) sliceval = np.sum(slicearray, axis=0) slicevalarray[0] = sliceval # Back face along dimension 1 somaslice = A[A.shape[0] - 1, :, :] slicearray = np.sum(somaslice, axis=0) sliceval = np.sum(slicearray, axis=0) slicevalarray[1] = sliceval # Front face along dimension 2 somaslice = A[:, 0, :] slicearray = np.sum(somaslice, axis=0) sliceval = np.sum(slicearray, axis=0) slicevalarray[2] = sliceval # Back face along dimension 2 somaslice = A[:, A.shape[1] - 1, :] slicearray = np.sum(somaslice, axis=0) sliceval = np.sum(slicearray, axis=0) slicevalarray[3] = sliceval # Front face along dimension 3 somaslice = A[:, :, 0] slicearray = np.sum(somaslice, axis=0) sliceval = np.sum(slicearray, axis=0) slicevalarray[4] = sliceval # Back face along dimension 3 somaslice = A[:, :, A.shape[2] - 1] slicearray = np.sum(somaslice, axis=0) sliceval = np.sum(slicearray, axis=0) slicevalarray[5] = sliceval # The maxval is used to compare the threshold(100 mentioned later) maxval = slicevalarray.max() # The maxind is the index of slicevalarray. # In addition, it determines which wall will be extended maxind = slicevalarray.argmax() # The size of binary data image sz1 = self.data.shape[0] # sz2 = self.data.shape[1] # sz3 = self.data.shape[2] # extend = enlrspt have value, not extend = (enlrspt=None) # 100 : A threshold of the total number of somatic voxels on each wall if (maxval > 100): self.enlrspt = self.startpoint.copy() self.enlrept = self.endpoint.copy() # The following code determines the most possible wall(face) # which requires the extension if (maxind == 0): self.enlrspt[0] = self.enlrspt[0] - (sz1 / 4) elif (maxind == 1): self.enlrept[0] = self.enlrept[0] + (sz1 / 4) elif (maxind == 2): self.enlrspt[1] = self.enlrspt[1] - (sz1 / 4) elif (maxind == 3): self.enlrept[1] = self.enlrept[1] + (sz1 / 4) elif (maxind == 4): self.enlrspt[2] = self.enlrspt[2] - (sz1 / 4) elif (maxind == 5): self.enlrept[2] = self.enlrept[2] + (sz1 / 4) # To constrain new bounding box inside the image size else: self.enlrspt = None self.enlrept = None def autosmooth(self): """The automatic smoothing of soma volume to remove dendrites""" # The autosmooth is the abbreviation of automatic smoothing iterations = 20 # Calculate the initial volume u = self._u ini_vol = sum(u[u > 0]) # The smooth operation make for i in range(iterations): self.step_sm() u = self._u volu = sum(u[u > 0]) vol_pct = volu / ini_vol # The criteria of the termination of soma growth # The somatic volume underwent dramatic change judge_one = vol_pct < 0.75 judge_two = vol_pct > 1.15 judge_criteria = np.logical_or(judge_one, judge_two) if judge_criteria: break def evolve_visual(msnake, levelset=None, num_iters=20, background=None): """ Visual evolution of a morphological snake. Parameters ---------- msnake : MorphGAC or MorphACWE instance The morphological snake solver. levelset : array-like, optional If given, the levelset of the solver is initialized to this. If not given, the evolution will use the levelset already set in msnake. num_iters : int, optional The number of iterations. background : array-like, optional If given, background will be shown behind the contours instead of msnake.data. """ from matplotlib import pyplot as ppl if levelset is not None: msnake.levelset = levelset # Prepare the visual environment. fig = ppl.gcf() fig.clf() ax1 = fig.add_subplot(1, 2, 1) if background is None: ax1.imshow(msnake.data, cmap=ppl.cm.gray) else: ax1.imshow(background, cmap=ppl.cm.gray) ax1.contour(msnake.levelset, [0.5], colors='r') ax2 = fig.add_subplot(1, 2, 2) ax_u = ax2.imshow(msnake.levelset) ppl.pause(0.001) # Iterate. for i in range(num_iters): # Evolve. msnake.step() # Update figure. del ax1.collections[0] ax1.contour(msnake.levelset, [0.5], colors='r') ax_u.set_data(msnake.levelset) fig.canvas.draw() # ppl.pause(0.001) # Return the last levelset. return msnake.levelset def evolve_visual3d(msnake, levelset=None, num_iters=20): """ Visual evolution of a three-dimensional morphological snake. Parameters ---------- msnake : MorphGAC or MorphACWE instance The morphological snake solver. levelset : array-like, optional If given, the levelset of the solver is initialized to this. If not given, the evolution will use the levelset already set in msnake. num_iters : int, optional The number of iterations. """ from mayavi import mlab # import matplotlib.pyplot as ppl if levelset is not None: msnake.levelset = levelset fig = mlab.gcf() mlab.clf() src = mlab.pipeline.scalar_field(msnake.data) mlab.pipeline.image_plane_widget( src, plane_orientation='x_axes', colormap='gray') cnt = mlab.contour3d(msnake.levelset, contours=[0.5]) @mlab.animate(ui=True) def anim(): for i in range(num_iters): msnake.step() cnt.mlab_source.scalars = msnake.levelset yield anim() mlab.show() # Return the last levelset. return msnake.levelset
24,107
34.040698
86
py
rivuletpy
rivuletpy-master/rivuletpy/__init__.py
0
0
0
py
rivuletpy
rivuletpy-master/rivuletpy/trace.py
import math from tqdm import tqdm import numpy as np import skfmm import msfm from scipy.interpolate import RegularGridInterpolator from scipy.ndimage.morphology import binary_dilation from skimage.morphology import skeletonize_3d from .soma import Soma from .swc import SWC class Tracer(object): def __init__(self): pass def reset(self): pass def trace(self): pass class R2Tracer(Tracer): def __init__(self, quality=False, silent=False, speed=False, clean=False, non_stop=False, skeletonize=False): self._quality = quality self._bimg = None self._dilated_bimg = None self._bsum = 0 # For counting the covered foreground self._bb = None # For making the erasing contour self._t = None # Original timemap self._tt = None # The copy of the timemap self._grad = None self._coverage = 0. self._soma = None # soma self._silent = silent # Disable all console outputs # Tracing stops when 98% of the foreground has been covered self._target_coverage = 0.98 self._cover_ctr_old = 0. self._cover_ctr_new = 0. # The type of speed image to use. Options are ['dt', 'ssm'] self._speed = speed self._erase_ratio = 1.5 # Whether the unconnected branches will be discarded self._clean = clean # Whether to ignore the gap and online confidence stopping criteria self._non_stop = non_stop self.skeletonize = skeletonize self._eps = 1e-5 def trace(self, img, threshold): ''' The main entry for Rivulet2 ''' self.img = img self._bimg = (img > threshold).astype('int') # Segment image if not self._silent: print('(1) -- Detecting Soma...', end='') self._soma = Soma() self._soma.detect(self._bimg, not self._quality, self._silent) self._prep() # Iterative Back Tracking with Erasing if not self._silent: print('(5) --Start Backtracking with {} ...'.format( 'non stop' if self._non_stop else 'standard stopping criteria')) swc = self._iterative_backtrack() if self._clean: swc.prune() return swc, self._soma def _prep(self): if self.skeletonize: print('Skeletonize the binary image...') self._bimg = skeletonize_3d(self._bimg) self._bimg = binary_dilation(self._bimg, iterations=1) self._nforeground = self._bimg.sum() # Dilate bimg to make it less strict for the big gap criteria # It is needed since sometimes the tracing goes along the # boundary of the thin fibre in the binary img self._dilated_bimg = binary_dilation(self._bimg) if not self._silent: print('(2) --Boundary DT...') self._make_dt() if not self._silent: print('(3) --Fast Marching with %s quality...' % ('high' if self._quality else 'low')) self._fast_marching() if not self._silent: print('(4) --Compute Gradients...') self._make_grad() # Make copy of the timemap self._tt = self._t.copy() self._tt[self._bimg <= 0] = -2 # Label all voxels of soma with -3 self._tt[self._soma.mask > 0] = -3 # For making a large tube to contain the last traced branch self._bb = np.zeros(shape=self._tt.shape) def _update_coverage(self): self._cover_ctr_new = np.logical_and( self._tt < 0, self._bimg > 0).sum() self._coverage = self._cover_ctr_new / self._nforeground if not self._silent: self._pbar.update(self._cover_ctr_new - self._cover_ctr_old) self._cover_ctr_old = self._cover_ctr_new def _make_grad(self): # Get the gradient of the Time-crossing map dx, dy, dz = self._dist_gradient() standard_grid = (np.arange(self._t.shape[0]), np.arange(self._t.shape[1]), np.arange(self._t.shape[2])) self._grad = (RegularGridInterpolator(standard_grid, dx), RegularGridInterpolator(standard_grid, dy), RegularGridInterpolator(standard_grid, dz)) def _make_dt(self): ''' Make the distance transform according to the speed type ''' if self._speed: self._dt = self.img.astype(float) # The input image self._dt /= self._dt.max() else: self._dt = skfmm.distance(self._bimg, dx=5e-2) # Boundary DT def _fast_marching(self): speed = self._make_speed() # # Fast Marching if self._quality: # if not self._silent: print('--MSFM...') self._t = msfm.run(speed, self._bimg.copy().astype( 'int64'), self._soma.centroid, True, True) else: # if not self._silent: print('--FM...') marchmap = np.ones(self._bimg.shape) marchmap[self._soma.centroid[0], self._soma.centroid[1], self._soma.centroid[2]] = -1 self._t = skfmm.travel_time(marchmap, speed, dx=5e-3) def _make_speed(self): F = self._dt ** 4 F[F <= 0] = 1e-10 return F def _dist_gradient(self): fx = np.zeros(shape=self._t.shape) fy = np.zeros(shape=self._t.shape) fz = np.zeros(shape=self._t.shape) J = np.zeros(shape=[s + 2 for s in self._t.shape]) # Padded Image J[:, :, :] = self._t.max() J[1:-1, 1:-1, 1:-1] = self._t Ne = [[-1, -1, -1], [-1, -1, 0], [-1, -1, 1], [-1, 0, -1], [-1, 0, 0], [-1, 0, 1], [-1, 1, -1], [-1, 1, 0], [-1, 1, 1], [0, -1, -1], [0, -1, 0], [0, -1, 1], [0, 0, -1], [0, 0, 1], [0, 1, -1], [0, 1, 0], [0, 1, 1], [1, -1, -1], [1, -1, 0], [1, -1, 1], [1, 0, -1], [1, 0, 0], [1, 0, 1], [1, 1, -1], [1, 1, 0], [1, 1, 1]] for n in Ne: In = J[1 + n[0]:J.shape[0] - 1 + n[0], 1 + n[1]:J.shape[1] - 1 + n[1], 1 + n[2]:J.shape[2] - 1 + n[2]] check = In < self._t self._t[check] = In[check] D = np.divide(n, np.linalg.norm(n)) fx[check] = D[0] fy[check] = D[1] fz[check] = D[2] return -fx, -fy, -fz def _step(self, branch): # RK4 Walk for one step p = rk4(branch.pts[-1], self._grad, self._t, 1) branch.update(p, self._bimg, self._dilated_bimg) def _erase(self, branch): # Erase it from the timemap for i in range(len(branch.pts)): n = [math.floor(n) for n in branch.pts[i]] r = 1 if branch.radius[i] < 1 else branch.radius[i] # To make sure all the foreground voxels are included in bb r = math.ceil(r * self._erase_ratio) X, Y, Z = np.meshgrid( constrain_range(n[0] - r, n[0] + r + 1, 0, self._tt.shape[0]), constrain_range(n[1] - r, n[1] + r + 1, 0, self._tt.shape[1]), constrain_range(n[2] - r, n[2] + r + 1, 0, self._tt.shape[2])) self._bb[X, Y, Z] = 1 startidx, endidx = [math.floor(p) for p in branch.pts[0]], [ math.floor(p) for p in branch.pts[-1]] if len(branch.pts) > 5 and self._t[endidx[0], endidx[1], endidx[2]] < self._t[ startidx[0], startidx[1], startidx[2]]: erase_region = np.logical_and( self._t[endidx[0], endidx[1], endidx[2]] <= self._t, self._t <= self._t[startidx[0], startidx[1], startidx[2]]) erase_region = np.logical_and(self._bb, erase_region) else: erase_region = self._bb.astype('bool') if np.count_nonzero(erase_region) > 0: self._tt[erase_region] = -2 if branch.low_conf else -1 self._bb.fill(0) def _iterative_backtrack(self): # Initialise swc with the soma centroid swc = SWC(self._soma) soma_node = np.asarray([0, 1, self._soma.centroid[0], self._soma.centroid[1], self._soma.centroid[2], self._soma.radius, -1, 1.]) swc.add(np.reshape(soma_node, (1, 8))) if not self._silent: self._pbar = tqdm(total=math.floor(self._nforeground * self._target_coverage)) # Loop for all branches while self._coverage < self._target_coverage: self._update_coverage() # Find the geodesic furthest point on foreground time-crossing-map srcpt = np.asarray(np.unravel_index( self._tt.argmax(), self._tt.shape)).astype('float64') branch = R2Branch() branch.add(srcpt, 1., 1.) # Erase the source point just in case self._tt[math.floor(srcpt[0]), math.floor( srcpt[1]), math.floor(srcpt[2])] = -2 keep = True # Loop for 1 back-tracking iteration while True: self._step(branch) head = branch.pts[-1] tt_head = self._tt[math.floor(head[0]), math.floor( head[1]), math.floor(head[2])] # 1. Check out of bound if not inbound(head, self._bimg.shape): branch.slice(0, -1) break # 2. Check for the large gap criterion if branch.gap > np.asarray(branch.radius).mean() * 8 and not self._non_stop: break else: branch.reset_gap() # 3. Check if Soma has been reached if tt_head == -3: keep = True if branch.branchlen > self._soma.radius * 3 else False branch.reached_soma = True break # 4. Check if not moved for 15 iterations if branch.is_stucked(): break # 5. Check for low online confidence if branch.low_conf and not self._non_stop: keep = False break # 6. Check for branch merge # Consider reaches previous explored area traced with branch # Note: when the area was traced due to noise points # (erased with -2), not considered as 'reached' if tt_head == -1: branch.touched = True if swc.size() == 1: break matched, matched_idx = swc.match(head, branch.radius[-1]) if matched > 0: branch.touch_idx = matched_idx break if branch.steps_after_reach > 200: break self._erase(branch) # Add to SWC if it was decided to be kept if keep: pidx = None if branch.reached_soma: pidx = 0 elif branch.touch_idx >= 0: pidx = branch.touch_idx swc.add_branch(branch, pidx) return swc class Branch(object): def __init__(self): self.pts = [] self.radius = [] class R2Branch(Branch): def __init__(self): self.pts = [] self.conf = [] self.radius = [] self.steps_after_reach = 0 self.low_conf = False self.touch_idx = -2 self.reached_soma = False self.branchlen = 0 self.gap = 0 self.online_voxsum = 0 self.stepsz = 0 self.touched = False self.ma_short = -1 self.ma_long = -1 self.ma_short_window = 4 self.ma_long_window = 10 self.in_valley = False def add(self, pt, conf, radius): self.pts.append(pt) self.conf.append(conf) self.radius.append(radius) def is_stucked(self): if self.stepsz == 0: return True if len(self.pts) > 15: if np.linalg.norm(np.asarray(self.pts[-1]) - np.asarray(self.pts[-15])) < 1: return True else: return False else: return False def reset_gap(self): self.gap = 0 def update(self, pt, bimg, dilated_bimg): eps = 1e-5 head = self.pts[-1] velocity = np.asarray(pt) - np.asarray(head) self.stepsz = np.linalg.norm(velocity) self.branchlen += self.stepsz b = dilated_bimg[math.floor(pt[0]), math.floor( pt[1]), math.floor(pt[2])] if b > 0: self.gap += self.stepsz self.online_voxsum += b oc = self.online_voxsum / (len(self.pts) + 1) self.update_ma(oc) # We are stepping in a valley if (self.ma_short < self.ma_long - eps and oc < 0.5 and not self.in_valley): self.in_valley = True # Cut at the valley if self.in_valley and self.ma_short > self.ma_long: valleyidx = np.asarray(self.conf).argmin() # Only cut if the valley confidence is below 0.5 if self.conf[valleyidx] < 0.5: self.slice(0, valleyidx) self.low_conf = True else: in_valley = False if oc <= 0.2: self.low_conf = True if self.touched: self.steps_after_reach += 1 r = estimate_radius(pt, bimg) self.add(pt, oc, r) def update_ma(self, oc): if len(self.pts) > self.ma_long_window: if self.ma_short == -1: self.ma_short = oc else: self.ma_short = exponential_moving_average( oc, self.ma_short, self.ma_short_window if len(self.pts) >= self.ma_short_window else len(self.pts)) if self.ma_long == -1: self.ma_long = oc else: self.ma_long = exponential_moving_average( oc, self.ma_long, self.ma_long_window if len(self.pts) >= self.ma_long_window else len(self.pts)) def slice(self, start, end): self.pts = self.pts[start: end] self.radius = self.radius[start: end] self.conf = self.conf[start: end] def estimate_radius(pt, bimg): r = 0 x = math.floor(pt[0]) y = math.floor(pt[1]) z = math.floor(pt[2]) while True: r += 1 try: if bimg[max(x - r, 0): min(x + r + 1, bimg.shape[0]), max(y - r, 0): min(y + r + 1, bimg.shape[1]), max(z - r, 0): min(z + r + 1, bimg.shape[2])].sum() / (2 * r + 1) ** 3 < .6: break except IndexError: break return r def exponential_moving_average(p, ema, n): ''' The exponential moving average (EMA) traditionally used in analysing stock market. EMA_{i+1} = (p * \alpha) + (EMA_{i} * (1 - \alpha)) where p is the new value; EMA_{i} is the last ema value; n is the time period; \alpha=2/(1+n) is the smoothing factor. --------------------------------------------- Parameters: p: The new value in the sequence ema: the last EMA value n: The period window size ''' alpha = 2 / (1 + n) return p * alpha + ema * (1 - alpha) def rk4(srcpt, ginterp, t, stepsize): # Compute K1 k1 = np.asarray([g(srcpt)[0] for g in ginterp]) k1 *= stepsize / max(np.linalg.norm(k1), 1.) tp = srcpt - 0.5 * k1 # Position of temporary point if not inbound(tp, t.shape): return srcpt # Compute K2 k2 = np.asarray([g(tp)[0] for g in ginterp]) k2 *= stepsize / max(np.linalg.norm(k2), 1.) tp = srcpt - 0.5 * k2 # Position of temporary point if not inbound(tp, t.shape): return srcpt # Compute K3 k3 = np.asarray([g(tp)[0] for g in ginterp]) k3 *= stepsize / max(np.linalg.norm(k3), 1.) tp = srcpt - k3 # Position of temporary point if not inbound(tp, t.shape): return srcpt # Compute K4 k4 = np.asarray([g(tp)[0] for g in ginterp]) k4 *= stepsize / max(np.linalg.norm(k4), 1.) return srcpt - (k1 + k2 * 2 + k3 * 2 + k4) / 6.0 # Compute final point def inbound(pt, shape): return all([True if 0 <= p <= s - 1 else False for p, s in zip(pt, shape)]) def constrain_range(min, max, minlimit, maxlimit): return list( range(min if min > minlimit else minlimit, max if max < maxlimit else maxlimit))
16,754
33.054878
96
py
rivuletpy
rivuletpy-master/rivuletpy/swc.py
import math import numpy as np from .utils.io import saveswc from collections import Counter from random import gauss from random import random from random import randrange from scipy.spatial.distance import cdist class SWC(object): def __init__(self, soma=None): self._data = np.zeros((1, 8)) if soma: self._data[0, :] = np.asarray( [ 0, 1, soma.centroid[0], soma.centroid[1], soma.centroid[2], soma.radius, -1, 1, ] ) def add(self, swc_nodes): np.vstack((self._data, swc_nodes)) def add_branch(self, branch, pidx=None, random_color=True): """ Add a branch to swc. Note: This swc is special with N X 8 shape. The 8-th column is the online confidence """ if random_color: rand_node_type = randrange(6, 257) new_branch = np.zeros((len(branch.pts), 8)) id_start = 1 if self._data.shape[0] == 1 else self._data[:, 0].max() + 1 for i in range(len(branch.pts)): p, r, c = branch.pts[i], branch.radius[i], branch.conf[i] id = id_start + i # 3 for basal dendrite; 4 for apical dendrite; # However now we cannot differentiate them automatically nodetype = 3 if i == len(branch.pts) - 1: # The end of this branch pid = self._data[pidx, 0] if pidx is not None else -2 if pid != -2 and pid != 0 and self._data.shape[0] != 1: # Its connected node is fork point self._data[self._data[:, 0] == pid, 1] = 5 else: pid = id_start + i + 1 if i == 0: nodetype = 6 # Endpoint assert pid != id new_branch[i] = np.asarray( [ id, rand_node_type if random_color else nodetype, p[0], p[1], p[2], r, pid, c, ] ) # Check if any tail should be connected to its tail tail = new_branch[0] matched, minidx = self.match(tail[2:5], tail[5]) if matched and self._data[minidx, 6] == -2: self._data[minidx, 6] = tail[0] self._data = np.vstack((self._data, new_branch)) def _prune_leaves(self): # Find all the leaves childctr = Counter(self._data[:, 6]) leafidlist = [id for id in self._data[:, 0] if id not in self._data[:, 6]] id2dump = [] rmean = self._data[:, 5].mean() # Mean radius for leafid in leafidlist: # Iterate each leaf node nodeid = leafid branch = [] while True: # Get the leaf branch out node = self._data[self._data[:, 0] == nodeid, :].flatten() if node.size == 0: break branch.append(node) parentid = node[6] if childctr[parentid] != 1: break # merged / unconnected nodeid = parentid # Get the length of the leaf leaflen = sum( [ np.linalg.norm(branch[i][2:5] - branch[i - 1][2:5]) for i in range(1, len(branch)) ] ) # Prune if the leave is too short or # the confidence of the leave branch is too low if leaflen <= 4 * rmean: id2dump.extend([node[0] for node in branch]) # Only keep the swc nodes not in the dump id list cutted = [] for nodeidx in range(self._data.shape[0]): if self._data[nodeidx, 0] not in id2dump: cutted.append(self._data[nodeidx, :]) cutted = np.squeeze(np.dstack(cutted)).T self._data = cutted def _prune_unreached(self): """ Only keep the largest connected component """ swcdict = {} for n in self._data: # Hash all the swc nodes swcdict[n[0]] = Node(n[0]) # Try to join all the unconnected branches at first for i, n in enumerate(self._data): if n[6] not in swcdict: # Try to match it matched, midx = self.match(n[2:5], n[5]) if matched: self._data[i, 6] = self._data[midx, 0] # Add mutual links for all nodes for n in self._data: id = n[0] pid = n[6] if pid >= 0: swcdict[id].add_link(swcdict[pid]) groups = connected_components(set(swcdict.values())) lenlist = [len(g) for g in groups] maxidx = lenlist.index(max(lenlist)) set2keep = groups[maxidx] id2keep = [n.id for n in set2keep] self._data = self._data[np.in1d(self._data[:, 0], np.asarray(id2keep)), :] def prune(self): self._prune_unreached() self._prune_leaves() def reset(self, crop_region, zoom_factor): """ Pad and rescale swc back to the original space """ tswc = self._data.copy() if zoom_factor != 1.0: # Pad the swc back to original space tswc[:, 2:5] *= 1.0 / zoom_factor # Pad the swc back tswc[:, 2] += crop_region[0, 0] tswc[:, 3] += crop_region[1, 0] tswc[:, 4] += crop_region[2, 0] self._data = tswc def get_id(self, idx): return self._data[idx, 0] def match(self, pos, radius): """ Find the closest ground truth node """ nodes = self._data[:, 2:5] distlist = np.squeeze(cdist(pos.reshape(1, 3), nodes)) if distlist.size == 0: return False, -2 minidx = distlist.argmin() minnode = self._data[minidx, 2:5] # See if either of them can cover each other with a ball of their own # radius mindist = np.linalg.norm(pos - minnode) return radius > mindist or self._data[minidx, 5] > mindist, minidx def size(self): return self._data.shape[0] def save(self, fname): saveswc(fname, self._data) def get_array(self): return self._data[:, :7] def view(self): from rivuletpy.utils.rendering3 import Viewer3, Line3 # Compute the center of mass center = self._data[:, 2:5].mean(axis=0) translated = self._data[:, 2:5] - np.tile(center, (self._data.shape[0], 1)) # Init viewer viewer = Viewer3(800, 800, 800) viewer.set_bounds( self._data[:, 2].min(), self._data[:, 2].max(), self._data[:, 3].min(), self._data[:, 3].max(), self._data[:, 4].min(), self._data[:, 4].max(), ) lid = self._data[:, 0] line_color = [random(), random(), random()] for i in range(self._data.shape[0]): # Change color if its a bifurcation if (self._data[i, 0] == self._data[:, -1]).sum() > 1: line_color = [random(), random(), random()] # Draw a line between this node and its parent if ( i < self._data.shape[0] - 1 and self._data[i, 0] == self._data[i + 1, -1] ): l = Line3(translated[i, :], translated[i + 1, :]) l.set_color(*line_color) viewer.add_geom(l) else: pid = self._data[i, -1] pidx = np.argwhere(pid == lid).flatten() if len(pidx) == 1: l = Line3(translated[i, :], translated[pidx, :].flatten()) l.set_color(*line_color) viewer.add_geom(l) while True: try: viewer.render(return_rgb_array=False) except KeyboardInterrupt: break def push_nodes_with_binary(self, b, step_ratio=0.1, niter=0): """ Push the nodes towards the center with the binary image boundaries """ lid = list(self._data[:, 0]) lpid = list(self._data[:, -2]) t_data = self._data.copy() children_idx = { pid: [i for i, p in enumerate(lpid) if p == t_data[i, 0]] for pid in lpid } for _ in range(niter): for i in range(t_data.shape[0]): pid, radius, (x, y, z) = ( int(t_data[i, -2]), t_data[i, -3], t_data[i, 2:5], ) cidx = children_idx[pid] if pid != i and pid in lid and len(cidx) <= 1: px, py, pz = t_data[t_data[:, 0] == pid, 2:5][0] vnorm = norm_vec(np.asarray([x - px, y - py, z - pz])) if len(cidx) == 1: cx, cy, cz = t_data[cidx[0], 2:5] vnorm = ( vnorm + norm_vec(np.asarray([cx - x, cy - y, cz - z])) ) / 2 if all([v == 0 for v in vnorm]): continue pt = np.asarray([x, y, z]) p_vectors = get_perpendicular_vectors(pt, vnorm) p_distances = [ get_distance_to_boundary(pt, pvec, b) for pvec in p_vectors ] dx, dy, dz = np.sum( [pv * pd for pv, pd in zip(p_vectors, p_distances)], 0 ) # Constrain the displacement by the nodo radii tx = x + dx * step_ratio ty = y + dy * step_ratio tz = z + dz * step_ratio dist = ( (tx - self._data[i, 2]) ** 2 + (ty - self._data[i, 3]) ** 2 + (tz - self._data[i, 4]) ** 2 ) ** 0.5 if dist <= radius / 2: t_data[i, 2] = tx t_data[i, 3] = ty t_data[i, 4] = tz else: pass self._data = t_data def get_distance_to_boundary(pt, vec, b): temp_pt = pt.copy() while True: next_pt = temp_pt + vec if ( b[math.floor(next_pt[0]), math.floor(next_pt[1]), math.floor(next_pt[2])] <= 0 ): return ((temp_pt - pt) ** 2).sum() ** 0.5 else: temp_pt = next_pt def norm_vec(vec): norm = (vec**2).sum() ** 0.5 return vec / norm def get_perpendicular_vectors(pt, vec): v1 = perpendicular_vector(vec) v2 = -v1 v3 = perpendicular_vector(vec, v1) v4 = -v3 return v1, v2, v3, v4 def make_rand_vector3d(): vec = [gauss(0, 1) for i in range(3)] mag = sum(x**2 for x in vec) ** 0.5 return [x / mag for x in vec] def perpendicular_vector(v, vr=None): return np.cross(v, make_rand_vector3d() if vr is None else vr) def get_subtree_nodeids(swc, node): subtreeids = np.array([]) # Find children chidx = np.argwhere(node[0] == swc[:, 6]) # Recursion stops when there this node is a # leaf with no children, return itself if chidx.size == 0: return node[0] else: # Get the node ids of each children for c in chidx: subids = get_subtree_nodeids(swc, swc[c, :].squeeze()) subtreeids = np.hstack((subtreeids, subids, node[0])) return subtreeids class Node(object): def __init__(self, id): self.__id = id self.__links = set() @property def id(self): return self.__id @property def links(self): return set(self.__links) def add_link(self, other): self.__links.add(other) other.__links.add(self) def connected_components(nodes): """ The function to look for connected components. Reference: https://breakingcode.wordpress.com/2013/04/08/finding-connected-components-in-a-graph/ """ # List of connected components found. The order is random. result = [] # Make a copy of the set, so we can modify it. nodes = set(nodes) # Iterate while we still have nodes to process. while nodes: # Get a random node and remove it from the global set. n = nodes.pop() # This set will contain the next group of nodes # connected to each other. group = {n} # Build a queue with this node in it. queue = [n] # Iterate the queue. # When it's empty, we finished visiting a group of connected nodes. while queue: # Consume the next item from the queue. n = queue.pop(0) # Fetch the neighbors. neighbors = n.links # Remove the neighbors we already visited. neighbors.difference_update(group) # Remove the remaining nodes from the global set. nodes.difference_update(neighbors) # Add them to the group of connected nodes. group.update(neighbors) # Add them to the queue, so we visit them in the next iterations. queue.extend(neighbors) # Add the group to the list of groups. result.append(group) # Return the list of groups. return result
13,630
30.407834
101
py
rivuletpy
rivuletpy-master/rivuletpy/utils/rendering3.py
import os import numpy as np import math # from gym.envs.classic_control.rendering import * from .rendering import * from PIL import Image # PIL library is required import pyglet from pyglet.gl import glu from .io import * # colors black = (0, 0, 0, 1) gray = (0.5, 0.5, 0.5) red = (1, 0, 0) def _add_attrs(geom, attrs): if "color" in attrs: geom.set_color(*attrs["color"]) if "linewidth" in attrs: geom.set_linewidth(attrs["linewidth"]) class Cylinder3(Geom): def __init__(self, centre=(0.0, 0.0, 0.0), radius=2, face=(1,0,0)): Geom.__init__(self) self._centre = centre self._radius = radius self._face = face/np.linalg.norm(face) def render1(self): # http://stackoverflow.com/questions/6992541/opengl-rotation-in-given-direction glPushMatrix() glTranslatef(*self._centre) # glRotatef(-90, 0, 1, 0) # Rotate to face (1,0,0) # print('== Rotate to face', self._face) # glRotatef(-np.arcsin(self._face[2]) * 180 / np.pi, 0, 1, 0) # glRotatef(np.arctan2(self._face[1], self._face[0]) * 180 / np.pi, 0, 0, 1) T = np.array((0., 0., 1.)) # TODO: Need to be face vector Y = np.array((0., 1., 0.)) # TODO: need to be up vector U = (T - np.dot(T, Y)) U /= np.linalg.norm(U) L = np.cross(U, T) M = np.zeros(shape=(4, 4)) M[0:3, 0] = L M[0:3, 1] = U M[0:3, 2] = T M[0:3, 3] = np.array(self._centre) M[-1, -1] = 1 print('M:', M) M = M.flatten('F') M = (GLfloat*len(M))(*M) # M = glGetFloatv(GL_MODELVIEW_MATRIX, M) glMultMatrixf(M) gluCylinder(gluNewQuadric(), self._radius, 0, 4*self._radius, 100, 100) glPopMatrix() class Ball3(Geom): def __init__(self, centre=(0.0, 0.0, 0.0), radius=2): Geom.__init__(self) self._centre = centre self._radius = radius def render1(self): glPushMatrix() glTranslatef(*self._centre) # translate to GL loc ppint gluSphere(gluNewQuadric(), self._radius, 100, 100) glPopMatrix() class Line3(Geom): def __init__(self, start=(0.0, 0.0, 0.0), end=(0.0, 0.0, 0.0)): Geom.__init__(self) self.start = start self.end = end self.linewidth = LineWidth(1) self.add_attr(self.linewidth) def render1(self): # pyglet.graphics.draw(2, gl.GL_LINES, ('v3f', self.start + self.end)) glBegin(GL_LINES) glVertex3f(*self.start) glVertex3f(*self.end) glEnd() def set_line_width(self, x): self.linewidth.stroke = x class Transform3(Transform): def __init__(self, translation=(0.0, 0.0, 0.0), rotation=(0.0, 0.0, 0.0), scale=(1, 1, 1)): self.set_translation(*translation) self.set_rotation(*rotation) self.set_scale(*scale) def enable(self): glPushMatrix() glLoadIdentity() glTranslatef(*self.translation) # translate to GL loc ppint glRotatef(self.rotation[0], 1, 0, 0) glRotatef(self.rotation[1], 0, 1, 0) glRotatef(self.rotation[2], 0, 0, 1) glScalef(*self.scale) def set_translation(self, newx, newy, newz): self.translation = (float(newx), float(newy), float(newz)) def set_rotation(self, newx, newy, newz): self.rotation = (float(newx), float(newy), float(newz)) def set_scale(self, newx, newy, newz): self.scale = (float(newx), float(newy), float(newz)) class Viewer3(Viewer): def __init__(self, width, height, depth, display=None): super(Viewer3, self).__init__(width, height, display) self.depth = depth self.transform = Transform3() @self.window.event def on_resize(width, height): # sets the viewport gl.glViewport(0, 0, width, height) # sets the projection gl.glMatrixMode(gl.GL_PROJECTION) gl.glLoadIdentity() glu.gluPerspective(90, width / float(height), 0.1, 2*self.depth) # sets the model view gl.glMatrixMode(gl.GL_MODELVIEW) gl.glLoadIdentity() return pyglet.event.EVENT_HANDLED @self.window.event def on_mouse_scroll(x, y, scroll_x, scroll_y): # scroll the MOUSE WHEEL to zoom self.transform.set_translation(self.transform.translation[0], self.transform.translation[1], self.transform.translation[2] - scroll_y *20) @self.window.event def on_mouse_drag(x, y, dx, dy, button, modifiers): # press the LEFT MOUSE BUTTON to rotate if button == pyglet.window.mouse.LEFT: self.transform.set_rotation(self.transform.rotation[0] - dy / 5.0, self.transform.rotation[1] + dx / 5.0, self.transform.rotation[2]) # press the LEFT and RIGHT MOUSE BUTTON simultaneously to pan if button == pyglet.window.mouse.LEFT | pyglet.window.mouse.RIGHT: self.transform.set_translation(self.transform.translation[0] + dx/2, self.transform.translation[1] + dy/2, self.transform.translation[2]) def set_bounds(self, left, right, bottom, top, towards, away): assert right > left and top > bottom and away > towards scalex = self.width / (right-left) scaley = self.height / (top-bottom) scalez = self.depth / (away-towards) self.transform = Transform3( translation=(0, 0, -1.5*away)) def draw_line(self, start, end, **attrs): geom = Line3(start, end) _add_attrs(geom, attrs) self.add_onetime(geom) return geom def render(self, return_rgb_array=False): glClear(GL_COLOR_BUFFER_BIT); self.window.clear() self.window.switch_to() self.window.dispatch_events() # glLoadIdentity() self.transform.enable() for geom in self.geoms: geom.render() for geom in self.onetime_geoms: geom.render() self.transform.disable() arr = None if return_rgb_array: image_data = pyglet.image.get_buffer_manager().get_color_buffer().get_image_data() arr = np.fromstring(image_data.data, dtype=np.uint8, sep='') arr = arr.reshape(self.height, self.width, 4) arr = arr[::-1,:,0:3] self.window.flip() self.onetime_geoms = [] return arr # super(Viewer3, self).render(return_rgb_array)
6,644
33.252577
95
py
rivuletpy
rivuletpy-master/rivuletpy/utils/metrics.py
from collections import deque import numpy as np from scipy.spatial.distance import cdist def precision_recall(swc1, swc2, dist1=4, dist2=4): ''' Calculate the precision, recall and F1 score between swc1 and swc2 (ground truth) It generates a new swc file with node types indicating the agreement between two input swc files In the output swc file: node type - 1. the node is in both swc1 agree with swc2 - 2. the node is in swc1, not in swc2 (over-traced) - 3. the node is in swc2, not in swc1 (under-traced) target: The swc from the tracing method gt: The swc from the ground truth dist1: The distance to consider for precision dist2: The distance to consider for recall ''' TPCOLOUR, FPCOLOUR, FNCOLOUR = 3, 2, 180 # COLOUR is the SWC node type defined for visualising in V3D d = cdist(swc1[:, 2:5], swc2[:, 2:5]) mindist1 = d.min(axis=1) tp = (mindist1 < dist1).sum() fp = swc1.shape[0] - tp mindist2 = d.min(axis=0) fn = (mindist2 > dist2).sum() precision = tp / (tp + fp) recall = tp / (tp + fn) f1 = 2 * precision * recall / (precision + recall) # Make the swc for visual comparison swc1[mindist1 <= dist1, 1] = TPCOLOUR swc1[mindist1 > dist1, 1] = FPCOLOUR swc2_fn = swc2[mindist2 > dist2, :] swc2_fn[:, 0] = swc2_fn[:, 0] + 100000 swc2_fn[:, -1] = swc2_fn[:, -1] + 100000 swc2_fn[:, 1] = FNCOLOUR swc_compare = np.vstack((swc1, swc2_fn)) swc_compare[:, -2] = 1 # Compute the SD, SSD, SSD% defined in Peng.et.al 2010 SD = (np.mean(mindist1) + np.mean(mindist2)) / 2 far1, far2 = mindist1[mindist1 > dist1], mindist2[mindist2 > dist2] SSD = (np.mean(far1) + np.mean(far2)) / 2 pSSD = (len(far1) / len(mindist1) + len(far2) / len(mindist2)) / 2 return (precision, recall, f1), (SD, SSD, pSSD), swc_compare def upsample_swc(swc): tswc = swc.copy() id_idx = {} # Build a nodeid->idx hash table for nodeidx in range(tswc.shape[0]): id_idx[tswc[nodeidx, 0]] = nodeidx newid = tswc[:,0].max() + 1 newnodes = [] for nodeidx in range(tswc.shape[0]): pid = tswc[nodeidx, -1] # parent id if pid not in id_idx: # raise Exception('Parent with id %d not found' % pid) continue nodepos = tswc[nodeidx, 2:5] parentpos = tswc[id_idx[pid], 2:5] if np.linalg.norm(nodepos - parentpos) > 1.: # Add a node in the middle if too far mid_pos = nodepos + 0.5 * (parentpos - nodepos) newnodes.append( np.asarray([newid, 2, mid_pos[0], mid_pos[1], mid_pos[2], 1, pid]) ) newid += 1 tswc[nodeidx, -1] = newid # Stack the new nodes to the end of the swc file newnodes = np.vstack(newnodes) tswc = np.vstack((tswc, newnodes)) return tswc def gaussian_distance(swc1, swc2, sigma=2.): ''' The geometric metrics of NetMets. The gaussian distances between the closest neighbours returns : (M1, M2) where M1 is the gaussian distances from the nodes in swc1 to their closest neighbour in swc2; vise versa for M2 D. Mayerich, C. Bjornsson, J. Taylor, and B. Roysam, “NetMets: software for quantifying and visualizing errors in biological network segmentation.,” BMC Bioinformatics, vol. 13 Suppl 8, no. Suppl 8, p. S7, 2012. ''' swc1 = upsample_swc(swc1) swc2 = upsample_swc(swc2) d = cdist(swc1[:, 2:5], swc2[:, 2:5]) # Pairwise distances between 2 swc files mindist1 = d.min(axis=1) M1 = 1 - np.exp(mindist1 ** 2 / (2 * sigma ** 2)) mindist2 = d.min(axis=0) M2 = 1 - np.exp(mindist2 ** 2 / (2 * sigma ** 2)) return M1, M2 def connectivity_distance(swc1, swc2, sigma=2., ignore_leaf=True): ''' The connectivity metrics of NetMets. Returns (midx1, midx2): the indices of nodes in each swc that have connection errors D. Mayerich, C. Bjornsson, J. Taylor, and B. Roysam, “NetMets: software for quantifying and visualizing errors in biological network segmentation.,” BMC Bioinformatics, vol. 13 Suppl 8, no. Suppl 8, p. S7, 2012. ''' # graph Initialisation d = cdist(swc1[:, 2:5], swc2[:, 2:5]) # Pairwise distances between 2 swc files mindist1, mindist2 = d.min(axis=1), d.min(axis=0) minidx1, minidx2 = d.argmin(axis=1), d.argmin(axis=0) # Colour nodes - matched nodes have the same colour cnodes1, cnodes2 = {}, {}# Coloured Nodes <id, colour> for i in range(swc1.shape[0]): if mindist1[i] < sigma: cnodes1[swc1[i, 0]] = i cnodes2[swc2[minidx1[i], 0]] = i # Build Initial graphs, Edge: <id_i, id_j>: 1 g1 = build_graph_from_swc(swc1) g2 = build_graph_from_swc(swc2) # BFS to build the core graph for both swc, returns the remaining edges not used to build the core graph dg1 = build_core_graph(g1, cnodes1) dg2 = build_core_graph(g2, cnodes2) # Find the diff edges with coloured nodes involved mid1 = set() for id in dg1: for nid in g1[id]: if nid in cnodes1: mid1.add(nid) mid2 = set() for id in dg2: for nid in g2[id]: if nid in cnodes2: mid2.add(nid) id_idx_hash1 = {} for i in range(swc1.shape[0]): id_idx_hash1[swc1[i, 0]] = i id_idx_hash2 = {} for i in range(swc2.shape[0]): id_idx_hash2[swc2[i, 0]] = i midx1 = [ int(id_idx_hash1[id]) for id in mid1 ] # Mistake coloured nodes in edges of dg1 midx2 = [ int(id_idx_hash2[id]) for id in mid2 ] # Mistake coloured nodes in edges of dg2 # Filter out the midx of nodes on leaf segments if ignore_leaf: leafidx1 = find_leaf_idx(swc1) midx1 = set(midx1) - set(leafidx1) leafidx2 = find_leaf_idx(swc2) midx2 = set(midx2) - set(leafidx2) return len(midx1) / len(mid1), len(midx2) / len(mid2) def find_leaf_idx(swc): # The degree of a node is the number of children + 1 except the root degree = np.zeros(swc.shape[0]) for i in range(swc.shape[0]): degree[i] = np.count_nonzero(swc[:, -1] == swc[i, 0]) + 1 # A node is a leaf node if it is parent to no other node leaf_segment_idx = [] leaf_node_idx = np.where(degree == 1)[0] for idx in leaf_node_idx: # Add its parent to the leaf segment idx list if its parent degree < 3 nodeidx = idx while degree[nodeidx] < 3: leaf_segment_idx.append(int(nodeidx)) if swc[nodeidx, -1] < 0: break nodeidx = np.where(swc[:, 0] == swc[nodeidx, -1])[0] return leaf_segment_idx def build_graph_from_swc(swc): g = {} for i in range(swc.shape[0]): id, pid = swc[i, 0], swc[i, -1] if id in g: g[id].append(pid) else: g[id] = [pid] if pid in g: g[pid].append(id) else: g[pid] = [id] for key, value in g.items(): g[key] = set(value) return g def build_core_graph(g, cnodes): ''' Returns the edges not used in building the core graph (topologically matched between two graphs) ''' cnodes = cnodes.copy() # Coloured node list to mark which have not been discovered dg = g.copy() while cnodes: root = next(iter(cnodes)) core_neighbours = find_core_neighbours_bfs(dg, root, cnodes) # BFS to discover the neighbour nodes_on_path = set() if core_neighbours: for id in core_neighbours: nodes_on_path = nodes_on_path.union(track_path_nodes_dijstra(dg, id, root)) else: nodes_on_path.add(root) cnodes.pop(root) # Remove the discovered coloured nodes for n in nodes_on_path: dg.pop(n, None) for n in dg: dg[n] = dg[n].difference(nodes_on_path) return dg def find_core_neighbours_bfs(g, root, cnodes): ''' Find the coloured neighbours of root node with BFS search ''' visited = {} node_queue = deque() visited[root] = True node_queue.append(root) core_neighbours = [] while node_queue: r = node_queue.popleft() if r in cnodes and r != root: core_neighbours.append(r) # If this node is coloured, bfs stops on it and add it to the core neighbours else: for n in g[r]: # visit all the neighbours of r if n not in visited: visited[n] = True node_queue.append(n) return core_neighbours def track_path_nodes_dijstra(g, target, source): path = {} visited = {source: 0} nodes = g.copy() while nodes: min_node = None for node in nodes: if node in visited: if min_node is None: min_node = node elif visited[node] < visited[min_node]: min_node = node if min_node is None: break nodes.pop(min_node) tweight = visited[min_node] for n in g[min_node]: weight = tweight + 1 if n not in visited or weight < visited[n]: visited[n] = weight path[n] = min_node if min_node == target: break nodes_on_path, n = set(), target while n != source: n = path[n] nodes_on_path.add(n) return nodes_on_path
9,515
31.367347
116
py
rivuletpy
rivuletpy-master/rivuletpy/utils/__init__.py
0
0
0
py
rivuletpy
rivuletpy-master/rivuletpy/utils/io.py
import os import numpy as np from scipy import io as sio import SimpleITK as sitk def loadimg(file, target_resolution): if file.endswith('.mat'): filecont = sio.loadmat(file) img = filecont['img'] for z in range(img.shape[-1]): # Flip the image upside down img[:, :, z] = np.flipud(img[:, :, z]) img = np.swapaxes(img, 0, 1) elif file.endswith('.tif'): img = loadtiff3d(file) elif file.endswith('.mhd'): from scipy.ndimage.interpolation import zoom mhd = sitk.ReadImage(file) img = sitk.GetArrayFromImage(mhd) # Resample the image to isotropic resolution print('Resample Image to isotropic resolution 1mmx1mmx1mm') sx, sy, sz = mhd.GetSpacing() img = zoom(img, (sz / target_resolution, sy / target_resolution, sx / target_resolution), order=0) img = np.transpose(img, (2, 1, 0)) elif file.endswith('.nii') or file.endswith('.nii.gz'): import nibabel as nib img = nib.load(file) img = img.get_data() else: raise IOError("The extension of " + file + 'is not supported. File extension supported are: *.tif, *.mat, *.nii') return img def loadtiff3d(filepath): """Load a tiff file into 3D numpy array""" from libtiff import TIFF tiff = TIFF.open(filepath, mode='r') stack = [] for sample in tiff.iter_images(): stack.append(np.rot90(np.fliplr(np.flipud(sample)))) out = np.dstack(stack) tiff.close() return out def writetiff3d(filepath, block): from libtiff import TIFF try: os.remove(filepath) except OSError: pass tiff = TIFF.open(filepath, mode='w') block = np.swapaxes(block, 0, 1) for z in range(block.shape[2]): tiff.write_image(np.flipud(block[:, :, z]), compression=None) tiff.close() def loadswc(filepath): ''' Load swc file as a N X 7 numpy array ''' swc = [] with open(filepath) as f: lines = f.read().split("\n") for l in lines: if not l.startswith('#'): cells = l.split(' ') if len(cells) == 7: cells = [float(c) for c in cells] # cells[2:5] = [c-1 for c in cells[2:5]] swc.append(cells) return np.array(swc) def saveswc(filepath, swc): if swc.shape[1] > 7: swc = swc[:, :7] with open(filepath, 'w') as f: for i in range(swc.shape[0]): print('%d %d %.3f %.3f %.3f %.3f %d' % tuple(swc[i, :].tolist()), file=f) def crop(img, thr): """Crop a 3D block with value > thr""" ind = np.argwhere(img > thr) x = ind[:, 0] y = ind[:, 1] z = ind[:, 2] xmin = max(x.min() - 10, 0) xmax = min(x.max() + 10, img.shape[0]) ymin = max(y.min() - 10, 1) ymax = min(y.max() + 10, img.shape[1]) zmin = max(z.min() - 10, 2) zmax = min(z.max() + 10, img.shape[2]) return img[xmin:xmax, ymin:ymax, zmin:zmax], np.array( [[xmin, xmax], [ymin, ymax], [zmin, zmax]]) def world2ras(voxpos): '''Get the vox2ras-tkr transform. Inspired by get_vox2ras_tkr in https://discourse.slicer.org/t/building-the-ijk-to-ras-transform-from-a-nrrd-file/1513 ''' x, y, z = voxpos lps_to_ras = np.diag([-1, -1, 1, 1]) p = lps_to_ras.dot(np.asarray([x, y, z, 1]).T) return np.squeeze(p)[:3] def swc2world(swc, mhd, spacing, slicer=False): # First transfer the image coordinates to the original image size sp = mhd.GetSpacing() swc[:, 2] *= spacing[0] / sp[0] swc[:, 3] *= spacing[1] / sp[1] swc[:, 4] *= spacing[2] / sp[2] # USe SimpleITK to transform back to physical coordinates for i in range(swc.shape[0]): swc[i, 2:5] = mhd.TransformContinuousIndexToPhysicalPoint(swc[i, 2:5]) swc[i, 5] *= spacing[0] # Transform to RAS spacing coordinates that can be rendered in 3D Slicer if requested if slicer: print('Converting the vtk coordinates to RAS space') for i in range(swc.shape[0]): swc[i, 2:5] = world2ras(swc[i, 2:5]) return swc def swc2vtk(swc, outvtkpath): swc_arr = swc.get_array() nnode = swc_arr.shape[0] vtkstr = '# vtk DataFile Version 2.0\n' vtkstr += 'Generated with Rivuletpy\n' vtkstr += 'ASCII\n' vtkstr += 'DATASET POLYDATA\n' vtkstr += 'POINTS {} float\n'.format(nnode) id2vtkidx = {} for i in range(nnode): vtkstr += '{} {} {}\n'.format(swc_arr[i, 2], swc_arr[i, 3], swc_arr[i, 4]) id2vtkidx[int(swc_arr[i, 0])] = i linectr = 0 vtklinestr = '' for i in range(nnode): id, pid = swc_arr[i, 0], swc_arr[i, -1] if pid >= 0 and int(pid) in id2vtkidx: linectr += 1 vtklinestr += '{} {} {}\n'.format(2, id2vtkidx[int(id)], id2vtkidx[int(pid)]) vtkstr += 'LINES {} {}\n'.format(linectr, linectr * 3) vtkstr += vtklinestr vtkstr += "POINT_DATA {}\n".format(nnode) vtkstr += "SCALARS contourArray double\n" vtkstr += "LOOKUP_TABLE default\n" for i in range(nnode): vtkstr += '{}\n'.format(swc_arr[i, -2]) vtkstr += "SCALARS indicatorArray char\n" vtkstr += "LOOKUP_TABLE default\n" for i in range(nnode): vtkstr += '0\n' with open(outvtkpath, 'w') as f: f.write(vtkstr)
5,620
29.22043
92
py
rivuletpy
rivuletpy-master/rivuletpy/utils/rendering.py
""" 2D rendering framework """ from __future__ import division import os import six import sys import pyglet from pyglet.gl import * import math import numpy as np RAD2DEG = 57.29577951308232 def get_display(spec): """Convert a display specification (such as :0) into an actual Display object. Pyglet only supports multiple Displays on Linux. """ if spec is None: return None elif isinstance(spec, six.string_types): return pyglet.canvas.Display(spec) else: raise error.Error( 'Invalid display specification: {}. (Must be a string like :0 or None.)'.format(spec)) class Viewer(object): def __init__(self, width, height, display=None): display = get_display(display) self.width = width self.height = height self.window = pyglet.window.Window( width=width, height=height, display=display) self.window.on_close = self.window_closed_by_user self.geoms = [] self.onetime_geoms = [] self.transform = Transform() glEnable(GL_BLEND) glBlendFunc(GL_SRC_ALPHA, GL_ONE_MINUS_SRC_ALPHA) def close(self): self.window.close() def window_closed_by_user(self): self.close() def set_bounds(self, left, right, bottom, top): assert right > left and top > bottom scalex = self.width/(right-left) scaley = self.height/(top-bottom) self.transform = Transform( translation=(-left*scalex, -bottom*scalex), scale=(scalex, scaley)) def add_geom(self, geom): self.geoms.append(geom) def add_onetime(self, geom): self.onetime_geoms.append(geom) def render(self, return_rgb_array=False): glClearColor(1, 1, 1, 1) self.window.clear() self.window.switch_to() self.window.dispatch_events() self.transform.enable() for geom in self.geoms: geom.render() for geom in self.onetime_geoms: geom.render() self.transform.disable() arr = None if return_rgb_array: image_data = pyglet.image.get_buffer_manager().get_color_buffer().get_image_data() arr = np.fromstring(image_data.data, dtype=np.uint8, sep='') arr = arr.reshape(self.height, self.width, 4) arr = arr[::-1, :, 0:3] self.window.flip() self.onetime_geoms = [] return arr # Convenience def draw_circle(self, radius=10, res=30, filled=True, **attrs): geom = make_circle(radius=radius, res=res, filled=filled) _add_attrs(geom, attrs) self.add_onetime(geom) return geom def draw_polygon(self, v, filled=True, **attrs): geom = make_polygon(v=v, filled=filled) _add_attrs(geom, attrs) self.add_onetime(geom) return geom def draw_polyline(self, v, **attrs): geom = make_polyline(v=v) _add_attrs(geom, attrs) self.add_onetime(geom) return geom def draw_line(self, start, end, **attrs): geom = Line(start, end) _add_attrs(geom, attrs) self.add_onetime(geom) return geom def get_array(self): self.window.flip() image_data = pyglet.image.get_buffer_manager().get_color_buffer().get_image_data() self.window.flip() arr = np.fromstring(image_data.data, dtype=np.uint8, sep='') arr = arr.reshape(self.height, self.width, 4) return arr[::-1, :, 0:3] def _add_attrs(geom, attrs): if "color" in attrs: geom.set_color(*attrs["color"]) if "linewidth" in attrs: geom.set_linewidth(attrs["linewidth"]) class Geom(object): def __init__(self): self._color = Color((0, 0, 0, 1.0)) self.attrs = [self._color] def render(self): for attr in reversed(self.attrs): attr.enable() self.render1() for attr in self.attrs: attr.disable() def render1(self): raise NotImplementedError def add_attr(self, attr): self.attrs.append(attr) def set_color(self, r, g, b): self._color.vec4 = (r, g, b, 1) class Attr(object): def enable(self): raise NotImplementedError def disable(self): pass class Transform(Attr): def __init__(self, translation=(0.0, 0.0), rotation=0.0, scale=(1, 1)): self.set_translation(*translation) self.set_rotation(rotation) self.set_scale(*scale) def enable(self): glPushMatrix() # translate to GL loc ppint glTranslatef(self.translation[0], self.translation[1], 0) glRotatef(RAD2DEG * self.rotation, 0, 0, 1.0) glScalef(self.scale[0], self.scale[1], 1) def disable(self): glPopMatrix() def set_translation(self, newx, newy): self.translation = (float(newx), float(newy)) def set_rotation(self, new): self.rotation = float(new) def set_scale(self, newx, newy): self.scale = (float(newx), float(newy)) class Color(Attr): def __init__(self, vec4): self.vec4 = vec4 def enable(self): glColor4f(*self.vec4) class LineStyle(Attr): def __init__(self, style): self.style = style def enable(self): glEnable(GL_LINE_STIPPLE) glLineStipple(1, self.style) def disable(self): glDisable(GL_LINE_STIPPLE) class LineWidth(Attr): def __init__(self, stroke): self.stroke = stroke def enable(self): glLineWidth(self.stroke) class Point(Geom): def __init__(self): Geom.__init__(self) def render1(self): glBegin(GL_POINTS) # draw point glVertex3f(0.0, 0.0, 0.0) glEnd() class FilledPolygon(Geom): def __init__(self, v): Geom.__init__(self) self.v = v def render1(self): if len(self.v) == 4: glBegin(GL_QUADS) elif len(self.v) > 4: glBegin(GL_POLYGON) else: glBegin(GL_TRIANGLES) for p in self.v: glVertex3f(p[0], p[1], 0) # draw each vertex glEnd() def make_circle(radius=10, res=30, filled=True): points = [] for i in range(res): ang = 2*math.pi*i / res points.append((math.cos(ang)*radius, math.sin(ang)*radius)) if filled: return FilledPolygon(points) else: return PolyLine(points, True) def make_polygon(v, filled=True): if filled: return FilledPolygon(v) else: return PolyLine(v, True) def make_polyline(v): return PolyLine(v, False) def make_capsule(length, width): l, r, t, b = 0, length, width/2, -width/2 box = make_polygon([(l, b), (l, t), (r, t), (r, b)]) circ0 = make_circle(width/2) circ1 = make_circle(width/2) circ1.add_attr(Transform(translation=(length, 0))) geom = Compound([box, circ0, circ1]) return geom class Compound(Geom): def __init__(self, gs): Geom.__init__(self) self.gs = gs for g in self.gs: g.attrs = [a for a in g.attrs if not isinstance(a, Color)] def render1(self): for g in self.gs: g.render() class PolyLine(Geom): def __init__(self, v, close): Geom.__init__(self) self.v = v self.close = close self.linewidth = LineWidth(1) self.add_attr(self.linewidth) def render1(self): glBegin(GL_LINE_LOOP if self.close else GL_LINE_STRIP) for p in self.v: glVertex3f(p[0], p[1], 0) # draw each vertex glEnd() def set_linewidth(self, x): self.linewidth.stroke = x class Line(Geom): def __init__(self, start=(0.0, 0.0), end=(0.0, 0.0)): Geom.__init__(self) self.start = start self.end = end self.linewidth = LineWidth(1) self.add_attr(self.linewidth) def render1(self): glBegin(GL_LINES) glVertex2f(*self.start) glVertex2f(*self.end) glEnd() class Image(Geom): def __init__(self, fname, width, height): Geom.__init__(self) self.width = width self.height = height img = pyglet.image.load(fname) self.img = img self.flip = False def render1(self): self.img.blit(-self.width/2, -self.height/2, width=self.width, height=self.height) # ================================================================ class SimpleImageViewer(object): def __init__(self, display=None): self.window = None self.isopen = False self.display = display def imshow(self, arr): if self.window is None: height, width, channels = arr.shape self.window = pyglet.window.Window( width=width, height=height, display=self.display) self.width = width self.height = height self.isopen = True assert arr.shape == ( self.height, self.width, 3), "You passed in an image with the wrong number shape" image = pyglet.image.ImageData( self.width, self.height, 'RGB', arr.tobytes(), pitch=self.width * -3) self.window.clear() self.window.switch_to() self.window.dispatch_events() image.blit(0, 0) self.window.flip() def close(self): if self.isopen: self.window.close() self.isopen = False def __del__(self): self.close()
9,508
24.980874
98
py
rivuletpy
rivuletpy-master/filtering/thresholding.py
import numpy as np def fuzzy(img, level=128, p=2): ''' Image auto thresholding with measure of fuzziness using the Yager's measure Implemented the algorithm following Eq.(11) in L. K. Huang and M. J. J. Wang, “Image thresholding by minimizing the measures of fuzziness,” Pattern Recognit., vol. 28, no. 1, pp. 41–51, 1995. img: a ND numpy matrix level: number of levels to partition the histogram p: order of metric - 1 for Hamming metric - 2 for Euclidean metric ''' assert (p == 1 or p == 2) high, low = img.max(), img.min(), step = (high - low) / level backweighted = foreweighted = 0 nvox = img.size count = np.asarray([((img >= t - step / 2) & (img < t + step / 2)).sum() for t in np.arange(low, high, step)]) weighted = np.arange(low, high, step) * count backcount = np.asarray( [(img < t + step / 2).sum() for t in np.arange(low, high, step)]) forecount = np.asarray([nvox] * level) - backcount backweighted = np.cumsum(weighted) # foreweighted = np.full((level,), backweighted[-1]) - backweighted foreweighted = np.cumsum(weighted[::-1])[::-1] muback = backweighted / backcount mufore = foreweighted / forecount yager = np.zeros((level, )) for i, t1 in enumerate(np.arange(low, high, step)): gsum = 0. for j, t2 in enumerate(np.arange(low, high, step)): mu = muback[i] if t2 <= t1 else mufore[i] C = np.abs(t1 - low) if t2 <= t1 else np.abs(high - t1) C = 1e-4 if C == 0 else C mux = 1. / (1. + np.abs(t2 - mu) / C) mux_reversed = 1. - mux gsum += np.abs(mux - mux_reversed)**p yager[i] = gsum**1 / p # The chosen threshold with least fuzziness return yager.argmin() * step + low, yager def suppress(img, threshold): img[img <= threshold] = 0 return img def rescale(img, overwrite=False): ''' Rescale image intensities linearly to 0~255 ''' if not overwrite: result = img.copy() else: result = img result = result.astype('float') result -= result.min() result /= result.max() result *= 255 return result
2,224
31.720588
79
py
rivuletpy
rivuletpy-master/filtering/anisotropic.py
import numpy as np from scipy.special import jv # Bessel Function of the first kind from scipy.linalg import eig from scipy.fftpack import fftn, ifftn, ifft # import progressbar from tqdm import tqdm from scipy.ndimage import filters as fi import math # An implementation of the Optimally Oriented # M.W.K. Law and A.C.S. Chung, ``Three Dimensional Curvilinear # Structure Detection using Optimally Oriented Flux'', ECCV 2008, pp. # 368--382. # Max W. K. Law et al., ``Dilated Divergence based Scale-Space # Representation for Curve Analysis'', ECCV 2012, pp. 557--571. # Author: Siqi Liu def response(img, rsptype='oof', **kwargs): eps = 1e-12 rsp = np.zeros(img.shape) # bar = progressbar.ProgressBar(max_value=kwargs['radii'].size) # bar.update(0) W = np.zeros((img.shape[0], img.shape[1], img.shape[2], 3)) # Eigen values to save V = np.zeros((img.shape[0], img.shape[1], img.shape[2], 3, 3)) # Eigen vectors to save if rsptype == 'oof' : rsptensor = ooftensor(img, kwargs['radii'], kwargs['memory_save']) elif rsptype == 'bg': rsptensor = bgtensor(img, kwargs['radii'], kwargs['rho']) pbar = tqdm(total=len(kwargs['radii'])) for i, tensorfield in enumerate(rsptensor): # Make the tensor from tensorfield f11, f12, f13, f22, f23, f33 = tensorfield tensor = np.stack((f11, f12, f13, f12, f22, f23, f13, f23, f33), axis=-1) del f11 del f12 del f13 del f22 del f23 del f33 tensor = tensor.reshape(img.shape[0], img.shape[1], img.shape[2], 3, 3) w, v = np.linalg.eigh(tensor) del tensor sume = w.sum(axis=-1) nvox = img.shape[0] * img.shape[1] * img.shape[2] sortidx = np.argsort(np.abs(w), axis=-1) sortidx = sortidx.reshape((nvox, 3)) # Sort eigenvalues according to their abs w = w.reshape((nvox, 3)) for j, (idx, value) in enumerate(zip(sortidx, w)): w[j,:] = value[idx] w = w.reshape(img.shape[0], img.shape[1], img.shape[2], 3) # Sort eigenvectors according to their abs v = v.reshape((nvox, 3, 3)) for j, (idx, vec) in enumerate(zip(sortidx, v)): v[j,:,:] = vec[:, idx] del sortidx v = v.reshape(img.shape[0], img.shape[1], img.shape[2], 3, 3) mine = w[:,:,:, 0] mide = w[:,:,:, 1] maxe = w[:,:,:, 2] if rsptype == 'oof': feat = maxe elif rsptype == 'bg': feat = -mide / maxe * (mide + maxe) # Medialness measure response cond = sume >= 0 feat[cond] = 0 # Filter the non-anisotropic voxels del mine del maxe del mide del sume cond = np.abs(feat) > np.abs(rsp) W[cond, :] = w[cond, :] V[cond, :, :] = v[cond, :, :] rsp[cond] = feat[cond] del v del w del tensorfield del feat del cond pbar.update(1) return rsp, V, W def bgkern3(kerlen, mu=0, sigma=3., rho=0.2): ''' Generate the bi-gaussian kernel ''' sigma_b = rho * sigma k = rho ** 2 kr = (kerlen - 1) / 2 X, Y, Z = np.meshgrid(np.arange(-kr, kr+1), np.arange(-kr, kr+1), np.arange(-kr, kr+1)) dist = np.linalg.norm(np.stack((X, Y, Z)), axis=0) G = gkern3(dist, mu, sigma) # Normal Gaussian with mean at origin Gb = gkern3(dist, sigma-sigma_b, sigma_b) c0 = k * Gb[0, 0, math.floor(sigma_b)] - G[0, 0, math.floor(sigma)] c1 = G[0, 0, math.floor(sigma)] - k * Gb[0, 0, math.floor(sigma_b)] + c0 G += c0 Gb = k * Gb + c1 # Inverse Gaussian with phase shift # Replace the centre of Gb with G central_region = dist <= sigma del dist X = (X[central_region] + kr).astype('int') Y = (Y[central_region] + kr).astype('int') Z = (Z[central_region] + kr).astype('int') Gb[X, Y, Z] = G[X, Y, Z] return Gb def eigh(a, UPLO='L'): # I Borrowed from Dipy """Iterate over `np.linalg.eigh` if it doesn't support vectorized operation Parameters ---------- a : array_like (..., M, M) Hermitian/Symmetric matrices whose eigenvalues and eigenvectors are to be computed. UPLO : {'L', 'U'}, optional Specifies whether the calculation is done with the lower triangular part of `a` ('L', default) or the upper triangular part ('U'). Returns ------- w : ndarray (..., M) The eigenvalues in ascending order, each repeated according to its multiplicity. v : ndarray (..., M, M) The column ``v[..., :, i]`` is the normalized eigenvector corresponding to the eigenvalue ``w[..., i]``. Raises ------ LinAlgError If the eigenvalue computation does not converge. See Also -------- np.linalg.eigh """ a = np.asarray(a) if a.ndim > 2 and NUMPY_LESS_1_8: shape = a.shape[:-2] a = a.reshape(-1, a.shape[-2], a.shape[-1]) evals = np.empty((a.shape[0], a.shape[1])) evecs = np.empty((a.shape[0], a.shape[1], a.shape[1])) for i, item in enumerate(a): evals[i], evecs[i] = np.linalg.eigh(item, UPLO) return (evals.reshape(shape + (a.shape[1], )), evecs.reshape(shape + (a.shape[1], a.shape[1]))) return np.linalg.eigh(a, UPLO) def gkern3(dist, mu=0., sigma=3.): ''' Make 3D gaussian kernel ''' # Make a dirac spherical function return np.exp(-0.5 * (((dist - mu) / sigma)**2)) / (sigma * np.sqrt(2. * np.pi)) def hessian3(x): """ Calculate the hessian matrix with finite differences Parameters: - x : ndarray Returns: an array of shape (x.dim, x.ndim) + x.shape where the array[i, j, ...] corresponds to the second derivative x_ij """ x_grad = np.gradient(x) tmpgrad = np.gradient(x_grad[0]) f11 = tmpgrad[0] f12 = tmpgrad[1] f13 = tmpgrad[2] tmpgrad = np.gradient(x_grad[1]) f22 = tmpgrad[1] f23 = tmpgrad[2] tmpgrad = np.gradient(x_grad[2]) f33 = tmpgrad[2] return [f11, f12, f13, f22, f23, f33] def bgtensor(img, lsigma, rho=0.2): eps = 1e-12 fimg = fftn(img, overwrite_x=True) for s in lsigma: jvbuffer = bgkern3(kerlen=math.ceil(s)*6+1, sigma=s, rho=rho) jvbuffer = fftn(jvbuffer, shape=fimg.shape, overwrite_x=True) * fimg fimg = ifftn(jvbuffer, overwrite_x=True) yield hessian3(np.real(fimg)) def eigval33(tensorfield): ''' Calculate the eigenvalues of massive 3x3 real symmetric matrices. ''' a11, a12, a13, a22, a23, a33 = tensorfield eps = 1e-50 b = a11 + eps d = a22 + eps j = a33 + eps c = - a12**2. - a13**2. - a23**2. + b * d + d * j + j* b d = - b * d * j + a23**2. * b + a12**2. * j - a13**2. * d + 2. * a13 * a12 * a23 b = - a11 - a22 - a33 - 3. * eps d = d + (2. * b**3. - 9. * b * c) / 27 c = b**2. / 3. - c c = c**3. c = c / 27 c[c < 0] = 0 c = np.sqrt(c) j = c ** (1./3.) c = c + (c==0).astype('float') d = -d /2. /c d[d>1] = 1 d[d<-1] = 1 d = np.real(np.arccos(d) / 3.) c = j * np.cos(d) d = j * np.sqrt(3.) * np.sin(d) b = -b / 3. j = -c - d + b d = -c + d + b b = 2. * c + b return b, j, d def oofftkernel(kernel_radius, r, sigma=1, ntype=1): eps = 1e-12 normalisation = 4/3 * np.pi * r**3 / (jv(1.5, 2*np.pi*r*eps) / eps ** (3/2)) / r**2 * \ (r / np.sqrt(2.*r*sigma - sigma**2)) ** ntype jvbuffer = normalisation * np.exp( (-2 * sigma**2 * np.pi**2 * kernel_radius**2) / (kernel_radius**(3/2) )) return (np.sin(2 * np.pi * r * kernel_radius) / (2 * np.pi * r * kernel_radius) - np.cos(2 * np.pi * r * kernel_radius)) * \ jvbuffer * np.sqrt( 1./ (np.pi**2 * r *kernel_radius )) def ooftensor(img, radii, memory_save=True): ''' type: oof, bg ''' # sigma = 1 # TODO: Pixel spacing eps = 1e-12 # ntype = 1 # The type of normalisation fimg = fftn(img, overwrite_x=True) shiftmat = ifftshiftedcoormatrix(fimg.shape) x, y, z = shiftmat x = x / fimg.shape[0] y = y / fimg.shape[1] z = z / fimg.shape[2] kernel_radius = np.sqrt(x ** 2 + y ** 2 + z ** 2) + eps # The distance from origin for r in radii: # Make the fourier convolutional kernel jvbuffer = oofftkernel(kernel_radius, r) * fimg if memory_save: # F11 buffer = ifftshiftedcoordinate(img.shape, 0) ** 2 * x * x * jvbuffer buffer = ifft(buffer, axis=0) buffer = ifft(buffer, axis=1) buffer = ifft(buffer, axis=2) f11 = buffer.copy() # F12 buffer = ifftshiftedcoordinate(img.shape, 0) * ifftshiftedcoordinate(img.shape, 1) * x * y * jvbuffer buffer = ifft(buffer, axis=0) buffer = ifft(buffer, axis=1) buffer = ifft(buffer, axis=2) f12 = buffer.copy() # F13 buffer = ifftshiftedcoordinate(img.shape, 0) * ifftshiftedcoordinate(img.shape, 2) * x * z * jvbuffer buffer = ifft(buffer, axis=0) buffer = ifft(buffer, axis=1) buffer = ifft(buffer, axis=2) f13 = buffer.copy() # F22 buffer = ifftshiftedcoordinate(img.shape, 1) ** 2 * y ** 2 * jvbuffer buffer = ifft(buffer, axis=0) buffer = ifft(buffer, axis=1) buffer = ifft(buffer, axis=2) f22 = buffer.copy() # F23 buffer = ifftshiftedcoordinate(img.shape, 1) * ifftshiftedcoordinate(img.shape, 2) * y * z * jvbuffer buffer = ifft(buffer, axis=0) buffer = ifft(buffer, axis=1) buffer = ifft(buffer, axis=2) f23 = buffer.copy() # F33 buffer = ifftshiftedcoordinate(img.shape, 2) * ifftshiftedcoordinate(img.shape, 2) * z * z * jvbuffer buffer = ifft(buffer, axis=0) buffer = ifft(buffer, axis=1) buffer = ifft(buffer, axis=2) f33 = buffer.copy() else: f11 = np.real(ifftn(x * x * jvbuffer)) f12 = np.real(ifftn(x * y * jvbuffer)) f13 = np.real(ifftn(x * z * jvbuffer)) f22 = np.real(ifftn(y * y * jvbuffer)) f23 = np.real(ifftn(y * z * jvbuffer)) f33 = np.real(ifftn(z * z * jvbuffer)) yield [f11, f12, f13, f22, f23, f33] # The dimension is a vector specifying the size of the returned coordinate # matrices. The number of output argument is equals to the dimensionality # of the vector "dimension". All the dimension is starting from "1" def ifftshiftedcoormatrix(shape): shape = np.asarray(shape) p = np.floor(np.asarray(shape) / 2).astype('int') coord = [] for i in range(shape.size): a = np.hstack((np.arange(p[i], shape[i]), np.arange(0, p[i]))) - p[i] - 1. repmatpara = np.ones((shape.size,)).astype('int') repmatpara[i] = shape[i] A = a.reshape(repmatpara) repmatpara = shape.copy() repmatpara[i] = 1 coord.append(np.tile(A, repmatpara)) return coord def ifftshiftedcoordinate(shape, axis): shape = np.asarray(shape) p = np.floor(np.asarray(shape) / 2).astype('int') a = (np.hstack((np.arange(p[axis], shape[axis]), np.arange(0, p[axis]))) - p[axis] - 1.).astype('float') a /= shape[axis].astype('float') reshapepara = np.ones((shape.size,)).astype('float'); reshapepara[axis] = shape[axis]; A = a.reshape(reshapepara); repmatpara = shape.copy(); repmatpara[axis] = 1; return np.tile(A, repmatpara) def nonmaximal_suppression3(img, evl, evt, radius, threshold=0): ''' Non-maximal suppression with oof eigen vector img: The input image or filter response V: The eigenvector generated by anisotropic filtering algorithm radius: the radius to consider for suppression ''' # THE METHOD with ROTATED STENCILS -- Deprecated for now # basevec = np.asarray([0., 0., 1.]) # basestencils = np.asarray([[1., 0., 0.], # [1./np.sqrt(2.), 1./np.sqrt(2.), 0.], # [0., 1., 0.], # [-1./np.sqrt(2.), 1./np.sqrt(2.), 0.], # [-1, 0., 0.], # [-1./np.sqrt(2.), -1./np.sqrt(2.), 0.], # [0., -1, 0.], # [1./np.sqrt(2.), -1./np.sqrt(2.), 0.]]) # fg = np.argwhere(img > threshold) # nvox = fg.shape[0] # imgnew = img.copy() # for idx in fg: # ev = V[idx[0], idx[1], idx[2], :, -1] # # Get rotation matrix (http://math.stackexchange.com/questions/180418/calculate-rotation-matrix-to-align-vector-a-to-vector-b-in-3d) # v = np.cross(basevec, ev) # s = np.linalg.norm(v) # c = np.dot(basevec, ev) # skw = np.asarray([[0., -v[2], v[1]], # [v[2], 0, -v[0]], # [-v[1], v[0], 0.]]) # R = np.identity(3) + skw + skw**2 * ((1.-c)/s**2) # # Rotate the stencils # rotstencils = np.dot(R, basestencils.T) # # Get neighbours based on the rotated stencils # neighbours = rotstencils.T + np.tile(idx, (basestencils.shape[0], 1)) # # suppress the current image if it is not local maxima # neighbourvox = [img[n[0], n[1], n[2]] for n in neighbours if ] # TODO check in bound # if np.any(neighbourvox > img[idx[0], idx[1], idx[2]]): # imgnew[idx[0], idx[1], idx[2]] = 0. suppressed = img.copy() suppressed[suppressed <= threshold] = 0 suppressed_ctr = -1 while suppressed_ctr is not 0: suppressed_ctr = 0 fgidx = np.argwhere(suppressed > threshold) # Find foreground voxels while fgidx.shape[0] > 0: randidx = np.random.randint(0, fgidx.shape[0]) v = fgidx[randidx, :] # Randomly choose a foreground voxel fgidx = np.delete(fgidx, randidx, 0) e = evt[v[0], v[1], v[2], :, 0] # The primary eigenvector on v # Select the voxels on the orthogonal plane of eigenvector and within a distance to v vtile = np.tile(v, (fgidx.shape[0], 1)) etile = np.tile(e, (fgidx.shape[0], 1)) cond1 = np.abs(etile * (fgidx - vtile)).sum(axis=-1) < (1.5 * np.sqrt(6) / 4.) # http://math.stackexchange.com/questions/82151/find-the-equation-of-the-plane-passing-through-a-point-and-a-vector-orthogonal radius = (evl[v[0], v[1], v[2], :]).sum() cond2 = np.linalg.norm(vtile - fgidx, axis=-1) < radius cond = np.logical_and(cond1, cond2) l = fgidx[cond, :] if l.size == 0: continue lv = np.asarray([suppressed[l[i, 0], l[i, 1], l[i, 2]] for i in range(l.shape[0])]) # The voxel values of the voxels in l if lv.max() > suppressed[v[0], v[1], v[2]]: suppressed[v[0], v[1], v[2]] = 0 suppressed_ctr += 1 return suppressed
15,197
34.180556
217
py
rivuletpy
rivuletpy-master/filtering/riveal.py
import numpy as np import math import skfmm from tqdm import tqdm from scipy.ndimage.morphology import binary_dilation from keras.models import Sequential from keras.layers import Dense, Activation, Flatten from keras.layers import Convolution2D, MaxPooling2D from keras.layers.noise import GaussianDropout, GaussianNoise from keras.layers.advanced_activations import SReLU def riveal(img, swc, K=9, nsample=8e4, epoch=20): print('-- oiginal image size: ', img.shape) K = math.floor(K) # Make sure K is integer to avoid warnings # Pad the image and swc margin = 3 * K img = padimg(img, margin) swc = padswc(swc, margin) # Make the skeleton distance transform print('-- Distance transform for swc...') dt, foreground_region = make_skdt(img.shape, swc, K) # Normalise data # img = standardise(img) # dt = standardise(dt) img /= img.max() # dt /= dt.max() # Make the confident region print('==swc shape:', swc.shape) print('-- Making the confidence regions...(1/4)') high_conf_region = make_conf_region( img.shape, swc, K, low_conf=0.5, high_conf=1.) # print('-- Making the confidence regions...(2/4)') # mid_conf_region = make_conf_region(img.shape, swc, K, # low_conf=0.25, high_conf=0.5) print('-- Making the confidence regions...(3/4)') low_conf_region = make_conf_region( img.shape, swc, K, low_conf=0., high_conf=0.25) # # Fill only the central part of background region print('-- Making the confidence regions...(4/4)') background_region = np.zeros(img.shape) bg = np.logical_not(foreground_region) bg = np.logical_and(foreground_region, img > 0) for i in range(3): bg = binary_dilation(bg) background_region[margin:-margin, margin:-margin, margin:-margin] = bg[ margin:-margin, margin:-margin, margin:-margin] from matplotlib import pyplot as plt plt.subplot(3, 1, 1) plt.imshow(high_conf_region.max(axis=-1)) plt.title('high conf') plt.subplot(3, 1, 2) plt.imshow(low_conf_region.max(axis=-1)) plt.title('low conf') plt.subplot(3, 1, 3) plt.imshow(background_region.max(axis=-1)) plt.title('bg') plt.show() # Randomly sample 2.5D blocks from the include region print('-- Sampling blocks') x1, y1 = sample_block(img, dt, high_conf_region, K, math.ceil(nsample * 0.75)) # x2, y2 = sample_block(img, dt, mid_conf_region, # K, math.ceil(nsample * 0.2)) x3, y3 = sample_block(img, dt, low_conf_region, K, math.ceil(nsample * 0.1)) y3.fill(0.) x4, y4 = sample_block(img, dt, background_region, K, math.ceil(nsample * 0.15)) y4.fill(0.) train_x = np.vstack((x1, x3, x4)) train_y = np.vstack((y1, y3, y4)) # Build the CNN with keras+tensorflow print('--Training CNN...') model = traincnn(train_x, train_y, K, epoch) # Make the prediction within an area larger than # the segmentation of the image print('-- Predicting...') bimg = img > 0 for i in range(6): bimg = binary_dilation(bimg) include_region = bimg > 0 include_idx = np.argwhere(include_region) nidx = include_idx.shape[0] predict_x = np.zeros((nsample, 2 * K + 1, 2 * K + 1, 3)) rest = nidx resultimg = np.zeros(img.shape) pbar = tqdm(total=nidx) # Predict every batch of blocks while rest > 0: startidx = -rest endidx = -rest + nsample if -rest + nsample < nidx else nidx rest -= nsample # Write the value to each include voxel for i, gidx in enumerate(range(int(startidx), int(endidx))): bx, by, bz = include_idx[gidx, :] predict_x[i, :, :, 0] = img[bx - K:bx + K + 1, by - K:by + K + 1, bz] predict_x[i, :, :, 1] = img[bx - K:bx + K + 1, by, bz - K:bz + K + 1] predict_x[i, :, :, 2] = img[bx, by - K:by + K + 1, bz - K:bz + K + 1] pd = model.predict(predict_x, batch_size=64, verbose=0).flatten() for i, gidx in enumerate(range(int(startidx), int(endidx))): bx, by, bz = include_idx[gidx, :] resultimg[bx, by, bz] = pd[i] pbar.update(nsample) resultimg = unpadimg(resultimg, margin) return resultimg def standardise(img, zeromean=True): img = (img - img.mean()) / img.std() return img def constrain_range(min, max, minlimit, maxlimit): return list( range(min if min > minlimit else minlimit, max if max < maxlimit else maxlimit)) def sample_block(img, dt, include_region, K, nsample): include_idx = np.argwhere(include_region) nidx = include_idx.shape[0] nsample = nidx if nsample > nidx else nsample idx2train = include_idx[np.random.choice(nidx, nsample), :] # Claim the memory for 2.5D blocks x = np.zeros((nsample, 2 * K + 1, 2 * K + 1, 3)) y = np.zeros((nsample, 1)) # Claim the memory for 2.5D blocks for i in range(idx2train.shape[0]): bx, by, bz = idx2train[i, :] x[i, :, :, 0] = img[bx - K:bx + K + 1, by - K:by + K + 1, bz] x[i, :, :, 1] = img[bx - K:bx + K + 1, by, bz - K:bz + K + 1] x[i, :, :, 2] = img[bx, by - K:by + K + 1, bz - K:bz + K + 1] y[i] = dt[bx, by, bz] return x, y def make_conf_region(imshape, swc, K, low_conf=0.0, high_conf=1.0): if low_conf != 0.0 or high_conf != 1.0: confswc = np.vstack((swc[np.logical_and(swc[:, 7] >= low_conf, swc[:, 7] <= high_conf), :])) region = np.zeros(imshape) r = math.ceil(K * 0.75) for i in range(confswc.shape[0]): node = confswc[i, :] n = [math.floor(n) for n in node[2:5]] rg1 = constrain_range(n[0] - r, n[0] + r + 1, 0, imshape[0]) rg2 = constrain_range(n[1] - r, n[1] + r + 1, 0, imshape[1]) rg3 = constrain_range(n[2] - r, n[2] + r + 1, 0, imshape[2]) X, Y, Z = np.meshgrid(rg1, rg2, rg3) # Skip if any node has empty box if len(X) == 0 or len(Y) == 0 or len(Z) == 0: continue region[X, Y, Z] = 1 # _, region = make_skdt(imshape, confswc, K) return region def make_skdt(imshape, swc, K, a=6): skimg = make_sk_img(imshape, swc) dm = math.floor(K / 2) dt = skfmm.distance(skimg, dx=1) include_region = dt <= 1.5 * dm zeromask = dt >= dm dt = np.exp(a * (1 - dt / dm)) - 1 dt[zeromask] = 0 return dt, include_region def makecnn(in_shape, K): model = Sequential() model.add( Convolution2D( 32, 3, 3, border_mode='same', input_shape=in_shape[1:])) model.add(SReLU()) model.add(MaxPooling2D(pool_size=(2, 2), dim_ordering='tf')) model.add(GaussianNoise(1)) model.add(GaussianDropout(0.4)) model.add(Convolution2D(32, 3, 3, border_mode='same')) model.add(SReLU()) model.add(MaxPooling2D(pool_size=(2, 2), dim_ordering='tf')) model.add(GaussianNoise(1)) model.add(GaussianDropout(0.4)) model.add(Flatten()) model.add(Dense(64)) model.add(SReLU()) model.add(Dense(64)) # model.add(SReLU()) model.add(Dense(1)) model.add(Activation('linear')) return model def traincnn(x, y, K, epoch): x = x.astype('float32') y = y.astype('float32') x /= x.max() y /= y.max() model = makecnn(x.shape, K) model.compile(loss='mse', optimizer='rmsprop') model.fit(x, y, batch_size=64, nb_epoch=epoch, validation_split=0.15, shuffle=True) return model def make_sk_img(imshape, swc): skimg = np.ones(imshape) for i in range(swc.shape[0]): node = [math.floor(n) for n in swc[i, 2:5]] skimg[node[0], node[1], node[2]] = 0 return skimg def padimg(img, margin): pimg = np.zeros((img.shape[0] + 2 * margin, img.shape[1] + 2 * margin, img.shape[2] + 2 * margin)) pimg[margin:margin + img.shape[0], margin:margin + img.shape[1], margin: margin + img.shape[2]] = img return pimg def unpadimg(img, margin): pimg = np.zeros((img.shape[0] - 2 * margin, img.shape[1] - 2 * margin, img.shape[2] - 2 * margin)) pimg = img[margin:margin + img.shape[0], margin:margin + img.shape[1], margin:margin + img.shape[2]] return pimg def padswc(swc, margin): swc[:, 2:5] = swc[:, 2:5] + margin return swc
8,673
32.233716
78
py
rivuletpy
rivuletpy-master/filtering/__init__.py
0
0
0
py
rivuletpy
rivuletpy-master/filtering/morphology.py
import numpy as np from scipy.ndimage import gaussian_filter1d from scipy.ndimage.filters import laplace try: from skimage import filters except ImportError: from skimage import filter as filters from tqdm import tqdm from functools import reduce from scipy.interpolate import RegularGridInterpolator import skfmm def ssmdt(dt, ssmiter): dt = ssm(dt, anisotropic=True, iterations=ssmiter) dt[dt < filters.threshold_otsu(dt)] = 0 dt = skfmm.distance(dt, dx=5e-2) dt = skfmm.distance(np.logical_not(dt), dx=5e-3) dt[dt > 0.04] = 0.04 dt = dt.max() - dt dt[dt <= 0.038] = 0 return dt def ssm(img, anisotropic=False, iterations=30): ''' Skeleton strength map img: the input image anisotropic: True if using anisotropic diffusion iterations: number of iterations to optimise the GVF ''' # f = gaussian_gradient_magnitude(img, 1) # f = 1 - gimg # Inverted version of the smoothed # gradient of the distance transform gvfmap = gvf(img, mu=0.001, iterations=iterations, anisotropic=anisotropic) shifts = [-1, 1, -1, -1, -1, 1] axis = [1, 1, 2, 2, 3, 3] shiftmat = np.zeros(gvfmap.shape) f = np.zeros(img.shape) # reuse f for saving the SSM for i, (s, a) in enumerate(zip(shifts, axis)): # Only the orthogonal neighbours shiftmat.fill(0) shiftmat[a - 1, :, :, :] = s # Dot product gvf and neighbour displacement fields / # distance between neighbour f += np.sum(np.roll( gvfmap, s, axis=a) * shiftmat, axis=0) / np.linalg.norm( shiftmat, axis=0) f[np.isnan(f)] = 0 f[f < 0] = 0 return f def nonmax(img, sigma=2, threshold=0): ''' Finds directional local maxima in a gradient array, as used in the Canny edge detector, but made separately accessible here for greater flexibility. The result is a logical array with the value true where the gradient magnitude is a local maximum along the gradient direction. ''' # Get normalised gaussian gradients eps = 1e-12 gx = gaussian_filter1d(img, sigma, axis=0, order=1) gy = gaussian_filter1d(img, sigma, axis=1, order=1) gz = gaussian_filter1d(img, sigma, axis=2, order=1) gmag = np.sqrt(gx**2 + gy**2 + gz**2) gx = gx / (gmag + eps) gy = gy / (gmag + eps) gz = gz / (gmag + eps) standard_grid = (np.arange(gmag.shape[0]), np.arange(gmag.shape[1]), np.arange(gmag.shape[2])) ginterp = RegularGridInterpolator(standard_grid, gmag, bounds_error=False) # Interpolate the graident magnitudes idx = np.argwhere( img > threshold ) # Double-check if the original image should be used to check xidx = idx[:, 0] yidx = idx[:, 1] zidx = idx[:, 2] dx = gx[xidx, yidx, zidx] dy = gy[xidx, yidx, zidx] dz = gz[xidx, yidx, zidx] gmag_0 = gmag[xidx, yidx, zidx] gmag_1 = ginterp(np.stack((xidx + dx, yidx + dy, zidx + dz), axis=-1)) gmag_2 = ginterp(np.stack((xidx - dx, yidx - dy, zidx - dz), axis=-1)) # Suppress nonmax voxels keep = np.logical_and(gmag_0 > gmag_1, gmag_0 > gmag_2) gmag.fill(0) gmag[xidx, yidx, zidx] = keep.astype('float') return gmag def d(x): ''' The difference between a voxel and its six neighbours The boundary is messed, however it does not matter ''' diff = np.zeros((6, x.shape[0], x.shape[1], x.shape[2])) shifts = [-1, 1, -1, -1, -1, 1] axis = [0, 0, 1, 1, 2, 2] for i, (s, a) in enumerate(zip(shifts, axis)): diff[i] = np.roll(x, s, axis=a) - x return diff # The decreasing funciton for angles def g_all(u, v, w): G = np.zeros((6, u.shape[0], u.shape[1], u.shape[2])) # Result cvec = np.stack((u, v, w), axis=0) # The flow vector on central voxel shifts = [-1, 1, -1, -1, -1, 1] axis = [0, 0, 1, 1, 2, 2] for i, (s, a) in enumerate(zip(shifts, axis)): G[i] = g( cvec, np.stack( ( np.roll( u, s, axis=a), # The flow vector on the surronding voxel np.roll( v, s, axis=a), np.roll( w, s, axis=a)), axis=0)) return G # Calculate the G function between two vector fields def g(cvec, svec, K=1): cnorm = np.linalg.norm(cvec, axis=0) snorm = np.linalg.norm(svec, axis=0) t = np.sum(cvec * svec, axis=0) / (cnorm * snorm + 1e-12) t -= 1 t = np.exp(K * t) t[np.logical_or(cnorm == 0, snorm == 0)] = 0 return t # Divergence def div(x): """ compute the divergence of n-D scalar field `F` """ return reduce( np.add, np.gradient(x)) # http://stackoverflow.com/a/21134289/1890513 def gvf(f, mu=0.05, iterations=30, anisotropic=False, ignore_second_term=False): # Gradient vector flow # Translated from https://github.com/smistad/3D-Gradient-Vector-Flow-for-Matlab f = (f - f.min()) / (f.max() - f.min()) f = enforce_mirror_boundary( f) # Enforce the mirror conditions on the boundary dx, dy, dz = np.gradient(f) # Initialse with normal gradients ''' Initialise the GVF vectors following S3 in Yu, Zeyun, and Chandrajit Bajaj. "A segmentation-free approach for skeletonization of gray-scale images via anisotropic vector diffusion." CVPR, 2004. CVPR 2004. It only uses one of the surronding neighbours with the lowest intensity ''' magsq = dx**2 + dy**2 + dz**2 # Set up the initial vector field u = dx.copy() v = dy.copy() w = dz.copy() for i in tqdm(range(iterations)): # The boundary might not matter here # u = enforce_mirror_boundary(u) # v = enforce_mirror_boundary(v) # w = enforce_mirror_boundary(w) # Update the vector field if anisotropic: G = g_all(u, v, w) u += mu / 6. * div(np.sum(G * d(u), axis=0)) v += mu / 6. * div(np.sum(G * d(v), axis=0)) w += mu / 6. * div(np.sum(G * d(w), axis=0)) else: u += mu * 6 * laplace(u) v += mu * 6 * laplace(v) w += mu * 6 * laplace(w) if not ignore_second_term: u -= (u - dx) * magsq v -= (v - dy) * magsq w -= (w - dz) * magsq return np.stack((u, v, w), axis=0) def enforce_mirror_boundary(f): ''' This function enforces the mirror boundary conditions on the 3D input image f. The values of all voxels at the boundary is set to the values of the voxels 2 steps inward ''' N, M, O = f.shape # Indices in the middle xi = np.arange(1, M - 1) yi = np.arange(1, N - 1) zi = np.arange(1, O - 1) # Corners f[[0, -1], [0, -1], [0, -1]] = f[[2, -3], [2, -3], [2, -3]] # Edges f[np.ix_([0, -1], [0, -1], zi)] = f[np.ix_([2, -3], [2, -3], zi)] f[np.ix_(yi, [0, -1], [0, -1])] = f[np.ix_(yi, [2, -3], [2, -3])] f[np.ix_([0, -1], xi, [0, -1])] = f[np.ix_([2, -3], xi, [2, -3])] # Faces f[np.ix_([0, -1], xi, zi)] = f[np.ix_([2, -3], xi, zi)] f[np.ix_(yi, [0, -1], zi)] = f[np.ix_(yi, [2, -3], zi)] f[np.ix_(yi, xi, [0, -1])] = f[np.ix_(yi, xi, [2, -3])] return f
7,344
30.122881
110
py
Mr.Right
Mr.Right-main/main.py
import yaml import os import utils import warnings from argparse import ArgumentParser from torch import nn from pytorch_lightning import Trainer,seed_everything from pytorch_lightning import loggers as pl_loggers from pytorch_lightning.loggers import WandbLogger from pytorch_lightning.callbacks import ModelCheckpoint,LearningRateMonitor from data.data_module import TextToMultiDataModule from pltrainer import TextToMultiTrainer from functools import partial from models.model import TextToMultiModel from transformers import ( BertTokenizer, RobertaTokenizerFast ) warnings.filterwarnings("ignore") def main(args,config): seed_everything(config.seed) # tokenizer if args.pretrain == "ALBEF" or args.pretrain == "ViLT": tokenizer = BertTokenizer.from_pretrained( config.text_encoder, cache_dir= args.cache_dir, ) elif args.pretrain == "METER": tokenizer = RobertaTokenizerFast.from_pretrained( config.text_encoder, cache_dir= args.cache_dir, ) # dataset print("Create Dataset") data_module = TextToMultiDataModule(args,config,tokenizer) if args.mode == "test": data_module.prepare_data(test=config['test_file'],document=config['document']) else: data_module.prepare_data(train=config['train_file'],val=config['val_file'],test=config['test_file'],document=config['document']) data_module.setup() # mutli model print("Create multi modal") model = TextToMultiModel(tokenizer=tokenizer,config=config,args=args) pltrainer = TextToMultiTrainer(args,config,model,tokenizer) # logger wandb_logger = WandbLogger(name=args.wandb_task_name,project="multimodalembedding", entity=args.wandb_entity) checkpoint_callback = ModelCheckpoint( filename= '{loss:.2f}-{val_loss:.2f}-{multi_r1:.2f}', save_top_k=3, verbose=False, monitor='multi_r1', mode='max' ) lr_monitor = LearningRateMonitor(logging_interval='step') trainer_config = { "default_root_dir": args.save_checkpoint, "fast_dev_run": False, # "gradient_clip_val": config.gradient_clip_value, # "replace_sampler_ddp":False, "strategy": "ddp", "logger": wandb_logger, "gpus": args.num_gpus, "max_epochs": config["schedular"]["epochs"], # "max_steps": config["schedular"]["max_steps"], "auto_scale_batch_size": 'binsearch', "progress_bar_refresh_rate": 1, "precision": 16, "check_val_every_n_epoch": 10, "log_every_n_steps": 1, "flush_logs_every_n_steps": 1, "callbacks":[checkpoint_callback, lr_monitor], } trainer = Trainer(**trainer_config) if args.mode == "train": trainer.fit(pltrainer, data_module,ckpt_path=args.pl_checkpoint) elif args.mode == "test": trainer.test(pltrainer, data_module,ckpt_path=args.pl_checkpoint) if __name__ == '__main__': parser = ArgumentParser() parser.add_argument('--wandb_task_name', default='testing') parser.add_argument('--wandb_entity', default='multimodalembedding') parser.add_argument('--num_gpus', default=1, type=int) parser.add_argument("--config", default="configs/ALBEF.yaml", type=str) parser.add_argument("--cache_dir", default="cache/", type=str) parser.add_argument('--log_dir', default='logs/',type=str) parser.add_argument("--mode", default="train", type=str,choices=['train','test'],help='choose your mode') parser.add_argument("--pretrain", default="ALBEF", type=str,choices=['ALBEF','ViLT','MDETR','METER'],help='choose pretrain work') parser.add_argument("--embeds_feats", default="avg", type=str,choices=['cls','avg','iavg_tcls'],help='how to deal with text and image embeddings') parser.add_argument("--pickle_output", default="./", type=str,help='directory of testing pickle files') parser.add_argument("--test_output", default="output.json", type=str,help='json files of testing result') parser.add_argument("--save_checkpoint", default="checkpoints", type=str) parser.add_argument('--pl_checkpoint', default=None,type=str,help='Load pytorch lightning checkpoint') parser.add_argument('--batch_size', type=int, default=32,help='The batch size of each dataloader') parser.add_argument('--num_workers', type=int, default=8, help='The number of workers in the DataLoader') parser.add_argument('--shuffle', type=bool, default=True,help='Whether shuffle dataloader') parser.add_argument('--ctx_prediction', action='store_true', help='Whether do context prediction') parser.add_argument('--neg_matching', action='store_true', help='Whether do negative matching') parser.add_argument('--neg_matchingv2', action='store_true', help='Whether do negative matching version2') parser.add_argument('--test_rank', default=10, type=int, help='Step1. Contrastive -> rank -> Step2. Matching') parser.add_argument('--re_ranking', action='store_true', help='Whether do re ranking for matching') args = parser.parse_args() if not os.path.exists(args.log_dir): os.makedirs(args.log_dir, exist_ok=True) if not os.path.exists(args.cache_dir): os.makedirs(args.cache_dir, exist_ok=True) if not os.path.exists(args.save_checkpoint): os.makedirs(args.save_checkpoint, exist_ok=True) if args.pretrain == "ALBEF": args.config = "configs/ALBEF.yaml" elif args.pretrain == "ViLT": args.config = "configs/ViLT.yaml" elif args.pretrain == "MDETR": args.config = "configs/MDETR.yaml" elif args.pretrain == "METER": args.config = "configs/METER.yaml" with open(args.config) as f: config = yaml.safe_load(f) config = utils.AttrDict(config) print(args) print(config) main(args,config)
5,871
42.496296
150
py
Mr.Right
Mr.Right-main/compute_pickle.py
import pickle import os from argparse import ArgumentParser from metric import score def main(args): print("Load pickle...") path = args.pickle_intput multi_doc_path = os.path.join(path,'multimodal_documents.pickle') img_embd_path = os.path.join(path,'img_query.pickle') txt_embd_path = os.path.join(path,'txt_query.pickle') mr_embd_path = os.path.join(path,'multi_query.pickle') true_label_path = os.path.join(path,'labels.pickle') with open(multi_doc_path, 'rb') as handle: all_docs = pickle.load(handle) with open(img_embd_path, 'rb') as handle: all_img_queries = pickle.load(handle) with open(txt_embd_path, 'rb') as handle: all_txt_queries = pickle.load(handle) with open(mr_embd_path, 'rb') as handle: all_multi_queries = pickle.load(handle) with open(true_label_path, 'rb') as handle: all_queries_doc_ids = pickle.load(handle) print(f"Document size: {all_docs.shape}") print(f"Text-related query size: {all_txt_queries.shape}") print(f"Image-related query size: {all_img_queries.shape}") print(f"Mixed query size: {all_multi_queries.shape}") img_sims_matrix = all_img_queries @ all_docs.t() img_output_score = score(img_sims_matrix,all_queries_doc_ids) print(f"IR score:{img_output_score}") del img_sims_matrix txt_sims_matrix = all_txt_queries @ all_docs.t() txt_output_score = score(txt_sims_matrix,all_queries_doc_ids) print(f"TR score:{txt_output_score}") del txt_sims_matrix multi_sims_matrix = all_multi_queries @ all_docs.t() multi_output_score = score(multi_sims_matrix,all_queries_doc_ids) print(f"MR score:{multi_output_score}") if __name__ == '__main__': parser = ArgumentParser() parser.add_argument('--pickle_intput', default='pickle-albef/') args = parser.parse_args() print(args) main(args)
1,894
34.754717
69
py
Mr.Right
Mr.Right-main/utils.py
import json def prepare_pretrain_data(files): print("\nReading json files") image_text_pairs = [] for f in files: print(f"File: {f}",end="\r") image_text_pairs += json.load(open(f,'r')) return image_text_pairs class AttrDict(dict): def __init__(self, *args, **kwargs): super(AttrDict, self).__init__(*args, **kwargs) self.__dict__ = self
367
25.285714
55
py
Mr.Right
Mr.Right-main/pltrainer.py
import pdb import utils import json import pickle import torch import os import torch.nn.functional as F import torch.distributed as dist import pytorch_lightning as pl from metric import score from scheduler import CosineLRScheduler from tqdm import tqdm class TextToMultiTrainer(pl.LightningModule): def __init__(self, args, config, model, tokenizer, ): super().__init__() self.args = args self.config = config self.arg_opt = utils.AttrDict(config['optimizer']) self.arg_sche = utils.AttrDict(config['schedular']) self.model = model self.tokenizer = tokenizer self.automatic_optimization = False # pytorch lightning turn off Optimize self.step_size = 100 self.warmup_iterations = self.arg_sche.warmup_epochs*self.step_size self.save_hyperparameters() def training_step(self, train_batch, idx): opt = self.optimizers() opt.zero_grad() query = train_batch['query_str_tensor'] doc_text = train_batch['doc_str_tensor'] doc_image = train_batch['doc_image_tensor'] doc_id = train_batch['doc_id'] context_labels = train_batch.get('context_labels', None) loss_ita, loss_ctx_labels, loss_itm = self.model.forward( query, doc_text, doc_image, doc_id, context_labels, self.args.neg_matching, self.args.neg_matchingv2) loss = loss_ita + loss_ctx_labels + loss_itm self.manual_backward(loss) torch.nn.utils.clip_grad_norm_(self.model.parameters(), self.config.gradient_clip_value) opt.step() self.log("loss", loss, on_step=True, on_epoch=True, prog_bar=True, logger=True, sync_dist=True) self.log("loss_ita", loss_ita, on_step=True, on_epoch=True, prog_bar=True, logger=True, sync_dist=True) if self.args.ctx_prediction: self.log("loss_ctx", loss_ctx_labels, on_step=True, on_epoch=True, prog_bar=True, logger=True, sync_dist=True) if self.args.neg_matching: self.log("loss_itm", loss_itm, on_step=True, on_epoch=True, prog_bar=True, logger=True, sync_dist=True) sch = self.lr_schedulers() # step every `n` epochs if self.current_epoch==0 and self.global_step%self.step_size==0 and self.global_step<=self.warmup_iterations: sch.step(self.global_step//self.step_size) if self.trainer.is_last_batch: sch.step(self.current_epoch+self.arg_sche.warmup_epochs+1) return loss def validation_step(self, val_batch, idx, dataloader_idx=0): if dataloader_idx == 0: # query img_query = val_batch['img_query_str_tensor'] txt_query = val_batch['txt_query_str_tensor'] multi_query = val_batch['multi_query_str_tensor'] doc_id = val_batch['doc_id'] img_query['input_ids'] = img_query['input_ids'].view(img_query['input_ids'].shape[0],-1) img_query['attention_mask'] = img_query['attention_mask'].view(img_query['input_ids'].shape[0],-1) if "token_type_ids" in img_query.keys(): img_query["token_type_ids"] = img_query["token_type_ids"].view(img_query['input_ids'].shape[0],-1) txt_query['input_ids'] = txt_query['input_ids'].view(txt_query['input_ids'].shape[0],-1) txt_query['attention_mask'] = txt_query['attention_mask'].view(txt_query['input_ids'].shape[0],-1) if "token_type_ids" in txt_query.keys(): txt_query["token_type_ids"] = txt_query["token_type_ids"].view(txt_query['input_ids'].shape[0],-1) multi_query['input_ids'] = multi_query['input_ids'].view(multi_query['input_ids'].shape[0],-1) multi_query['attention_mask'] = multi_query['attention_mask'].view(multi_query['input_ids'].shape[0],-1) if "token_type_ids" in multi_query.keys(): multi_query["token_type_ids"] = multi_query["token_type_ids"].view(multi_query['input_ids'].shape[0],-1) img_query_embeds, img_query_feats = self.model.output_query_feats(img_query) txt_query_embeds, txt_query_feats = self.model.output_query_feats(txt_query) multi_query_embeds, multi_query_feats = self.model.output_query_feats(multi_query) result = { "img_query_text": val_batch['img_query_str'], "img_query_feats": img_query_feats, "txt_query_text": val_batch['txt_query_str'], "txt_query_feats": txt_query_feats, "multi_query_text": val_batch['multi_query_str'], "multi_query_feats": multi_query_feats, "docs_id": doc_id, } if self.args.re_ranking: result["img_query_embeds"] = img_query_embeds result["img_query_att"] = img_query['attention_mask'] result["txt_query_embeds"] = txt_query_embeds result["txt_query_att"] = txt_query['attention_mask'] result["multi_query_embeds"] = multi_query_embeds result["multi_query_att"] = multi_query['attention_mask'] elif dataloader_idx == 1: # document doc_text = val_batch['doc_str_tensor'] doc_id = val_batch['doc_id'] doc_image = val_batch['doc_image_tensor'] doc_text['input_ids'] = doc_text['input_ids'].view(doc_text['input_ids'].shape[0],-1) doc_text['attention_mask'] = doc_text['attention_mask'].view(doc_text['input_ids'].shape[0],-1) if "token_type_ids" in doc_text.keys(): doc_text["token_type_ids"] = doc_text["token_type_ids"].view(doc_text['input_ids'].shape[0],-1) docs_embeds,docs_feats,doc_masks = self.model.output_doc_feats(doc_text,doc_image) result = { "docs_text": val_batch['doc_str'], "docs_image": val_batch['image_path'], "docs_feats": docs_feats, "docs_id": doc_id } # ===== Context Prediction ===== if self.args.ctx_prediction: prediction_scores = self.model.context(docs_embeds) mean_prediction_scores = torch.mean(prediction_scores,1) top_context = torch.topk(mean_prediction_scores,20,dim=1) result["top_context"] = top_context # ===== Re-rank ===== if self.args.re_ranking: result["doc_embeds"] = docs_embeds result["doc_att"] = doc_masks return result def validation_epoch_end(self, validation_step_outputs): queries = validation_step_outputs[0] all_queries_doc_ids = torch.stack([feat for output in queries for feat in output["docs_id"]]) all_img_queries = torch.stack([feat for output in queries for feat in output["img_query_feats"]]) all_img_queries_text = [str(feat) for output in queries for feat in output["img_query_text"]] all_txt_queries = torch.stack([feat for output in queries for feat in output["txt_query_feats"]]) all_txt_queries_text = [str(feat) for output in queries for feat in output["txt_query_text"]] all_multi_queries = torch.stack([feat for output in queries for feat in output["multi_query_feats"]]) all_multi_queries_text = [str(feat) for output in queries for feat in output["multi_query_text"]] docs = validation_step_outputs[1] all_docs_ids = torch.stack([feat for output in docs for feat in output["docs_id"]]) all_docs = torch.stack([feat for output in docs for feat in output["docs_feats"]]) all_docs_captions = [str(feat) for output in docs for feat in output["docs_text"]] all_docs_images = [str(feat) for output in docs for feat in output["docs_image"]] if self.local_rank != None: # id all_queries_doc_ids_list = [torch.zeros_like(all_queries_doc_ids) for _ in range(dist.get_world_size())] dist.all_gather(all_queries_doc_ids_list, all_queries_doc_ids) all_queries_doc_ids = torch.cat(all_queries_doc_ids_list, dim=0) temp_all_queries_doc_ids = all_queries_doc_ids all_queries_doc_ids, rm_repeat_indices = self.unique(all_queries_doc_ids) # image_query all_img_queries_list = [torch.zeros_like(all_img_queries) for _ in range(dist.get_world_size())] dist.all_gather(all_img_queries_list, all_img_queries) all_img_queries = torch.cat(all_img_queries_list, dim=0) all_img_queries = all_img_queries[rm_repeat_indices] all_img_queries_text_list = [None for _ in range(dist.get_world_size())] dist.all_gather_object(all_img_queries_text_list, all_img_queries_text) all_img_queries_text = [text for queries_text in all_img_queries_text_list for text in queries_text] all_img_queries_text = [all_img_queries_text[i] for i in rm_repeat_indices.tolist()] # text_query all_txt_queries_list = [torch.zeros_like(all_txt_queries) for _ in range(dist.get_world_size())] dist.all_gather(all_txt_queries_list, all_txt_queries) all_txt_queries = torch.cat(all_txt_queries_list, dim=0) all_txt_queries = all_txt_queries[rm_repeat_indices] all_txt_queries_text_list = [None for _ in range(dist.get_world_size())] dist.all_gather_object(all_txt_queries_text_list, all_txt_queries_text) all_txt_queries_text = [text for queries_text in all_txt_queries_text_list for text in queries_text] all_txt_queries_text = [all_txt_queries_text[i] for i in rm_repeat_indices.tolist()] # multi_query all_multi_queries_list = [torch.zeros_like(all_multi_queries) for _ in range(dist.get_world_size())] dist.all_gather(all_multi_queries_list, all_multi_queries) all_multi_queries = torch.cat(all_multi_queries_list, dim=0) all_multi_queries = all_multi_queries[rm_repeat_indices] all_multi_queries_text_list = [None for _ in range(dist.get_world_size())] dist.all_gather_object(all_multi_queries_text_list, all_multi_queries_text) all_multi_queries_text = [text for queries_text in all_multi_queries_text_list for text in queries_text] all_multi_queries_text = [all_multi_queries_text[i] for i in rm_repeat_indices.tolist()] # multimodal_doc all_doc_ids_list = [torch.zeros_like(all_docs_ids) for _ in range(dist.get_world_size())] dist.all_gather(all_doc_ids_list, all_docs_ids) all_docs_ids = torch.cat(all_doc_ids_list, dim=0) all_docs_ids, rm_repeat_doc_indices = self.unique(all_docs_ids) all_docs_list = [torch.zeros_like(all_docs) for _ in range(dist.get_world_size())] dist.all_gather(all_docs_list, all_docs) all_docs = torch.cat(all_docs_list, dim=0) all_docs = all_docs[rm_repeat_doc_indices] all_docs_captions_list = [None for _ in range(dist.get_world_size())] dist.all_gather_object(all_docs_captions_list, all_docs_captions) all_docs_captions = [caption for docs_captions in all_docs_captions_list for caption in docs_captions] all_docs_captions = [all_docs_captions[i] for i in rm_repeat_doc_indices.tolist()] all_docs_image_list = [None for _ in range(dist.get_world_size())] dist.all_gather_object(all_docs_image_list, all_docs_images) all_docs_images = [image for docs_images in all_docs_image_list for image in docs_images] all_docs_images = [all_docs_images[i] for i in rm_repeat_doc_indices.tolist()] if self.args.re_ranking: all_docs_embeds = torch.stack([feat for output in docs for feat in output["doc_embeds"]]) all_docs_masks = torch.stack([feat for output in docs for feat in output["doc_att"]]) all_docs_embeds_list = [torch.zeros_like(all_docs_embeds) for _ in range(dist.get_world_size())] dist.all_gather(all_docs_embeds_list, all_docs_embeds) all_docs_embeds = torch.cat(all_docs_embeds_list, dim=0) all_docs_embeds = all_docs_embeds[all_docs_ids] all_docs_masks_list = [torch.zeros_like(all_docs_masks) for _ in range(dist.get_world_size())] dist.all_gather(all_docs_masks_list, all_docs_masks) all_docs_masks = torch.cat(all_docs_masks_list, dim=0) all_docs_masks = all_docs_masks[all_docs_ids] img_sims_matrix = all_img_queries @ all_docs.t() txt_sims_matrix = all_txt_queries @ all_docs.t() multi_sims_matrix = all_multi_queries @ all_docs.t() matrix_list = [img_sims_matrix,txt_sims_matrix,multi_sims_matrix] labels = F.one_hot(all_queries_doc_ids, len(all_docs)).to(self.device) if self.args.re_ranking: all_img_queries_embeds = torch.stack([feat for output in queries for feat in output["img_query_embeds"]]) all_img_queries_masks = torch.stack([feat for output in queries for feat in output["img_query_att"]]) all_txt_queries_embeds = torch.stack([feat for output in queries for feat in output["txt_query_embeds"]]) all_txt_queries_masks = torch.stack([feat for output in queries for feat in output["txt_query_att"]]) all_multi_queries_embeds = torch.stack([feat for output in queries for feat in output["multi_query_embeds"]]) all_multi_queries_masks = torch.stack([feat for output in queries for feat in output["multi_query_att"]]) score_matrix_i2m = torch.full((len(all_img_queries),len(all_docs)),-100.0).to(self.device) score_matrix_t2m = torch.full((len(all_txt_queries),len(all_docs)),-100.0).to(self.device) score_matrix_m2m = torch.full((len(all_multi_queries),len(all_docs)),-100.0).to(self.device) for type in range(3): temp_matrix = matrix_list[type] for i,sims in enumerate(tqdm(temp_matrix)): topk_sim, topk_idx = sims.topk(k=self.args.test_rank, dim=0) if type == 0: queries = all_img_queries_embeds[i].unsqueeze(0).repeat(self.args.test_rank,1,1) queries_mask = all_img_queries_masks[i].unsqueeze(0).repeat(self.args.test_rank,1) elif type == 1: queries = all_txt_queries_embeds[i].unsqueeze(0).repeat(self.args.test_rank,1,1) queries_mask = all_txt_queries_masks[i].unsqueeze(0).repeat(self.args.test_rank,1) else: queries = all_multi_queries_embeds[i].unsqueeze(0).repeat(self.args.test_rank,1,1) queries_mask = all_multi_queries_masks[i].unsqueeze(0).repeat(self.args.test_rank,1) docs = all_docs_embeds[topk_idx] docs_masks = all_docs_masks[topk_idx] itm_logits = self.model.matching_classifier( query_embeds=queries, query_attns=queries_mask, multi_embeds=docs, multi_attns=docs_masks ) if type == 0: score_matrix_i2m[i,topk_idx] = itm_logits[:,1] elif type == 1: score_matrix_t2m[i,topk_idx] = itm_logits[:,1] else: score_matrix_m2m[i,topk_idx] = itm_logits[:,1] if type == 0: img_sims_matrix = score_matrix_i2m img_sims_matrix = img_sims_matrix.cpu() elif type == 1: txt_sims_matrix = score_matrix_t2m txt_sims_matrix = txt_sims_matrix.cpu() else: multi_sims_matrix = score_matrix_m2m multi_sims_matrix = multi_sims_matrix.cpu() matrix_list = [img_sims_matrix,txt_sims_matrix,multi_sims_matrix] loss_i2m = -torch.sum(F.log_softmax(img_sims_matrix / self.model.temp, dim=1)*labels,dim=1).mean() loss_t2m = -torch.sum(F.log_softmax(txt_sims_matrix / self.model.temp, dim=1)*labels,dim=1).mean() loss_m2m = -torch.sum(F.log_softmax(multi_sims_matrix / self.model.temp, dim=1)*labels,dim=1).mean() img_output_score = score(img_sims_matrix,all_queries_doc_ids) txt_output_score = score(txt_sims_matrix,all_queries_doc_ids) multi_output_score = score(multi_sims_matrix,all_queries_doc_ids) val_loss = (loss_i2m + loss_t2m + loss_m2m) / 3 if self.args.ctx_prediction: loss_ctx_pred = torch.stack([output["loss_ctx_pred"] for output in validation_step_outputs]).mean() val_loss = (val_loss+loss_ctx_pred) / 2 self.log("val_loss_t2m", loss_t2m, on_epoch=True, prog_bar=True, logger=True, sync_dist=True) self.log("val_loss_ctx_pred", loss_ctx_pred, on_epoch=True, prog_bar=True, logger=True, sync_dist=True) self.log("val_loss", val_loss, on_epoch=True, prog_bar=True, logger=True, sync_dist=True) self.log("img_r1", img_output_score['r1'], on_epoch=True, prog_bar=False, logger=True, sync_dist=True) self.log("img_r5", img_output_score['r5'], on_epoch=True, prog_bar=False, logger=True, sync_dist=True) self.log("img_r10", img_output_score['r10'], on_epoch=True, prog_bar=False, logger=True, sync_dist=True) self.log("img_r_mean", img_output_score['r_mean'], on_epoch=True, prog_bar=False, logger=True, sync_dist=True) self.log("img_mrr10", img_output_score['mrr10'], on_epoch=True, prog_bar=True, logger=True, sync_dist=True) self.log("txt_r1", txt_output_score['r1'], on_epoch=True, prog_bar=True, logger=True, sync_dist=True) self.log("txt_r5", txt_output_score['r5'], on_epoch=True, prog_bar=False, logger=True, sync_dist=True) self.log("txt_r10", txt_output_score['r10'], on_epoch=True, prog_bar=False, logger=True, sync_dist=True) self.log("txt_r_mean", txt_output_score['r_mean'], on_epoch=True, prog_bar=False, logger=True, sync_dist=True) self.log("txt_mrr10", txt_output_score['mrr10'], on_epoch=True, prog_bar=True, logger=True, sync_dist=True) self.log("multi_r1", multi_output_score['r1'], on_epoch=True, prog_bar=False, logger=True, sync_dist=True) self.log("multi_r5", multi_output_score['r5'], on_epoch=True, prog_bar=False, logger=True, sync_dist=True) self.log("multi_r10", multi_output_score['r10'], on_epoch=True, prog_bar=False, logger=True, sync_dist=True) self.log("multi_r_mean", multi_output_score['r_mean'], on_epoch=True, prog_bar=False, logger=True, sync_dist=True) self.log("multi_mrr10", multi_output_score['mrr10'], on_epoch=True, prog_bar=True, logger=True, sync_dist=True) return { "val_loss":val_loss, "loss_t2m":loss_t2m, "loss_i2m":loss_i2m, "loss_m2m":loss_m2m, "img_r1":img_output_score['r1'], "img_r5":img_output_score['r5'], "img_r10":img_output_score['r10'], "img_r_mean":img_output_score['r_mean'], "img_mrr10":img_output_score['mrr10'], "txt_r1":txt_output_score['r1'], "txt_r5":txt_output_score['r5'], "txt_r10":txt_output_score['r10'], "txt_r_mean":txt_output_score['r_mean'], "txt_mrr10":txt_output_score['mrr10'], "multi_r1":multi_output_score['r1'], "multi_r5":multi_output_score['r5'], "multi_r10":multi_output_score['r10'], "multi_r_mean":multi_output_score['r_mean'], "multi_mrr10":multi_output_score['mrr10'] } def test_step(self, test_batch, idx, dataloader_idx=0): if dataloader_idx == 0: # query img_query = test_batch['img_query_str_tensor'] txt_query = test_batch['txt_query_str_tensor'] multi_query = test_batch['multi_query_str_tensor'] doc_id = test_batch['doc_id'] img_query['input_ids'] = img_query['input_ids'].view(img_query['input_ids'].shape[0],-1) img_query['attention_mask'] = img_query['attention_mask'].view(img_query['input_ids'].shape[0],-1) if "token_type_ids" in img_query.keys(): img_query["token_type_ids"] = img_query["token_type_ids"].view(img_query['input_ids'].shape[0],-1) txt_query['input_ids'] = txt_query['input_ids'].view(txt_query['input_ids'].shape[0],-1) txt_query['attention_mask'] = txt_query['attention_mask'].view(txt_query['input_ids'].shape[0],-1) if "token_type_ids" in txt_query.keys(): txt_query["token_type_ids"] = txt_query["token_type_ids"].view(txt_query['input_ids'].shape[0],-1) multi_query['input_ids'] = multi_query['input_ids'].view(multi_query['input_ids'].shape[0],-1) multi_query['attention_mask'] = multi_query['attention_mask'].view(multi_query['input_ids'].shape[0],-1) if "token_type_ids" in multi_query.keys(): multi_query["token_type_ids"] = multi_query["token_type_ids"].view(multi_query['input_ids'].shape[0],-1) img_query_embeds, img_query_feats = self.model.output_query_feats(img_query) txt_query_embeds, txt_query_feats = self.model.output_query_feats(txt_query) multi_query_embeds, multi_query_feats = self.model.output_query_feats(multi_query) result = { "img_query_text": test_batch['img_query_str'], "img_query_feats": img_query_feats, "txt_query_text": test_batch['txt_query_str'], "txt_query_feats": txt_query_feats, "multi_query_text": test_batch['multi_query_str'], "multi_query_feats": multi_query_feats, "docs_id": doc_id, } if self.args.re_ranking: result["img_query_embeds"] = img_query_embeds result["img_query_att"] = img_query['attention_mask'] result["txt_query_embeds"] = txt_query_embeds result["txt_query_att"] = txt_query['attention_mask'] result["multi_query_embeds"] = multi_query_embeds result["multi_query_att"] = multi_query['attention_mask'] elif dataloader_idx == 1: # document doc_text = test_batch['doc_str_tensor'] doc_id = test_batch['doc_id'] doc_image = test_batch['doc_image_tensor'] doc_text['input_ids'] = doc_text['input_ids'].view(doc_text['input_ids'].shape[0],-1) doc_text['attention_mask'] = doc_text['attention_mask'].view(doc_text['input_ids'].shape[0],-1) if "token_type_ids" in doc_text.keys(): doc_text["token_type_ids"] = doc_text["token_type_ids"].view(doc_text['input_ids'].shape[0],-1) docs_embeds,docs_feats,doc_masks = self.model.output_doc_feats(doc_text,doc_image) result = { "docs_text": test_batch['doc_str'], "docs_image": test_batch['image_path'], "docs_feats": docs_feats, "docs_id": doc_id } # ===== Context Prediction ===== if self.args.ctx_prediction: prediction_scores = self.model.context(docs_embeds) mean_prediction_scores = torch.mean(prediction_scores,1) top_context = torch.topk(mean_prediction_scores,20,dim=1) result["top_context"] = top_context # ===== Re-rank ===== if self.args.re_ranking: result["doc_embeds"] = docs_embeds result["doc_att"] = doc_masks return result def test_epoch_end(self, test_step_outputs): queries = test_step_outputs[0] all_queries_doc_ids = torch.stack([feat for output in queries for feat in output["docs_id"]]) all_img_queries = torch.stack([feat for output in queries for feat in output["img_query_feats"]]) all_img_queries_text = [str(feat) for output in queries for feat in output["img_query_text"]] all_txt_queries = torch.stack([feat for output in queries for feat in output["txt_query_feats"]]) all_txt_queries_text = [str(feat) for output in queries for feat in output["txt_query_text"]] all_multi_queries = torch.stack([feat for output in queries for feat in output["multi_query_feats"]]) all_multi_queries_text = [str(feat) for output in queries for feat in output["multi_query_text"]] docs = test_step_outputs[1] all_docs_ids = torch.stack([feat for output in docs for feat in output["docs_id"]]) all_docs = torch.stack([feat for output in docs for feat in output["docs_feats"]]) all_docs_captions = [str(feat) for output in docs for feat in output["docs_text"]] all_docs_images = [str(feat) for output in docs for feat in output["docs_image"]] if self.local_rank != None: # id all_queries_doc_ids_list = [torch.zeros_like(all_queries_doc_ids) for _ in range(dist.get_world_size())] dist.all_gather(all_queries_doc_ids_list, all_queries_doc_ids) all_queries_doc_ids = torch.cat(all_queries_doc_ids_list, dim=0) temp_all_queries_doc_ids = all_queries_doc_ids all_queries_doc_ids, rm_repeat_indices = self.unique(all_queries_doc_ids) # image_query all_img_queries_list = [torch.zeros_like(all_img_queries) for _ in range(dist.get_world_size())] dist.all_gather(all_img_queries_list, all_img_queries) all_img_queries = torch.cat(all_img_queries_list, dim=0) all_img_queries = all_img_queries[rm_repeat_indices] all_img_queries_text_list = [None for _ in range(dist.get_world_size())] dist.all_gather_object(all_img_queries_text_list, all_img_queries_text) all_img_queries_text = [text for queries_text in all_img_queries_text_list for text in queries_text] all_img_queries_text = [all_img_queries_text[i] for i in rm_repeat_indices.tolist()] # text_query all_txt_queries_list = [torch.zeros_like(all_txt_queries) for _ in range(dist.get_world_size())] dist.all_gather(all_txt_queries_list, all_txt_queries) all_txt_queries = torch.cat(all_txt_queries_list, dim=0) all_txt_queries = all_txt_queries[rm_repeat_indices] all_txt_queries_text_list = [None for _ in range(dist.get_world_size())] dist.all_gather_object(all_txt_queries_text_list, all_txt_queries_text) all_txt_queries_text = [text for queries_text in all_txt_queries_text_list for text in queries_text] all_txt_queries_text = [all_txt_queries_text[i] for i in rm_repeat_indices.tolist()] # multi_query all_multi_queries_list = [torch.zeros_like(all_multi_queries) for _ in range(dist.get_world_size())] dist.all_gather(all_multi_queries_list, all_multi_queries) all_multi_queries = torch.cat(all_multi_queries_list, dim=0) all_multi_queries = all_multi_queries[rm_repeat_indices] all_multi_queries_text_list = [None for _ in range(dist.get_world_size())] dist.all_gather_object(all_multi_queries_text_list, all_multi_queries_text) all_multi_queries_text = [text for queries_text in all_multi_queries_text_list for text in queries_text] all_multi_queries_text = [all_multi_queries_text[i] for i in rm_repeat_indices.tolist()] # multimodal_doc all_doc_ids_list = [torch.zeros_like(all_docs_ids) for _ in range(dist.get_world_size())] dist.all_gather(all_doc_ids_list, all_docs_ids) temp_all_queries_list = all_doc_ids_list all_docs_ids = torch.cat(all_doc_ids_list, dim=0) temp_all_queries_list_fix = all_docs_ids all_docs_ids, rm_repeat_doc_indices = self.unique(all_docs_ids) all_docs_list = [torch.zeros_like(all_docs) for _ in range(dist.get_world_size())] dist.all_gather(all_docs_list, all_docs) all_docs = torch.cat(all_docs_list, dim=0) temp_all_docs = all_docs all_docs = all_docs[rm_repeat_doc_indices] all_docs_captions_list = [None for _ in range(dist.get_world_size())] dist.all_gather_object(all_docs_captions_list, all_docs_captions) all_docs_captions = [caption for docs_captions in all_docs_captions_list for caption in docs_captions] temp_all_docs_captions = all_docs_captions all_docs_captions = [all_docs_captions[i] for i in rm_repeat_doc_indices.tolist()] all_docs_image_list = [None for _ in range(dist.get_world_size())] dist.all_gather_object(all_docs_image_list, all_docs_images) all_docs_images = [image for docs_images in all_docs_image_list for image in docs_images] all_docs_images = [all_docs_images[i] for i in rm_repeat_doc_indices.tolist()] if self.args.re_ranking: all_docs_embeds = torch.stack([feat for output in docs for feat in output["doc_embeds"]]) all_docs_masks = torch.stack([feat for output in docs for feat in output["doc_att"]]) all_docs_embeds_list = [torch.zeros_like(all_docs_embeds) for _ in range(dist.get_world_size())] dist.all_gather(all_docs_embeds_list, all_docs_embeds) all_docs_embeds = torch.cat(all_docs_embeds_list, dim=0) all_docs_embeds = all_docs_embeds[all_docs_ids] all_docs_masks_list = [torch.zeros_like(all_docs_masks) for _ in range(dist.get_world_size())] dist.all_gather(all_docs_masks_list, all_docs_masks) all_docs_masks = torch.cat(all_docs_masks_list, dim=0) all_docs_masks = all_docs_masks[all_docs_ids] if self.local_rank == 0: output_dir = self.args.pickle_output doc_path = os.path.join(output_dir,"multimodal_documents.pickle") img_path = os.path.join(output_dir,"img_query.pickle") txt_path = os.path.join(output_dir,"txt_query.pickle") multi_path = os.path.join(output_dir,"multi_query.pickle") labels_path = os.path.join(output_dir,"labels.pickle") with open(doc_path, 'wb') as handle: pickle.dump(all_docs, handle, protocol=pickle.HIGHEST_PROTOCOL) with open(img_path, 'wb') as handle: pickle.dump(all_img_queries, handle, protocol=pickle.HIGHEST_PROTOCOL) with open(txt_path, 'wb') as handle: pickle.dump(all_txt_queries, handle, protocol=pickle.HIGHEST_PROTOCOL) with open(multi_path, 'wb') as handle: pickle.dump(all_multi_queries, handle, protocol=pickle.HIGHEST_PROTOCOL) with open(labels_path, 'wb') as handle: pickle.dump(all_queries_doc_ids, handle, protocol=pickle.HIGHEST_PROTOCOL) print("Finish saving pickle. Now you can compute the score.") return {"img":True} img_sims_matrix = all_img_queries @ all_docs.t() txt_sims_matrix = all_txt_queries @ all_docs.t() multi_sims_matrix = all_multi_queries @ all_docs.t() matrix_list = [img_sims_matrix,txt_sims_matrix,multi_sims_matrix] if self.args.re_ranking: all_img_queries_embeds = torch.stack([feat for output in queries for feat in output["img_query_embeds"]]) all_img_queries_masks = torch.stack([feat for output in queries for feat in output["img_query_att"]]) all_txt_queries_embeds = torch.stack([feat for output in queries for feat in output["txt_query_embeds"]]) all_txt_queries_masks = torch.stack([feat for output in queries for feat in output["txt_query_att"]]) all_multi_queries_embeds = torch.stack([feat for output in queries for feat in output["multi_query_embeds"]]) all_multi_queries_masks = torch.stack([feat for output in queries for feat in output["multi_query_att"]]) score_matrix_i2m = torch.full((len(all_img_queries),len(all_docs)),-100.0).to(self.device) score_matrix_t2m = torch.full((len(all_txt_queries),len(all_docs)),-100.0).to(self.device) score_matrix_m2m = torch.full((len(all_multi_queries),len(all_docs)),-100.0).to(self.device) for type in range(3): temp_matrix = matrix_list[type] for i,sims in enumerate(tqdm(temp_matrix)): topk_sim, topk_idx = sims.topk(k=self.args.test_rank, dim=0) if type == 0: queries = all_img_queries_embeds[i].unsqueeze(0).repeat(self.args.test_rank,1,1) queries_mask = all_img_queries_masks[i].unsqueeze(0).repeat(self.args.test_rank,1) elif type == 1: queries = all_txt_queries_embeds[i].unsqueeze(0).repeat(self.args.test_rank,1,1) queries_mask = all_txt_queries_masks[i].unsqueeze(0).repeat(self.args.test_rank,1) else: queries = all_multi_queries_embeds[i].unsqueeze(0).repeat(self.args.test_rank,1,1) queries_mask = all_multi_queries_masks[i].unsqueeze(0).repeat(self.args.test_rank,1) docs = all_docs_embeds[topk_idx] docs_masks = all_docs_masks[topk_idx] itm_logits = self.model.matching_classifier( query_embeds=queries, query_attns=queries_mask, multi_embeds=docs, multi_attns=docs_masks ) if type == 0: score_matrix_i2m[i,topk_idx] = itm_logits[:,1] elif type == 1: score_matrix_t2m[i,topk_idx] = itm_logits[:,1] else: score_matrix_m2m[i,topk_idx] = itm_logits[:,1] if type == 0: img_sims_matrix = score_matrix_i2m img_sims_matrix = img_sims_matrix.cpu() elif type == 1: txt_sims_matrix = score_matrix_t2m txt_sims_matrix = txt_sims_matrix.cpu() else: multi_sims_matrix = score_matrix_m2m multi_sims_matrix = multi_sims_matrix.cpu() matrix_list = [img_sims_matrix,txt_sims_matrix,multi_sims_matrix] # img_output_score = score(img_sims_matrix,labels) # txt_output_score = score(txt_sims_matrix,labels) # multi_output_score = score(multi_sims_matrix,labels) img_output_score = score_v2(img_sims_matrix,all_queries_doc_ids) txt_output_score = score_v2(txt_sims_matrix,all_queries_doc_ids) multi_output_score = score_v2(multi_sims_matrix,all_queries_doc_ids) # output context if self.args.ctx_prediction: all_topk_context_values = torch.stack([feat for output in docs for feat in output["top_context"].values]) all_topk_context_indices = torch.stack([feat for output in docs for feat in output["top_context"].indices]) output_data = {"img_score":img_output_score,"txt_score":txt_output_score,"multi_score":multi_output_score,"result":[]} # output result for index,doc_scores in enumerate(img_sims_matrix): temp_dict = {"img":{},"txt":{},"multi":{}} for type in range(3): if type == 0: inds = torch.argsort(img_sims_matrix[index], descending=True)[:10].tolist() temp_all_queries_text = all_img_queries_text elif type == 1: inds = torch.argsort(txt_sims_matrix[index], descending=True)[:10].tolist() temp_all_queries_text = all_txt_queries_text else: inds = torch.argsort(multi_sims_matrix[index], descending=True)[:10].tolist() temp_all_queries_text = all_multi_queries_text temp = {"id": all_queries_doc_ids[index].item(), "query":temp_all_queries_text[index], "true_doc": all_docs_captions[all_queries_doc_ids[index]], "true_image":all_docs_images[all_queries_doc_ids[index]], } rank = 1e20 true_id = all_queries_doc_ids[index] if inds[0] == true_id: temp["correct"] = True else: temp["correct"] = False for rank,top_id in enumerate(inds): dic_key_cap = str(rank)+"_doc" dic_key_img = str(rank)+"_img" temp[dic_key_cap] = all_docs_captions[top_id] temp[dic_key_img] = all_docs_images[top_id] if self.args.ctx_prediction: context = self.tokenizer.decode(all_topk_context_indices[top_id]) dic_key_ctx = str(rank)+"_ctx" temp[dic_key_ctx] = context if type == 0: temp_dict["img"] = temp elif type == 1: temp_dict["txt"] = temp else: temp_dict["multi"] = temp output_data["result"].append(temp_dict) with open(self.args.test_output, "w") as outfile: json.dump(output_data, outfile, indent = 4) self.log("img_r1", img_output_score['r1'], on_epoch=True, prog_bar=True, logger=True, sync_dist=True) self.log("img_r5", img_output_score['r5'], on_epoch=True, prog_bar=True, logger=True, sync_dist=True) self.log("img_r10", img_output_score['r10'], on_epoch=True, prog_bar=True, logger=True, sync_dist=True) self.log("img_r_mean", img_output_score['r_mean'], on_epoch=True, prog_bar=True, logger=True, sync_dist=True) self.log("img_mrr10", img_output_score['mrr10'], on_epoch=True, prog_bar=True, logger=True, sync_dist=True) self.log("txt_r1", txt_output_score['r1'], on_epoch=True, prog_bar=True, logger=True, sync_dist=True) self.log("txt_r5", txt_output_score['r5'], on_epoch=True, prog_bar=True, logger=True, sync_dist=True) self.log("txt_r10", txt_output_score['r10'], on_epoch=True, prog_bar=True, logger=True, sync_dist=True) self.log("txt_r_mean", txt_output_score['r_mean'], on_epoch=True, prog_bar=True, logger=True, sync_dist=True) self.log("txt_mrr10", txt_output_score['mrr10'], on_epoch=True, prog_bar=True, logger=True, sync_dist=True) self.log("multi_r1", multi_output_score['r1'], on_epoch=True, prog_bar=True, logger=True, sync_dist=True) self.log("multi_r5", multi_output_score['r5'], on_epoch=True, prog_bar=True, logger=True, sync_dist=True) self.log("multi_r10", multi_output_score['r10'], on_epoch=True, prog_bar=True, logger=True, sync_dist=True) self.log("multi_r_mean", multi_output_score['r_mean'], on_epoch=True, prog_bar=True, logger=True, sync_dist=True) self.log("multi_mrr10", multi_output_score['mrr10'], on_epoch=True, prog_bar=True, logger=True, sync_dist=True) return { "img_r1":img_output_score['r1'], "img_r5":img_output_score['r5'], "img_r10":img_output_score['r10'], "img_r_mean":img_output_score['r_mean'], "img_mrr10":img_output_score['mrr10'], "txt_r1":txt_output_score['r1'], "txt_r5":txt_output_score['r5'], "txt_r10":txt_output_score['r10'], "txt_r_mean":txt_output_score['r_mean'], "txt_mrr10":txt_output_score['mrr10'], "multi_r1":multi_output_score['r1'], "multi_r5":multi_output_score['r5'], "multi_r10":multi_output_score['r10'], "multi_r_mean":multi_output_score['r_mean'], "multi_mrr10":multi_output_score['mrr10'] } def configure_optimizers(self): # optimizer opt_args = dict(lr=float(self.arg_opt.lr), weight_decay=float(self.arg_opt.weight_decay)) if hasattr(self.arg_opt, 'opt_eps') and self.arg_opt.opt_eps is not None: opt_args['eps'] = self.args.opt_eps if hasattr(self.arg_opt, 'opt_betas') and self.arg_opt.opt_betas is not None: opt_args['betas'] = self.args.opt_betas if hasattr(self.arg_opt, 'opt_args') and self.arg_opt.opt_args is not None: opt_args.update(self.args.opt_args) if self.arg_opt.opt == "adamW": optimizer = torch.optim.AdamW(self.model.parameters(),**opt_args) # scheduler lr_scheduler = None if self.arg_sche.sched == 'cosine': lr_scheduler = CosineLRScheduler( optimizer, t_initial=self.trainer.max_epochs, t_mul=getattr(self.arg_sche, 'lr_cycle_mul', 1.), lr_min=float(self.arg_sche.min_lr), decay_rate=self.arg_sche.decay_rate, warmup_lr_init=float(self.arg_sche.warmup_lr), warmup_t=self.arg_sche.warmup_epochs, cycle_limit=getattr(self.arg_sche, 'lr_cycle_limit', 1), ) # num_epochs = lr_scheduler.get_cycle_length() + self.arg_sche.cooldown_epochs # self.trainer.max_epochs = num_epochs return {'optimizer': optimizer, 'lr_scheduler': lr_scheduler} # https://github.com/pytorch/pytorch/issues/36748 def unique(self,x, dim=-1): unique, inverse = torch.unique(x, return_inverse=True, dim=dim) perm = torch.arange(inverse.size(dim), dtype=inverse.dtype, device=inverse.device) inverse, perm = inverse.flip([dim]), perm.flip([dim]) return unique, inverse.new_empty(unique.size(dim)).scatter_(dim, inverse, perm)
43,280
57.095302
126
py
Mr.Right
Mr.Right-main/metric.py
import numpy as np import torch import pdb from torchmetrics.functional import retrieval_recall,retrieval_reciprocal_rank @torch.no_grad() def score(scores_t2m, query_doc_id): """ scores_t2m: (q_size, d_size) query_doc_id: (q_size) """ ids = query_doc_id.unsqueeze(1) top1_i = torch.topk(scores_t2m, k=1, dim=1).indices top5_i = torch.topk(scores_t2m, k=5, dim=1).indices top10_v, top10_i = torch.topk(scores_t2m, k=10, dim=1) r1 = torch.mean(torch.sum(top1_i == ids, dim=1).float()).item() * 100 r5 = torch.mean(torch.sum(top5_i == ids, dim=1).float()).item() * 100 r10 = torch.mean(torch.sum(top10_i == ids, dim=1).float()).item() * 100 rmean = np.mean([r1, r5, r10]) top10_m = (top10_i==ids) mrr10 = np.mean([retrieval_reciprocal_rank(v, m).item() for v, m in zip(top10_v, top10_m)]) * 100 r1, r5, r10, mrr10, rmean eval_result = {'r1': r1, 'r5': r5, 'r10': r10, 'mrr10': mrr10, 'r_mean': rmean, } return eval_result
1,097
30.371429
103
py
Mr.Right
Mr.Right-main/scheduler/cosine_lr.py
""" Cosine Scheduler Cosine LR schedule with warmup, cycle/restarts, noise. Hacked together by / Copyright 2020 Ross Wightman """ import logging import math import numpy as np import torch from .scheduler import Scheduler from pdb import set_trace as breakpoint _logger = logging.getLogger(__name__) class CosineLRScheduler(Scheduler): """ Cosine decay with restarts. This is described in the paper https://arxiv.org/abs/1608.03983. Inspiration from https://github.com/allenai/allennlp/blob/master/allennlp/training/learning_rate_schedulers/cosine.py """ def __init__(self, optimizer: torch.optim.Optimizer, t_initial: int, t_mul: float = 1., lr_min: float = 0., decay_rate: float = 1., warmup_t=0, warmup_lr_init=0, warmup_prefix=True, cycle_limit=0, t_in_epochs=True, noise_range_t=None, noise_pct=0.67, noise_std=1.0, noise_seed=42, initialize=True) -> None: super().__init__( optimizer, param_group_field="lr", noise_range_t=noise_range_t, noise_pct=noise_pct, noise_std=noise_std, noise_seed=noise_seed, initialize=initialize) assert t_initial > 0 assert lr_min >= 0 if t_initial == 1 and t_mul == 1 and decay_rate == 1: _logger.warning("Cosine annealing scheduler will have no effect on the learning " "rate since t_initial = t_mul = eta_mul = 1.") self.t_initial = t_initial self.t_mul = t_mul self.lr_min = lr_min self.decay_rate = decay_rate self.cycle_limit = cycle_limit self.warmup_t = warmup_t self.warmup_lr_init = warmup_lr_init self.warmup_prefix = warmup_prefix self.t_in_epochs = t_in_epochs if self.warmup_t: self.warmup_steps = [(v - warmup_lr_init) / self.warmup_t for v in self.base_values] super().update_groups(self.warmup_lr_init) else: self.warmup_steps = [1 for _ in self.base_values] def _get_lr(self, t): if t < self.warmup_t: lrs = [self.warmup_lr_init + t * s for s in self.warmup_steps] else: if self.warmup_prefix: t = t - self.warmup_t if self.t_mul != 1: i = math.floor(math.log(1 - t / self.t_initial * (1 - self.t_mul), self.t_mul)) t_i = self.t_mul ** i * self.t_initial t_curr = t - (1 - self.t_mul ** i) / (1 - self.t_mul) * self.t_initial else: i = t // self.t_initial t_i = self.t_initial t_curr = t - (self.t_initial * i) gamma = self.decay_rate ** i lr_min = self.lr_min * gamma lr_max_values = [v * gamma for v in self.base_values] if self.cycle_limit == 0 or (self.cycle_limit > 0 and i < self.cycle_limit): lrs = [ lr_min + 0.5 * (lr_max - lr_min) * (1 + math.cos(math.pi * t_curr / t_i)) for lr_max in lr_max_values ] else: lrs = [self.lr_min for _ in self.base_values] return lrs def get_epoch_values(self, epoch: int): if self.t_in_epochs: return self._get_lr(epoch) else: return None def get_update_values(self, num_updates: int): if not self.t_in_epochs: return self._get_lr(num_updates) else: return None def get_cycle_length(self, cycles=0): if not cycles: cycles = self.cycle_limit cycles = max(1, cycles) if self.t_mul == 1.0: return self.t_initial * cycles else: return int(math.floor(-self.t_initial * (self.t_mul ** cycles - 1) / (1 - self.t_mul)))
4,027
33.135593
121
py
Mr.Right
Mr.Right-main/scheduler/scheduler.py
from typing import Dict, Any import torch class Scheduler: """ Parameter Scheduler Base Class A scheduler base class that can be used to schedule any optimizer parameter groups. Unlike the builtin PyTorch schedulers, this is intended to be consistently called * At the END of each epoch, before incrementing the epoch count, to calculate next epoch's value * At the END of each optimizer update, after incrementing the update count, to calculate next update's value The schedulers built on this should try to remain as stateless as possible (for simplicity). This family of schedulers is attempting to avoid the confusion of the meaning of 'last_epoch' and -1 values for special behaviour. All epoch and update counts must be tracked in the training code and explicitly passed in to the schedulers on the corresponding step or step_update call. Based on ideas from: * https://github.com/pytorch/fairseq/tree/master/fairseq/optim/lr_scheduler * https://github.com/allenai/allennlp/tree/master/allennlp/training/learning_rate_schedulers """ def __init__(self, optimizer: torch.optim.Optimizer, param_group_field: str, noise_range_t=None, noise_type='normal', noise_pct=0.67, noise_std=1.0, noise_seed=None, initialize: bool = True) -> None: self.optimizer = optimizer self.param_group_field = param_group_field self._initial_param_group_field = f"initial_{param_group_field}" if initialize: for i, group in enumerate(self.optimizer.param_groups): if param_group_field not in group: raise KeyError(f"{param_group_field} missing from param_groups[{i}]") group.setdefault(self._initial_param_group_field, group[param_group_field]) else: for i, group in enumerate(self.optimizer.param_groups): if self._initial_param_group_field not in group: raise KeyError(f"{self._initial_param_group_field} missing from param_groups[{i}]") self.base_values = [group[self._initial_param_group_field] for group in self.optimizer.param_groups] self.metric = None # any point to having this for all? self.noise_range_t = noise_range_t self.noise_pct = noise_pct self.noise_type = noise_type self.noise_std = noise_std self.noise_seed = noise_seed if noise_seed is not None else 42 self.update_groups(self.base_values) def state_dict(self) -> Dict[str, Any]: return {key: value for key, value in self.__dict__.items() if key != 'optimizer'} def load_state_dict(self, state_dict: Dict[str, Any]) -> None: self.__dict__.update(state_dict) def get_epoch_values(self, epoch: int): return None def get_update_values(self, num_updates: int): return None def step(self, epoch: int, metric: float = None) -> None: self.metric = metric values = self.get_epoch_values(epoch) if values is not None: values = self._add_noise(values, epoch) self.update_groups(values) def step_update(self, num_updates: int, metric: float = None): self.metric = metric values = self.get_update_values(num_updates) if values is not None: values = self._add_noise(values, num_updates) self.update_groups(values) def update_groups(self, values): if not isinstance(values, (list, tuple)): values = [values] * len(self.optimizer.param_groups) for param_group, value in zip(self.optimizer.param_groups, values): param_group[self.param_group_field] = value def _add_noise(self, lrs, t): if self.noise_range_t is not None: if isinstance(self.noise_range_t, (list, tuple)): apply_noise = self.noise_range_t[0] <= t < self.noise_range_t[1] else: apply_noise = t >= self.noise_range_t if apply_noise: g = torch.Generator() g.manual_seed(self.noise_seed + t) if self.noise_type == 'normal': while True: # resample if noise out of percent limit, brute force but shouldn't spin much noise = torch.randn(1, generator=g).item() if abs(noise) < self.noise_pct: break else: noise = 2 * (torch.rand(1, generator=g).item() - 0.5) * self.noise_pct lrs = [v + v * noise for v in lrs] return lrs
4,750
43.820755
112
py
Mr.Right
Mr.Right-main/scheduler/__init__.py
from .cosine_lr import CosineLRScheduler
41
20
40
py
Mr.Right
Mr.Right-main/models/matching.py
import torch import torch.nn as nn import torch.nn.functional as F class MatchingModel(nn.Module): def __init__(self, args, config, text_width, n_layers): super().__init__() self.config = config from models.ALBEF.models.xbert import BertModel self.config.num_hidden_layers = 4 self.config.fusion_layer = 0 self.itm_transformer = BertModel(self.config, add_pooling_layer=False) self.itm_head = nn.Linear(text_width, 2) def forward(self, query_embeds, query_attns, multi_embeds, multi_attns): output = self.itm_transformer( encoder_embeds = query_embeds, attention_mask = query_attns, encoder_hidden_states = multi_embeds, encoder_attention_mask = multi_attns, return_dict = True, mode = 'fusion', ) embeddings = output.last_hidden_state[:, 0, :] logits = self.itm_head(embeddings) return logits if __name__ == "__main__": from models.ALBEF.models.xbert import BertConfig import yaml class AttrDict(dict): def __init__(self, *args, **kwargs): super(AttrDict, self).__init__(*args, **kwargs) self.__dict__ = self with open('configs/model_retrieval.yaml') as f: config = yaml.safe_load(f) config = AttrDict(config) bert_config = BertConfig.from_json_file(config['bert_config']) m = MatchingModel(config=bert_config, text_width=bert_config.hidden_size, n_layers=1) e1 = torch.rand(5, 10, 768) e2 = torch.rand(5, 7, 768) a1 = torch.ones(e1.size()[:-1],dtype=torch.long) a2 = torch.ones(e2.size()[:-1],dtype=torch.long) o = m(e1, a1, e2, a2) print(o.shape)
1,817
29.3
89
py
Mr.Right
Mr.Right-main/models/model.py
import pdb import torch import torch.nn.functional as F from torch import nn from models.ALBEF.models.model_retrieval import ALBEF from models.ALBEF.models.vit import interpolate_pos_embed from models.ALBEF.models.xbert import BertOnlyMLMHead,BertConfig from models.ViLT.vilt.modules import ViLTransformerSS from models.METER.meter.modules import METERTransformerSS from models.matching import MatchingModel class TextToMultiModel(nn.Module): def __init__(self, args = None, config = None, tokenizer = None, ): super().__init__() self.args = args self.tokenizer = tokenizer embed_dim = config.embed_dim # Choose pretrain if args.pretrain == "ALBEF": self.model = ALBEF(config.text_encoder,tokenizer,config) text_width = self.model.text_encoder.config.hidden_size bert_config = self.model.bert_config if config.checkpoint!="": checkpoint = torch.load(config.checkpoint, map_location='cpu') state_dict = checkpoint['model'] # reshape positional embedding to accomodate for image resolution change pos_embed_reshaped = interpolate_pos_embed(state_dict['visual_encoder.pos_embed'],self.model.visual_encoder) state_dict['visual_encoder.pos_embed'] = pos_embed_reshaped for key in list(state_dict.keys()): if 'bert' in key: encoder_key = key.replace('bert.','') state_dict[encoder_key] = state_dict[key] del state_dict[key] msg = self.model.load_state_dict(state_dict,strict=False) print('load checkpoint from %s'%config.checkpoint) print(msg) elif args.pretrain == "ViLT": self.model = ViLTransformerSS(config) text_width = config.hidden_size bert_config = BertConfig.from_json_file(config.bert_config) elif args.pretrain == "METER": self.model = METERTransformerSS(config) text_width = config.hidden_size bert_config = BertConfig.from_json_file(config.bert_config) if args.pretrain == "METER" and (self.args.embeds_feats == "cls" or self.args.embeds_feats == "iavg_tcls"): self.multi_proj = nn.Linear(config.hidden_size*2, embed_dim) else: self.multi_proj = nn.Linear(config.hidden_size, embed_dim) self.query_proj = nn.Linear(text_width, embed_dim) if 'vocab_size' in config: bert_config.vocab_size = config.vocab_size self.context = BertOnlyMLMHead(bert_config) self.temp = nn.Parameter(torch.ones([]) * config['temp']) self.queue_size = config['queue_size'] # Matching self.matching_classifier = MatchingModel(args = args, config=bert_config, text_width=text_width, n_layers=1) # create the queue self.register_buffer("multi_queue", torch.randn(embed_dim, self.queue_size)) # self.register_buffer("query_queue", torch.randn(embed_dim, self.queue_size)) self.register_buffer("idx_queue", torch.full((1,self.queue_size),-100)) self.register_buffer("queue_num", torch.zeros(1, dtype=torch.long)) self.multi_queue = nn.functional.normalize(self.multi_queue, dim=0) # self.query_queue = nn.functional.normalize(self.query_queue, dim=0) def forward_feats(self, doc_image, doc_text, query): result = {} if self.args.pretrain == "ALBEF": output = self.model(doc_image,doc_text,query) elif self.args.pretrain == "ViLT": output = self.model(batch={ "image":doc_image, "text_ids":doc_text['input_ids'], "text_masks":doc_text['attention_mask'], "query_ids":query['input_ids'], "query_masks":query['attention_mask'], }) elif self.args.pretrain == "METER": output = self.model(batch={ "image":doc_image, "text_ids":doc_text['input_ids'], "text_masks":doc_text['attention_mask'], "query_ids":query['input_ids'], "query_masks":query['attention_mask'], }) result["query_embeds"] = output["query_embeds"] result["query_atts"] = output["query_atts"] result["multi_embeds"] = output["multi_embeds"] result["multi_atts"] = output["multi_atts"] # How to get feature if self.args.embeds_feats == "avg": avg_query_embeds = (result["query_embeds"] * result["query_atts"].unsqueeze(-1)).sum(dim=1) / result["query_atts"].sum(dim=1).unsqueeze(-1) result["query_feat"] = F.normalize(self.query_proj(avg_query_embeds),dim=-1) avg_multi_embeds = (result["multi_embeds"] * result["multi_atts"].unsqueeze(-1)).sum(dim=1) / result["multi_atts"].sum(dim=1).unsqueeze(-1) result["multi_feat"] = F.normalize(self.multi_proj(avg_multi_embeds),dim=-1) elif self.args.embeds_feats == "cls": result["query_feat"] = output["query_cls"].float() if self.args.pretrain == "METER": # METER has two cls token multi_embeds = torch.cat([output["text_cls"], output["img_cls"]], dim=-1).float() result["multi_feat"] = F.normalize(self.multi_proj(multi_embeds),dim=-1) else: result["multi_feat"] = output["multi_cls"].float() elif self.args.embeds_feats == "iavg_tcls": result["query_feat"] = output["query_cls"].float() text_cls = output["text_cls"] avg_img_embeds = (output["image_feats"] * output["image_masks"].unsqueeze(-1)).sum(dim=1) / output["image_masks"].sum(dim=1).unsqueeze(-1) concat_embeds = torch.cat([text_cls, avg_img_embeds], dim=-1).float() avg_multi_feat = F.normalize(self.multi_proj(concat_embeds),dim=-1) result["multi_feat"] = avg_multi_feat.float() return result def forward(self, query, doc_text, doc_image, doc_id, context_labels=None, matching=None, matchingv2=None): query['input_ids'] = query['input_ids'].view(query['input_ids'].shape[0],-1) query['attention_mask'] = query['attention_mask'].view(query['input_ids'].shape[0],-1) if "token_type_ids" in query: query['token_type_ids'] = query['token_type_ids'].view(query['input_ids'].shape[0],-1) doc_text['input_ids'] = doc_text['input_ids'].view(doc_text['input_ids'].shape[0],-1) doc_text['attention_mask'] = doc_text['attention_mask'].view(doc_text['input_ids'].shape[0],-1) if "token_type_ids" in doc_text: doc_text['token_type_ids'] = doc_text['token_type_ids'].view(doc_text['input_ids'].shape[0],-1) result = self.forward_feats(doc_image, doc_text, query) multi_feat = result['multi_feat'] # B, 1, H query_feat = result['query_feat'] # B, 1, H multi_embeds = result['multi_embeds'] # B, L, H multi_atts = result['multi_atts'] # B, L, H query_embeds = result['query_embeds'] query_atts = result['query_atts'] # [TODO] # why only query with doc similarity # why not add doc with query similarity? with torch.no_grad(): multi_feat_all = torch.cat([multi_feat.t(),self.multi_queue.clone().detach()],dim=1) # query_feat_all = torch.cat([query_feat.t(),self.query_queue.clone().detach()],dim=1) sim_q2m = query_feat @ multi_feat_all / self.temp # sim_m2q = multi_feat @ query_feat_all / self.temp idx = doc_id.view(-1,1) # batch_size, id [0,1,2,0] idx_all = torch.cat([idx.t(), self.idx_queue.clone().detach()],dim=1) # 1, queue_size [[0,1,2,0,12,13,1,0,2]] pos_idx = torch.eq(idx, idx_all).float() # 1,queue_size [[1,0,0,1,0,0,0,1,0],...] sim_targets = pos_idx / pos_idx.sum(1,keepdim=True) # normalzie [[0.33,0,0,0.33,0,0,0,0.33,0],...] loss_q2m = -torch.sum(F.log_softmax(sim_q2m, dim=1)*sim_targets,dim=1).mean() # loss_m2q = -torch.sum(F.log_softmax(sim_m2q, dim=1)*sim_targets,dim=1).mean() # loss_ita = (loss_q2m + loss_m2q)/2 loss_ita = loss_q2m self._dequeue_and_enqueue(multi_feat, None, idx) # text_feat_m # self._dequeue_and_enqueue(multi_feat, query_feat, idx) # text_feat_m # ===== Matching Loss ===== # Matching Classification # Pos (multi_embeds, query_embeds) # Neg (multi_embeds_neg, query_embeds) # Neg (multi_embeds_image+multi_embeds_neg_text ,query_embeds) # Neg (multi_embeds_text+multi_embeds_neg_image ,query_embeds) # Neg (multi_embeds, query_embeds_neg) loss_itm = 0.0 if matching: with torch.no_grad(): bs = doc_image.size(0) mask = torch.eq(idx, idx.T) weights_q2m = F.softmax(sim_q2m[:, :bs]+1e-4, dim=1) weights_q2m.masked_fill_(mask, 1e-10) # weights_m2q = F.softmax(sim_m2q[:, :bs]+1e-4, dim=1) # weights_m2q.masked_fill_(mask, 1e-10) # [New] if matchingv2: neg_doc_text_inps = [] neg_doc_text_atts = [] neg_doc_image = [] multi_embeds_neg = [] multi_atts_neg = [] for b in range(bs): neg_idx = torch.multinomial(weights_q2m[b], 1).item() multi_embeds_neg.append(multi_embeds[neg_idx]) multi_atts_neg.append(multi_atts[neg_idx]) #[New] if matchingv2: neg_doc_text_inps.append(doc_text['input_ids'][neg_idx]) neg_doc_text_atts.append(doc_text['attention_mask'][neg_idx]) neg_doc_image.append(doc_image[neg_idx]) multi_embeds_neg = torch.stack(multi_embeds_neg, dim=0) multi_atts_neg = torch.stack(multi_atts_neg, dim=0) # [New] if matchingv2: neg_doc_text_inps = torch.stack(neg_doc_text_inps, dim=0) neg_doc_text_atts = torch.stack(neg_doc_text_atts, dim=0) neg_doc_text = {'input_ids': neg_doc_text_inps, 'attention_mask': neg_doc_text_atts} neg_doc_image = torch.stack(neg_doc_image, dim=0) result_neg_image = self.forward_feats(neg_doc_image, doc_text ,query) multi_embeds_image_neg = result_neg_image['multi_embeds'] multi_atts_image_neg = result_neg_image['multi_atts'] result_neg_text = self.forward_feats(doc_image, neg_doc_text ,query) multi_embeds_text_neg = result_neg_text['multi_embeds'] multi_atts_text_neg = result_neg_text['multi_atts'] # Quadra if matchingv2: query_embeds_matching = torch.cat([query_embeds, query_embeds, query_embeds, query_embeds], dim=0) query_attn_matching = torch.cat([query_atts, query_atts, query_atts, query_atts], dim=0) multi_embeds_matching = torch.cat([multi_embeds, multi_embeds_neg, multi_embeds_image_neg, multi_embeds_text_neg], dim=0) multi_attn_matching = torch.cat([multi_atts, multi_atts_neg, multi_atts_image_neg, multi_atts_text_neg], dim=0) else: # Binary query_embeds_matching = torch.cat([query_embeds, query_embeds], dim=0) query_attn_matching = torch.cat([query_atts, query_atts], dim=0) multi_embeds_matching = torch.cat([multi_embeds, multi_embeds_neg], dim=0) multi_attn_matching = torch.cat([multi_atts, multi_atts_neg], dim=0) # Triple # query_embeds_matching = torch.cat([query_embeds, query_embeds, query_embeds_neg], dim=0) # query_attn_matching = torch.cat([query_atts, query_atts, query_atts_neg], dim=0) # multi_embeds_matching = torch.cat([multi_embeds, multi_embeds_neg, multi_embeds], dim=0) # multi_attn_matching = torch.cat([multi_atts, multi_atts_neg, multi_atts], dim=0) itm_logits = self.matching_classifier( query_embeds=query_embeds_matching, query_attns=query_attn_matching, multi_embeds=multi_embeds_matching, multi_attns=multi_attn_matching ) itm_labels = torch.cat([torch.ones(bs,dtype=torch.long), torch.zeros(bs * 3 if matchingv2 else bs * 1,dtype=torch.long)], dim=0).to(doc_image.device) # [TODO] Binary_CE loss_itm = F.cross_entropy(itm_logits, itm_labels) loss_ctx_pred = 0.0 # ===== Context Prediction Loss ===== if context_labels is not None: ctx_targets = context_labels / context_labels.sum(1,keepdim=True) prediction_scores = self.context(multi_embeds) mean_prediction_scores = torch.mean(prediction_scores,1) loss_ctx_pred = -torch.sum(F.log_softmax(mean_prediction_scores, dim=1)*ctx_targets,dim=1).mean() return loss_ita, loss_ctx_pred, loss_itm def output_itm_logits(self, query, doc_text, doc_image): query['input_ids'] = query['input_ids'].view(query['input_ids'].shape[0],-1) query['attention_mask'] = query['attention_mask'].view(query['input_ids'].shape[0],-1) if "token_type_ids" in query: query['token_type_ids'] = query['token_type_ids'].view(query['input_ids'].shape[0],-1) doc_text['input_ids'] = doc_text['input_ids'].view(doc_text['input_ids'].shape[0],-1) doc_text['attention_mask'] = doc_text['attention_mask'].view(doc_text['input_ids'].shape[0],-1) if "token_type_ids" in doc_text: doc_text['token_type_ids'] = doc_text['token_type_ids'].view(doc_text['input_ids'].shape[0],-1) result = self.forward_feats(doc_image, doc_text, query) multi_feat = result['multi_feat'] # B, 1, H query_feat = result['query_feat'] # B, 1, H multi_embeds = result['multi_embeds'] # B, L, H multi_atts = result['multi_atts'] # B, L, H query_embeds = result['query_embeds'] query_atts = result['query_atts'] itm_logits = self.matching_classifier( query_embeds=query_embeds, query_attns=query_atts, multi_embeds=multi_embeds, multi_attns=multi_atts ) return itm_logits @torch.no_grad() def output_query_feats(self,query): if self.args.pretrain == "ALBEF": query_output = self.model.text_encoder(query['input_ids'], attention_mask = query["attention_mask"], mode='text') query_embeds = query_output.last_hidden_state query_masks = query["attention_mask"] query_cls = F.normalize(self.model.text_proj(query_embeds[:,0,:]),dim=-1) elif self.args.pretrain == "ViLT": query_embeds = self.model.text_embeddings(query['input_ids']) query_masks = query['attention_mask'] for i, blk in enumerate(self.model.transformer.blocks): query_embeds, _ = blk(query_embeds, mask=query_masks) query_embeds = self.model.transformer.norm(query_embeds) query_cls = F.normalize(self.query_proj(query_embeds[:,0,:]),dim=-1) elif self.args.pretrain == "METER": query_embeds = self.model.text_transformer.embeddings(input_ids=query['input_ids']) query_masks = query['attention_mask'] device = query_embeds.device input_shape = query_masks.size() extend_query_masks = self.model.text_transformer.get_extended_attention_mask(query_masks, input_shape, device) for layer in self.model.text_transformer.encoder.layer: query_embeds = layer(query_embeds, extend_query_masks)[0] query_embeds = self.model.cross_modal_text_transform(query_embeds) query_cls = self.model.cross_modal_text_pooler(query_embeds) if self.args.embeds_feats == "avg": avg_query_embeds = (query_embeds * query_masks.unsqueeze(-1)).sum(dim=1) / query_masks.sum(dim=1).unsqueeze(-1) query_feat = F.normalize(self.query_proj(avg_query_embeds),dim=-1) elif self.args.embeds_feats == "cls": query_feat = query_cls.float() elif self.args.embeds_feats == "iavg_tcls": query_feat = query_cls.float() return query_embeds,query_feat @torch.no_grad() def output_doc_feats(self,doc_text,doc_image): if self.args.pretrain == "ALBEF": image_embeds = self.model.visual_encoder(doc_image) image_atts = torch.ones(image_embeds.size()[:-1],dtype=torch.long).to(doc_image.device) doc_text_output = self.model.text_encoder(doc_text['input_ids'], attention_mask = doc_text["attention_mask"], mode='text') doc_text_embeds = doc_text_output.last_hidden_state output = self.model.text_encoder(encoder_embeds = doc_text_embeds, attention_mask = doc_text["attention_mask"], encoder_hidden_states = image_embeds, encoder_attention_mask = image_atts, return_dict = True, mode = 'fusion', ) multi_embeds = output.last_hidden_state multi_atts = doc_text["attention_mask"] multi_cls = F.normalize(self.model.text_proj(multi_embeds[:,0,:]),dim=-1) elif self.args.pretrain == "ViLT": output = self.model.output_multi(batch={ "image":doc_image, "text_ids":doc_text['input_ids'], "text_masks":doc_text['attention_mask'], }) multi_embeds = output["multi_embeds"] multi_atts = output["multi_atts"] multi_cls = output["multi_cls"] elif self.args.pretrain == "METER": output = self.model.output_multi(batch={ "image":doc_image, "text_ids":doc_text['input_ids'], "text_masks":doc_text['attention_mask'], }) multi_embeds = output["multi_embeds"] multi_atts = output["multi_atts"] text_cls = output["text_cls"] img_cls = output["cls_feats_image"] if self.args.embeds_feats == "avg": avg_multi_embeds = (multi_embeds * multi_atts.unsqueeze(-1)).sum(dim=1) / multi_atts.sum(dim=1).unsqueeze(-1) avg_multi_feat = F.normalize(self.multi_proj(avg_multi_embeds),dim=-1) elif self.args.embeds_feats == "cls": if self.args.pretrain == "METER": # METER has two cls token multi_embeds = torch.cat([text_cls, img_cls], dim=-1).float() avg_multi_feat = F.normalize(self.multi_proj(multi_embeds),dim=-1) else: avg_multi_feat = multi_cls.float() elif self.args.embeds_feats == "iavg_tcls": text_cls = output["text_cls"] img_embeds = output["image_feats"] avg_img_embeds = (output["image_feats"] * output["image_masks"].unsqueeze(-1)).sum(dim=1) / output["image_masks"].sum(dim=1).unsqueeze(-1) concat_embeds = torch.cat([text_cls, avg_img_embeds], dim=-1).float() avg_multi_feat = F.normalize(self.multi_proj(concat_embeds),dim=-1) return multi_embeds,avg_multi_feat, multi_atts @torch.no_grad() def _dequeue_and_enqueue(self, multi_feat, query_feat, idx): # gather keys before updating queue idxs = concat_all_gather(idx) if multi_feat is not None: multi_feats = concat_all_gather(multi_feat) batch_size = multi_feats.shape[0] ptr = int(self.queue_num) assert self.queue_size % batch_size == 0 # for simplicity self.multi_queue[:, ptr:ptr + batch_size] = multi_feats.T if query_feat is not None: query_feats = concat_all_gather(query_feat) batch_size = query_feats.shape[0] ptr = int(self.queue_num) assert self.queue_size % batch_size == 0 # for simplicity self.query_queue[:, ptr:ptr + batch_size] = query_feats.T # replace the keys at ptr (dequeue and enqueue) self.idx_queue[:, ptr:ptr + batch_size] = idxs.T ptr = (ptr + batch_size) % self.queue_size # move pointer self.queue_num[0] = ptr @torch.no_grad() def concat_all_gather(tensor): """ Performs all_gather operation on the provided tensors. *** Warning ***: torch.distributed.all_gather has no gradient. """ group = torch.distributed.group.WORLD world_size = torch.distributed.get_world_size(group) tensors_gather = [torch.ones_like(tensor) for _ in range(world_size)] torch.distributed.all_gather(tensors_gather, tensor, group, async_op=False) output = torch.cat(tensors_gather, dim=0) return output
21,799
50.294118
151
py
Mr.Right
Mr.Right-main/models/METER/azure_distributed_run.py
import os import copy import pytorch_lightning as pl import os os.environ["NCCL_DEBUG"] = "INFO" from meter.config import ex from meter.modules import METERTransformerSS from meter.datamodules.multitask_datamodule import MTDataModule import resource rlimit = resource.getrlimit(resource.RLIMIT_NOFILE) resource.setrlimit(resource.RLIMIT_NOFILE, (20480, rlimit[1])) from pytorch_lightning.plugins.environments import ClusterEnvironment from pytorch_lightning.plugins.training_type import DDPPlugin import torch.distributed as dist class MyCluster(ClusterEnvironment): def creates_children(self) -> bool: # return True if the cluster is managed (you don't launch processes yourself) return True def master_address(self): return os.environ['MASTER_ADDR'] def master_port(self) -> int: return int(os.environ["MASTER_PORT"]) def world_size(self): return int(os.environ['OMPI_COMM_WORLD_SIZE']) def global_rank(self) -> int: return int(os.environ['OMPI_COMM_WORLD_RANK']) def local_rank(self) -> int: return int(os.environ['OMPI_COMM_WORLD_LOCAL_RANK']) def node_rank(self) -> int: return int(os.environ["OMPI_COMM_WORLD_NODE_RANK"]) def set_global_rank(self, rank: int) -> None: pass def set_world_size(self, size: int) -> None: pass class MyDDPPlugin(DDPPlugin): def init_ddp_connection(self, global_rank = None, world_size = None) -> None: master_uri = "tcp://%s:%s" % (os.environ['MASTER_ADDR'], os.environ['MASTER_PORT']) dist.init_process_group( backend=self.torch_distributed_backend, init_method=master_uri, world_size=int(os.environ['OMPI_COMM_WORLD_SIZE']), rank=int(os.environ['OMPI_COMM_WORLD_RANK']), ) @ex.automain def main(_config): os.environ["NCCL_DEBUG"] = "INFO" world_size = int(os.environ['OMPI_COMM_WORLD_SIZE']) local_size = int(os.environ['OMPI_COMM_WORLD_LOCAL_SIZE']) global_rank = int(os.environ['OMPI_COMM_WORLD_RANK']) local_rank = int(os.environ['OMPI_COMM_WORLD_LOCAL_RANK']) master_addr = os.environ['MASTER_ADDR'] master_port = os.environ['MASTER_PORT'] # set environment variables for 'env://' os.environ['WORLD_SIZE'] = str(world_size) os.environ['NODE_RANK'] = str(os.environ["OMPI_COMM_WORLD_NODE_RANK"]) _config = copy.deepcopy(_config) pl.seed_everything(_config["seed"]) dm = MTDataModule(_config, dist=True) model = METERTransformerSS(_config) exp_name = f'{_config["exp_name"]}' os.makedirs(_config["log_dir"], exist_ok=True) checkpoint_callback = pl.callbacks.ModelCheckpoint( save_top_k=1, verbose=True, monitor="val/the_metric", mode="max", save_last=True, ) logger = pl.loggers.TensorBoardLogger( _config["log_dir"], name=f'{exp_name}_seed{_config["seed"]}_from_{_config["load_path"].split("/")[-1][:-5]}', ) lr_callback = pl.callbacks.LearningRateMonitor(logging_interval="step") callbacks = [checkpoint_callback, lr_callback] num_gpus = ( _config["num_gpus"] if isinstance(_config["num_gpus"], int) else len(_config["num_gpus"]) ) grad_steps = max(_config["batch_size"] // ( _config["per_gpu_batchsize"] * num_gpus * _config["num_nodes"] ), 1) max_steps = _config["max_steps"] if _config["max_steps"] is not None else None trainer = pl.Trainer( plugins=[MyCluster(), MyDDPPlugin()], gpus=_config["num_gpus"], num_nodes=_config["num_nodes"], precision=_config["precision"], accelerator="ddp", benchmark=True, deterministic=True, max_epochs=_config["max_epoch"] if max_steps is None else 1000, max_steps=max_steps, callbacks=callbacks, logger=logger, prepare_data_per_node=False, replace_sampler_ddp=False, accumulate_grad_batches=grad_steps, log_every_n_steps=10, flush_logs_every_n_steps=10, resume_from_checkpoint=_config["resume_from"], weights_summary="top", fast_dev_run=_config["fast_dev_run"], val_check_interval=_config["val_check_interval"], ) if not _config["test_only"]: trainer.fit(model, datamodule=dm) else: trainer.test(model, datamodule=dm)
4,388
31.272059
97
py
Mr.Right
Mr.Right-main/models/METER/setup.py
from setuptools import setup, find_packages setup( name="meter", packages=find_packages( exclude=[".dfc", ".vscode", "dataset", "notebooks", "result", "scripts"] ), version="0.1.0", license="MIT", description="METER: Multimodal End-to-end TransformER", author="Microsoft Corporation", author_email="zdou0830@gmail.com", url="https://github.com/zdou0830/METER", keywords=["vision and language pretraining"], install_requires=["torch", "pytorch_lightning"], )
511
29.117647
80
py
Mr.Right
Mr.Right-main/models/METER/run.py
import os import copy import pytorch_lightning as pl import os os.environ["NCCL_DEBUG"] = "INFO" from meter.config import ex from meter.modules import METERTransformerSS from meter.datamodules.multitask_datamodule import MTDataModule import resource rlimit = resource.getrlimit(resource.RLIMIT_NOFILE) resource.setrlimit(resource.RLIMIT_NOFILE, (20480, rlimit[1])) @ex.automain def main(_config): _config = copy.deepcopy(_config) pl.seed_everything(_config["seed"]) dm = MTDataModule(_config, dist=True) model = METERTransformerSS(_config) exp_name = f'{_config["exp_name"]}' os.makedirs(_config["log_dir"], exist_ok=True) checkpoint_callback = pl.callbacks.ModelCheckpoint( save_top_k=1, verbose=True, monitor="val/the_metric", mode="max", save_last=True, ) logger = pl.loggers.TensorBoardLogger( _config["log_dir"], name=f'{exp_name}_seed{_config["seed"]}_from_{_config["load_path"].split("/")[-1][:-5]}', ) lr_callback = pl.callbacks.LearningRateMonitor(logging_interval="step") callbacks = [checkpoint_callback, lr_callback] num_gpus = ( _config["num_gpus"] if isinstance(_config["num_gpus"], int) else len(_config["num_gpus"]) ) grad_steps = max(_config["batch_size"] // ( _config["per_gpu_batchsize"] * num_gpus * _config["num_nodes"] ), 1) max_steps = _config["max_steps"] if _config["max_steps"] is not None else None trainer = pl.Trainer( gpus=_config["num_gpus"], num_nodes=_config["num_nodes"], precision=_config["precision"], accelerator="ddp", benchmark=True, deterministic=True, max_epochs=_config["max_epoch"] if max_steps is None else 1000, max_steps=max_steps, callbacks=callbacks, logger=logger, prepare_data_per_node=False, replace_sampler_ddp=False, accumulate_grad_batches=grad_steps, log_every_n_steps=10, flush_logs_every_n_steps=10, resume_from_checkpoint=_config["resume_from"], weights_summary="top", fast_dev_run=_config["fast_dev_run"], val_check_interval=_config["val_check_interval"], ) if not _config["test_only"]: trainer.fit(model, datamodule=dm) else: trainer.test(model, datamodule=dm)
2,373
29.050633
97
py
Mr.Right
Mr.Right-main/models/METER/meter/config.py
from sacred import Experiment ex = Experiment("METER") def _loss_names(d): ret = { "itm": 0, "mlm": 0, "mpp": 0, "vqa": 0, "vcr": 0, "vcr_qar": 0, "nlvr2": 0, "irtr": 0, "contras": 0, "snli": 0, } ret.update(d) return ret @ex.config def config(): exp_name = "meter" seed = 0 datasets = ["coco", "vg", "sbu", "gcc"] loss_names = _loss_names({"itm": 1, "mlm": 1}) batch_size = 4096 # this is a desired batch size; pl trainer will accumulate gradients when per step batch is smaller. # Image setting train_transform_keys = ["clip"] val_transform_keys = ["clip"] image_size = 224 patch_size = 32 draw_false_image = 1 image_only = False resolution_before = 224 # Text Setting vqav2_label_size = 3129 max_text_len = 40 tokenizer = "bert-base-uncased" vocab_size = 30522 whole_word_masking = False # note that whole_word_masking does not work for RoBERTa mlm_prob = 0.15 draw_false_text = 0 # Transformer Setting num_top_layer = 6 input_image_embed_size = 768 input_text_embed_size = 768 vit = 'ViT-B/32' hidden_size = 768 num_heads = 12 num_layers = 6 mlp_ratio = 4 drop_rate = 0.1 # Optimizer Setting optim_type = "adamw" learning_rate = 1e-5 weight_decay = 0.01 decay_power = 1 max_epoch = 100 max_steps = 100000 warmup_steps = 10000 end_lr = 0 lr_mult_head = 5 # multiply lr for downstream heads lr_mult_cross_modal = 5 # multiply lr for the cross-modal module # Downstream Setting get_recall_metric = False # PL Trainer Setting resume_from = None fast_dev_run = False val_check_interval = 1.0 test_only = False # below params varies with the environment data_root = "" log_dir = "result" per_gpu_batchsize = 0 # you should define this manually with per_gpu_batch_size=# num_gpus = 8 num_nodes = 1 load_path = "" num_workers = 8 precision = 32 @ex.named_config def task_mlm_itm_clip_bert(): exp_name = "mlm_itm" datasets = ["coco", "vg", "sbu", "gcc"] loss_names = _loss_names({"itm": 1, "mlm": 1}) batch_size = 4096 max_epoch = 10 max_steps = 100000 warmup_steps = 0.1 whole_word_masking = True vocab_size = 30522 max_text_len = 50 image_size = 224 tokenizer = "bert-base-uncased" train_transform_keys = ["clip"] val_transform_keys = ["clip"] learning_rate = 1e-5 val_check_interval = 1.0 lr_mult_head = 5 lr_mult_cross_modal = 5 num_top_layer = 6 hidden_size = 768 num_heads = 12 @ex.named_config def task_finetune_nlvr2_clip_bert(): exp_name = "finetune_nlvr2" datasets = ["nlvr2"] loss_names = _loss_names({"nlvr2": 1}) batch_size = 256 max_epoch = 10 max_steps = None warmup_steps = 0.1 draw_false_image = 0 learning_rate = 1e-5 lr_mult_head = 10 lr_mult_cross_modal = 5 tokenizer = "bert-base-uncased" max_text_len = 50 input_text_embed_size = 768 vit = 'ViT-B/32' train_transform_keys = ["clip"] val_transform_keys = ["clip"] input_image_embed_size = 768 image_size = 288 @ex.named_config def task_finetune_vqa_clip_bert(): exp_name = "finetune_vqa" datasets = ["vqa"] loss_names = _loss_names({"vqa": 1}) batch_size = 512 max_epoch = 10 max_steps = None warmup_steps = 0.1 draw_false_image = 0 learning_rate = 5e-6 val_check_interval = 0.1 lr_mult_head = 50 lr_mult_cross_modal = 5 tokenizer = "bert-base-uncased" max_text_len = 50 input_text_embed_size = 768 vit = 'ViT-B/32' train_transform_keys = ["clip"] val_transform_keys = ["clip"] input_image_embed_size = 768 image_size = 576 @ex.named_config def task_finetune_irtr_coco_clip_bert(): exp_name = "finetune_irtr_coco" datasets = ["coco"] loss_names = _loss_names({"itm": 0.5, "irtr": 1}) batch_size = 512 max_epoch = 10 max_steps = None warmup_steps = 0.1 get_recall_metric = True draw_false_text = 15 learning_rate = 5e-6 lr_mult_head = 5 lr_mult_cross_modal = 5 tokenizer = "bert-base-uncased" input_text_embed_size = 768 vit = 'ViT-B/32' train_transform_keys = ["clip"] val_transform_keys = ["clip"] input_image_embed_size = 768 image_size = 384 @ex.named_config def task_finetune_irtr_f30k_clip_bert(): exp_name = "finetune_irtr_f30k" datasets = ["f30k"] loss_names = _loss_names({"itm": 0.5, "irtr": 1}) batch_size = 512 max_epoch = 10 max_steps = None warmup_steps = 0.1 get_recall_metric = True draw_false_text = 15 learning_rate = 5e-6 lr_mult_head = 5 lr_mult_cross_modal = 5 tokenizer = "bert-base-uncased" input_text_embed_size = 768 vit = 'ViT-B/32' train_transform_keys = ["clip"] val_transform_keys = ["clip"] input_image_embed_size = 768 image_size = 384 @ex.named_config def task_finetune_snli_clip_bert(): exp_name = "finetune_snli" datasets = ["snli"] loss_names = _loss_names({"snli": 1}) batch_size = 64 max_epoch = 5 max_steps = None warmup_steps = 0.1 draw_false_image = 0 learning_rate = 2e-6 lr_mult_head = 10 lr_mult_cross_modal = 5 tokenizer = "bert-base-uncased" max_text_len = 50 input_text_embed_size = 768 vit = 'ViT-B/32' train_transform_keys = ["clip"] val_transform_keys = ["clip"] input_image_embed_size = 768 image_size = 384 # Named configs for "etc" which are orthogonal to "env" and "task", need to be added at the end # vision encoder @ex.named_config def swin32_base224(): vit = "swin_base_patch4_window7_224_in22k" patch_size = 32 image_size = 224 train_transform_keys = ["imagenet"] val_transform_keys = ["imagenet"] input_image_embed_size = 1024 resolution_before = 224 @ex.named_config def swin32_base384(): vit = "swin_base_patch4_window12_384_in22k" patch_size = 32 image_size = 384 train_transform_keys = ["imagenet"] val_transform_keys = ["imagenet"] input_image_embed_size = 1024 resolution_before = 384 @ex.named_config def swin32_large384(): vit = "swin_large_patch4_window12_384_in22k" patch_size = 32 image_size = 384 train_transform_keys = ["imagenet"] val_transform_keys = ["imagenet"] input_image_embed_size = 1536 resolution_before = 384 @ex.named_config def clip32(): vit = 'ViT-B/32' image_size = 224 patch_size = 32 train_transform_keys = ["clip"] val_transform_keys = ["clip"] input_image_embed_size = 768 @ex.named_config def clip16(): vit = 'ViT-B/16' image_size = 224 patch_size = 16 train_transform_keys = ["clip"] val_transform_keys = ["clip"] input_image_embed_size = 768 # text encoder @ex.named_config def text_roberta(): tokenizer = "roberta-base" vocab_size = 50265 input_text_embed_size = 768 @ex.named_config def text_roberta_large(): tokenizer = "roberta-large" vocab_size = 50265 input_text_embed_size = 1024 # random augmentation @ex.named_config def imagenet_randaug(): train_transform_keys = ["imagenet_randaug"] @ex.named_config def clip_randaug(): train_transform_keys = ["clip_randaug"]
7,425
23.671096
123
py
Mr.Right
Mr.Right-main/models/METER/meter/__init__.py
0
0
0
py
Mr.Right
Mr.Right-main/models/METER/meter/modules/clip_model.py
from collections import OrderedDict from typing import Tuple, Union import numpy as np import torch import torch.nn.functional as F from torch import nn class LayerNorm(nn.LayerNorm): """Subclass torch's LayerNorm to handle fp16.""" def forward(self, x: torch.Tensor): orig_type = x.dtype ret = super().forward(x.type(torch.float32)) return ret.type(orig_type) class QuickGELU(nn.Module): def forward(self, x: torch.Tensor): return x * torch.sigmoid(1.702 * x) class ResidualAttentionBlock(nn.Module): def __init__(self, d_model: int, n_head: int, attn_mask: torch.Tensor = None): super().__init__() self.attn = nn.MultiheadAttention(d_model, n_head) self.ln_1 = LayerNorm(d_model) self.mlp = nn.Sequential(OrderedDict([ ("c_fc", nn.Linear(d_model, d_model * 4)), ("gelu", QuickGELU()), ("c_proj", nn.Linear(d_model * 4, d_model)) ])) self.ln_2 = LayerNorm(d_model) self.attn_mask = attn_mask def attention(self, x: torch.Tensor, x_mask:torch.Tensor): if x_mask is not None: x_mask = x_mask.to(dtype=torch.bool, device=x.device) self.attn_mask = self.attn_mask.to(dtype=x.dtype, device=x.device) if self.attn_mask is not None else None return self.attn(x, x, x, need_weights=False, attn_mask=self.attn_mask, key_padding_mask=x_mask)[0] def forward(self, x: torch.Tensor, x_mask:torch.Tensor=None): x = x + self.attention(self.ln_1(x), x_mask) x = x + self.mlp(self.ln_2(x)) return x class Transformer(nn.Module): def __init__(self, width: int, layers: int, heads: int, attn_mask: torch.Tensor = None): super().__init__() self.width = width self.layers = layers self.resblocks = nn.Sequential(*[ResidualAttentionBlock(width, heads, attn_mask) for _ in range(layers-1)]) def forward(self, x: torch.Tensor, x_mask: torch.Tensor=None): for block in self.resblocks: x = block(x, x_mask) return x class VisualTransformer(nn.Module): def __init__(self, input_resolution: int, patch_size: int, width: int, layers: int, heads: int, output_dim: int, resolution_after: int): super().__init__() self.input_resolution = input_resolution self.output_dim = output_dim self.conv1 = nn.Conv2d(in_channels=3, out_channels=width, kernel_size=patch_size, stride=patch_size, bias=False) scale = width ** -0.5 self.class_embedding = nn.Parameter(scale * torch.randn(width)) self.positional_embedding = nn.Parameter(scale * torch.randn((resolution_after // patch_size) ** 2 + 1, width)) self.ln_pre = LayerNorm(width) self.transformer = Transformer(width, layers, heads) self.ln_post = LayerNorm(width) def forward(self, x: torch.Tensor, x_mask): x = self.conv1(x) # shape = [*, width, grid, grid] x = x.reshape(x.shape[0], x.shape[1], -1) # shape = [*, width, grid ** 2] x = x.permute(0, 2, 1) # shape = [*, grid ** 2, width] t=self.class_embedding.to(x.dtype) + torch.zeros(x.shape[0], 1, x.shape[-1], dtype=x.dtype, device=x.device) x = torch.cat([t, x], dim=1) # shape = [*, grid ** 2 + 1, width] x = x + self.positional_embedding.to(x.dtype) x = self.ln_pre(x) x = x.permute(1, 0, 2) # NLD -> LND x = self.transformer(x, x_mask) x = x.permute(1, 0, 2) # LND -> NLD x = self.ln_post(x) return x class CLIP(nn.Module): def __init__(self, embed_dim: int, # vision image_resolution: int, vision_layers: Union[Tuple[int, int, int, int], int], vision_width: int, vision_patch_size: int, # text context_length: int, vocab_size: int, transformer_width: int, transformer_heads: int, transformer_layers: int, resolution_after=224, ): super().__init__() self.context_length = context_length vision_heads = vision_width // 64 self.visual = VisualTransformer( input_resolution=image_resolution, patch_size=vision_patch_size, width=vision_width, layers=vision_layers, heads=vision_heads, output_dim=embed_dim, resolution_after=resolution_after, ) self.vocab_size = vocab_size self.token_embedding = nn.Embedding(vocab_size, transformer_width) self.positional_embedding = nn.Parameter(torch.empty(self.context_length, transformer_width)) self.ln_final = LayerNorm(transformer_width) self.initialize_parameters() def initialize_parameters(self): nn.init.normal_(self.token_embedding.weight, std=0.02) nn.init.normal_(self.positional_embedding, std=0.01) proj_std = (self.visual.transformer.width ** -0.5) * ((2 * self.visual.transformer.layers) ** -0.5) attn_std = self.visual.transformer.width ** -0.5 fc_std = (2 * self.visual.transformer.width) ** -0.5 for block in self.visual.transformer.resblocks: nn.init.normal_(block.attn.in_proj_weight, std=attn_std) nn.init.normal_(block.attn.out_proj.weight, std=proj_std) nn.init.normal_(block.mlp.c_fc.weight, std=fc_std) nn.init.normal_(block.mlp.c_proj.weight, std=proj_std) @property def dtype(self): return self.visual.conv1.weight.dtype def forward(self, image, image_mask=None): return self.visual(image.type(self.dtype), image_mask) _MODELS = { "ViT-B/32": "https://openaipublic.azureedge.net/clip/models/40d365715913c9da98579312b702a82c18be219cc2a73407c4526f58eba950af/ViT-B-32.pt", "ViT-B/16": "https://openaipublic.azureedge.net/clip/models/5806e77cd80f8b59890b7e101eabd078d9fb84e6937f9e85e4ecb61988df416f/ViT-B-16.pt", } import os import hashlib import urllib from tqdm import tqdm import warnings def _download(url: str, root: str = os.path.expanduser("~/.cache/clip")): os.makedirs(root, exist_ok=True) filename = os.path.basename(url) expected_sha256 = url.split("/")[-2] download_target = os.path.join(root, filename) if os.path.exists(download_target) and not os.path.isfile(download_target): raise RuntimeError(f"{download_target} exists and is not a regular file") if os.path.isfile(download_target): if hashlib.sha256(open(download_target, "rb").read()).hexdigest() == expected_sha256: return download_target else: warnings.warn(f"{download_target} exists, but the SHA256 checksum does not match; re-downloading the file") with urllib.request.urlopen(url) as source, open(download_target, "wb") as output: with tqdm(total=int(source.info().get("Content-Length")), ncols=80, unit='iB', unit_scale=True) as loop: while True: buffer = source.read(8192) if not buffer: break output.write(buffer) loop.update(len(buffer)) if hashlib.sha256(open(download_target, "rb").read()).hexdigest() != expected_sha256: raise RuntimeError(f"Model has been downloaded but the SHA256 checksum does not not match") return download_target def adapt_position_encoding(model, patch_size=32, after=384, suffix='visual.positional_embedding'): keys = [k for k in model if k.endswith(suffix)] assert len(keys) == 1 key = keys[0] origin_pos_embed = model[key] origin_dim2 = False if len(origin_pos_embed.shape) == 2: origin_dim2 = True origin_pos_embed = origin_pos_embed.unsqueeze(0) grid_before = int(np.sqrt(origin_pos_embed.shape[1] - 1)) before = int(grid_before*patch_size) assert (before % patch_size) == 0 grid_after = after // patch_size assert (after % patch_size) == 0 embed_dim = origin_pos_embed.shape[-1] pos_embed = origin_pos_embed[0, 1:, :].reshape((grid_before, grid_before, embed_dim)) new_size = (grid_after, grid_after) pos_embed = torch.nn.functional.interpolate(pos_embed.permute((2, 0, 1)).unsqueeze(0), size=new_size, mode='bicubic') pos_embed = pos_embed.squeeze(0).permute((1, 2, 0)).reshape((-1, embed_dim)) pos_embed = torch.cat((origin_pos_embed[0, 0:1, :], pos_embed), dim=0).unsqueeze(0) assert pos_embed.shape == (1, grid_after * grid_after + 1, embed_dim) if origin_dim2: assert pos_embed.shape[0] == 1 pos_embed = pos_embed.squeeze(0) model[key] = pos_embed return model def build_model(name, resolution_after=224): if name in _MODELS: model_path = _download(_MODELS[name]) elif os.path.isfile(name): model_path = name else: raise RuntimeError(f"Model {name} not found; available models = {available_models()}" ) try: model = torch.jit.load(model_path, map_location="cpu") state_dict = None except RuntimeError: if jit: warnings.warn(f"File {model_path} is not a JIT archive. Loading as a state dict instead") jit = False state_dict = torch.load(model_path, map_location="cpu") state_dict = state_dict or model.state_dict() vit = "visual.proj" in state_dict vision_width = state_dict["visual.conv1.weight"].shape[0] vision_layers = len([k for k in state_dict.keys() if k.startswith("visual.") and k.endswith(".attn.in_proj_weight")]) vision_patch_size = state_dict["visual.conv1.weight"].shape[-1] grid_size = round((state_dict["visual.positional_embedding"].shape[0] - 1) ** 0.5) image_resolution = vision_patch_size * grid_size embed_dim = state_dict["text_projection"].shape[1] context_length = state_dict["positional_embedding"].shape[0] vocab_size = state_dict["token_embedding.weight"].shape[0] transformer_width = state_dict["ln_final.weight"].shape[0] transformer_heads = transformer_width // 64 transformer_layers = len(set(k.split(".")[2] for k in state_dict if k.startswith(f"transformer.resblocks"))) model = CLIP( embed_dim, image_resolution, vision_layers, vision_width, vision_patch_size, context_length, vocab_size, transformer_width, transformer_heads, transformer_layers, resolution_after, ) for key in ["input_resolution", "context_length", "vocab_size"]: if key in state_dict: del state_dict[key] model_dict = model.state_dict() pretrained_dict = state_dict if resolution_after != image_resolution: pretrained_dict = adapt_position_encoding(pretrained_dict, after=resolution_after, patch_size=vision_patch_size) # 1. filter out unnecessary keys pretrained_dict = {k: v for k, v in pretrained_dict.items() if k in model_dict} # 2. overwrite entries in the existing state dict model_dict.update(pretrained_dict) # 3. load the new state dict model.load_state_dict(model_dict) return model
11,209
39.179211
142
py
Mr.Right
Mr.Right-main/models/METER/meter/modules/meter_utils.py
import torch import random from transformers.optimization import AdamW from transformers import ( get_polynomial_decay_schedule_with_warmup, get_cosine_schedule_with_warmup, ) from .dist_utils import all_gather from .objectives import compute_irtr_recall from ..gadgets.my_metrics import Accuracy, VQAScore, Scalar def set_metrics(pl_module): for split in ["train", "val"]: for k, v in pl_module.hparams.config["loss_names"].items(): if v <= 0: continue if k == "vqa": setattr(pl_module, f"{split}_vqa_score", VQAScore()) setattr(pl_module, f"{split}_{k}_loss", Scalar()) elif k == "nlvr2": if split == "train": setattr(pl_module, f"train_{k}_accuracy", Accuracy()) setattr(pl_module, f"train_{k}_loss", Scalar()) else: setattr(pl_module, f"dev_{k}_accuracy", Accuracy()) setattr(pl_module, f"dev_{k}_loss", Scalar()) setattr(pl_module, f"test_{k}_accuracy", Accuracy()) setattr(pl_module, f"test_{k}_loss", Scalar()) elif k == "snli": if split == "train": setattr(pl_module, f"train_{k}_accuracy", Accuracy()) setattr(pl_module, f"train_{k}_loss", Scalar()) else: setattr(pl_module, f"dev_{k}_accuracy", Accuracy()) setattr(pl_module, f"dev_{k}_loss", Scalar()) setattr(pl_module, f"test_{k}_accuracy", Accuracy()) setattr(pl_module, f"test_{k}_loss", Scalar()) elif k == "irtr": setattr(pl_module, f"{split}_irtr_loss", Scalar()) elif k == "mppd" or k == "mpfr": setattr(pl_module, f"{split}_{k}_loss", Scalar()) elif k == "itm": setattr(pl_module, f"{split}_{k}_accuracy", Accuracy()) setattr(pl_module, f"{split}_{k}_loss", Scalar()) else: setattr(pl_module, f"{split}_{k}_accuracy", Accuracy()) setattr(pl_module, f"{split}_{k}_loss", Scalar()) def epoch_wrapup(pl_module): phase = "train" if pl_module.training else "val" the_metric = 0 if pl_module.hparams.config["get_recall_metric"] and not pl_module.training: (ir_r1, ir_r5, ir_r10, tr_r1, tr_r5, tr_r10) = compute_irtr_recall(pl_module) print((ir_r1, ir_r5, ir_r10, tr_r1, tr_r5, tr_r10), pl_module.global_step) pl_module.logger.experiment.add_scalar( "recalls/ir_r1", ir_r1, pl_module.global_step ) pl_module.logger.experiment.add_scalar( "recalls/ir_r5", ir_r5, pl_module.global_step ) pl_module.logger.experiment.add_scalar( "recalls/ir_r10", ir_r10, pl_module.global_step ) pl_module.logger.experiment.add_scalar( "recalls/tr_r1", tr_r1, pl_module.global_step ) pl_module.logger.experiment.add_scalar( "recalls/tr_r5", tr_r5, pl_module.global_step ) pl_module.logger.experiment.add_scalar( "recalls/tr_r10", tr_r10, pl_module.global_step ) the_metric += ir_r1.item() + tr_r1.item() for loss_name, v in pl_module.hparams.config["loss_names"].items(): if v <= 0: continue value = 0 if loss_name == "vqa": value = getattr(pl_module, f"{phase}_{loss_name}_score").compute() pl_module.log(f"{loss_name}/{phase}/score_epoch", value) getattr(pl_module, f"{phase}_{loss_name}_score").reset() pl_module.log( f"{loss_name}/{phase}/loss_epoch", getattr(pl_module, f"{phase}_{loss_name}_loss").compute(), ) getattr(pl_module, f"{phase}_{loss_name}_loss").reset() elif loss_name == "nlvr2" or loss_name == 'snli': if phase == "train": value = getattr(pl_module, f"train_{loss_name}_accuracy").compute() pl_module.log(f"{loss_name}/train/accuracy_epoch", value) getattr(pl_module, f"train_{loss_name}_accuracy").reset() pl_module.log( f"{loss_name}/train/loss_epoch", getattr(pl_module, f"train_{loss_name}_loss").compute(), ) getattr(pl_module, f"train_{loss_name}_loss").reset() else: value = getattr(pl_module, f"test_{loss_name}_accuracy").compute() pl_module.log(f"{loss_name}/test/accuracy_epoch", value) getattr(pl_module, f"test_{loss_name}_accuracy").reset() pl_module.log( f"{loss_name}/test/loss_epoch", getattr(pl_module, f"test_{loss_name}_loss").compute(), ) getattr(pl_module, f"test_{loss_name}_loss").reset() value = getattr(pl_module, f"dev_{loss_name}_accuracy").compute() pl_module.log(f"{loss_name}/dev/accuracy_epoch", value) getattr(pl_module, f"dev_{loss_name}_accuracy").reset() pl_module.log( f"{loss_name}/dev/loss_epoch", getattr(pl_module, f"dev_{loss_name}_loss").compute(), ) getattr(pl_module, f"dev_{loss_name}_loss").reset() elif loss_name == "irtr": pl_module.log( f"{loss_name}/{phase}/irtr_loss_epoch", getattr(pl_module, f"{phase}_irtr_loss").compute(), ) getattr(pl_module, f"{phase}_irtr_loss").reset() elif loss_name == "mppd" or loss_name == "mpfr": pl_module.log( f"{loss_name}/{phase}/loss_epoch", getattr(pl_module, f"{phase}_{loss_name}_loss").compute(), ) getattr(pl_module, f"{phase}_{loss_name}_loss").reset() elif loss_name == "itm": value = getattr(pl_module, f"{phase}_{loss_name}_accuracy").compute() pl_module.log(f"{loss_name}/{phase}/accuracy_epoch", value) getattr(pl_module, f"{phase}_{loss_name}_accuracy").reset() pl_module.log( f"{loss_name}/{phase}/loss_epoch", getattr(pl_module, f"{phase}_{loss_name}_loss").compute(), ) getattr(pl_module, f"{phase}_{loss_name}_loss").reset() else: value = getattr(pl_module, f"{phase}_{loss_name}_accuracy").compute() pl_module.log(f"{loss_name}/{phase}/accuracy_epoch", value) getattr(pl_module, f"{phase}_{loss_name}_accuracy").reset() pl_module.log( f"{loss_name}/{phase}/loss_epoch", getattr(pl_module, f"{phase}_{loss_name}_loss").compute(), ) getattr(pl_module, f"{phase}_{loss_name}_loss").reset() the_metric += value pl_module.log(f"{phase}/the_metric", the_metric) def check_non_acc_grad(pl_module): if pl_module.token_type_embeddings.weight.grad is None: return True else: grad = pl_module.token_type_embeddings.weight.grad return (grad.sum() == 0).item() def set_task(pl_module): pl_module.current_tasks = [ k for k, v in pl_module.hparams.config["loss_names"].items() if v > 0 ] return def set_schedule(pl_module): lr = pl_module.hparams.config["learning_rate"] wd = pl_module.hparams.config["weight_decay"] no_decay = [ "bias", "LayerNorm.bias", "LayerNorm.weight", "norm.bias", "norm.weight", "norm1.bias", "norm1.weight", "norm2.bias", "norm2.weight", ] head_names = ["vqa_classifier", "nlvr2_classifier", "mlm_score", "itm_score", "snli_classifier"] cross_modal_names = ['cross_modal'] lr_mult_head = pl_module.hparams.config["lr_mult_head"] lr_mult_cross_modal = pl_module.hparams.config["lr_mult_cross_modal"] end_lr = pl_module.hparams.config["end_lr"] decay_power = pl_module.hparams.config["decay_power"] optim_type = pl_module.hparams.config["optim_type"] optimizer_grouped_parameters = [ { "params": [ p for n, p in pl_module.named_parameters() if not any(nd in n for nd in no_decay) and not any(bb in n for bb in head_names) and not any(ht in n for ht in cross_modal_names) ], "weight_decay": wd, "lr": lr, }, { "params": [ p for n, p in pl_module.named_parameters() if any(nd in n for nd in no_decay) and not any(bb in n for bb in head_names) and not any(ht in n for ht in cross_modal_names) ], "weight_decay": 0.0, "lr": lr, }, { "params": [ p for n, p in pl_module.named_parameters() if not any(nd in n for nd in no_decay) and any(bb in n for bb in head_names) and not any(ht in n for ht in cross_modal_names) ], "weight_decay": wd, "lr": lr * lr_mult_head, }, { "params": [ p for n, p in pl_module.named_parameters() if any(nd in n for nd in no_decay) and any(bb in n for bb in head_names) and not any(ht in n for ht in cross_modal_names) ], "weight_decay": 0.0, "lr": lr * lr_mult_head, }, { "params": [ p for n, p in pl_module.named_parameters() if not any(nd in n for nd in no_decay) and not any(bb in n for bb in head_names) and any(ht in n for ht in cross_modal_names) ], "weight_decay": wd, "lr": lr * lr_mult_cross_modal, }, { "params": [ p for n, p in pl_module.named_parameters() if any(nd in n for nd in no_decay) and not any(bb in n for bb in head_names) and any(ht in n for ht in cross_modal_names) ], "weight_decay": 0.0, "lr": lr * lr_mult_cross_modal, }, ] if optim_type == "adamw": optimizer = AdamW( optimizer_grouped_parameters, lr=lr, eps=1e-8, betas=(0.9, 0.98) ) elif optim_type == "adam": optimizer = torch.optim.Adam(optimizer_grouped_parameters, lr=lr) elif optim_type == "sgd": optimizer = torch.optim.SGD(optimizer_grouped_parameters, lr=lr, momentum=0.9) if pl_module.trainer.max_steps is None: max_steps = ( len(pl_module.trainer.datamodule.train_dataloader()) * pl_module.trainer.max_epochs // pl_module.trainer.accumulate_grad_batches ) else: max_steps = pl_module.trainer.max_steps warmup_steps = pl_module.hparams.config["warmup_steps"] if isinstance(pl_module.hparams.config["warmup_steps"], float): warmup_steps = int(max_steps * warmup_steps) if decay_power == "cosine": scheduler = get_cosine_schedule_with_warmup( optimizer, num_warmup_steps=warmup_steps, num_training_steps=max_steps, ) else: scheduler = get_polynomial_decay_schedule_with_warmup( optimizer, num_warmup_steps=warmup_steps, num_training_steps=max_steps, lr_end=end_lr, power=decay_power, ) sched = {"scheduler": scheduler, "interval": "step"} return ( [optimizer], [sched], )
11,926
38.363036
100
py
Mr.Right
Mr.Right-main/models/METER/meter/modules/swin_transformer.py
""" Swin Transformer A PyTorch impl of : `Swin Transformer: Hierarchical Vision Transformer using Shifted Windows` - https://arxiv.org/pdf/2103.14030 Code/weights from https://github.com/microsoft/Swin-Transformer, original copyright/license info below """ # -------------------------------------------------------- # Swin Transformer # Copyright (c) 2021 Microsoft # Licensed under The MIT License [see LICENSE for details] # Written by Ze Liu # -------------------------------------------------------- import logging import math from copy import deepcopy from typing import Optional import torch import torch.nn as nn import torch.utils.checkpoint as checkpoint from timm.data import IMAGENET_DEFAULT_MEAN, IMAGENET_DEFAULT_STD from timm.models.helpers import build_model_with_cfg, overlay_external_default_cfg from .swin_helpers import swin_build_model_with_cfg from timm.models.layers import PatchEmbed, Mlp, DropPath, to_2tuple, trunc_normal_ from timm.models.registry import register_model from timm.models.vision_transformer import checkpoint_filter_fn, _init_vit_weights _logger = logging.getLogger(__name__) def _cfg(url='', **kwargs): return { 'url': url, 'num_classes': 1000, 'input_size': (3, 224, 224), 'pool_size': None, 'crop_pct': .9, 'interpolation': 'bicubic', 'fixed_input_size': True, 'mean': IMAGENET_DEFAULT_MEAN, 'std': IMAGENET_DEFAULT_STD, 'first_conv': 'patch_embed.proj', 'classifier': 'head', **kwargs } default_cfgs = { # patch models (my experiments) 'swin_base_patch4_window12_384': _cfg( url='https://github.com/SwinTransformer/storage/releases/download/v1.0.0/swin_base_patch4_window12_384_22kto1k.pth', input_size=(3, 384, 384), crop_pct=1.0), 'swin_base_patch4_window7_224': _cfg( url='https://github.com/SwinTransformer/storage/releases/download/v1.0.0/swin_base_patch4_window7_224_22kto1k.pth', ), 'swin_large_patch4_window12_384': _cfg( url='https://github.com/SwinTransformer/storage/releases/download/v1.0.0/swin_large_patch4_window12_384_22kto1k.pth', input_size=(3, 384, 384), crop_pct=1.0), 'swin_large_patch4_window7_224': _cfg( url='https://github.com/SwinTransformer/storage/releases/download/v1.0.0/swin_large_patch4_window7_224_22kto1k.pth', ), 'swin_small_patch4_window7_224': _cfg( url='https://github.com/SwinTransformer/storage/releases/download/v1.0.0/swin_small_patch4_window7_224.pth', ), 'swin_tiny_patch4_window7_224': _cfg( url='https://github.com/SwinTransformer/storage/releases/download/v1.0.0/swin_tiny_patch4_window7_224.pth', ), 'swin_base_patch4_window12_384_in22k': _cfg( url='https://github.com/SwinTransformer/storage/releases/download/v1.0.0/swin_base_patch4_window12_384_22k.pth', input_size=(3, 384, 384), crop_pct=1.0, num_classes=21841), 'swin_base_patch4_window7_224_in22k': _cfg( url='https://github.com/SwinTransformer/storage/releases/download/v1.0.0/swin_base_patch4_window7_224_22k.pth', num_classes=21841), 'swin_large_patch4_window12_384_in22k': _cfg( url='https://github.com/SwinTransformer/storage/releases/download/v1.0.0/swin_large_patch4_window12_384_22k.pth', input_size=(3, 384, 384), crop_pct=1.0, num_classes=21841), 'swin_large_patch4_window7_224_in22k': _cfg( url='https://github.com/SwinTransformer/storage/releases/download/v1.0.0/swin_large_patch4_window7_224_22k.pth', num_classes=21841), } def window_partition(x, window_size: int): """ Args: x: (B, H, W, C) window_size (int): window size Returns: windows: (num_windows*B, window_size, window_size, C) """ B, H, W, C = x.shape x = x.view(B, H // window_size, window_size, W // window_size, window_size, C) windows = x.permute(0, 1, 3, 2, 4, 5).contiguous().view(-1, window_size, window_size, C) return windows def window_reverse(windows, window_size: int, H: int, W: int): """ Args: windows: (num_windows*B, window_size, window_size, C) window_size (int): Window size H (int): Height of image W (int): Width of image Returns: x: (B, H, W, C) """ B = int(windows.shape[0] / (H * W / window_size / window_size)) x = windows.view(B, H // window_size, W // window_size, window_size, window_size, -1) x = x.permute(0, 1, 3, 2, 4, 5).contiguous().view(B, H, W, -1) return x class WindowAttention(nn.Module): r""" Window based multi-head self attention (W-MSA) module with relative position bias. It supports both of shifted and non-shifted window. Args: dim (int): Number of input channels. window_size (tuple[int]): The height and width of the window. num_heads (int): Number of attention heads. qkv_bias (bool, optional): If True, add a learnable bias to query, key, value. Default: True attn_drop (float, optional): Dropout ratio of attention weight. Default: 0.0 proj_drop (float, optional): Dropout ratio of output. Default: 0.0 """ def __init__(self, dim, window_size, num_heads, qkv_bias=True, attn_drop=0., proj_drop=0.): super().__init__() self.dim = dim self.window_size = window_size # Wh, Ww self.num_heads = num_heads head_dim = dim // num_heads self.scale = head_dim ** -0.5 # define a parameter table of relative position bias self.relative_position_bias_table = nn.Parameter( torch.zeros((2 * window_size[0] - 1) * (2 * window_size[1] - 1), num_heads)) # 2*Wh-1 * 2*Ww-1, nH # get pair-wise relative position index for each token inside the window coords_h = torch.arange(self.window_size[0]) coords_w = torch.arange(self.window_size[1]) coords = torch.stack(torch.meshgrid([coords_h, coords_w])) # 2, Wh, Ww coords_flatten = torch.flatten(coords, 1) # 2, Wh*Ww relative_coords = coords_flatten[:, :, None] - coords_flatten[:, None, :] # 2, Wh*Ww, Wh*Ww relative_coords = relative_coords.permute(1, 2, 0).contiguous() # Wh*Ww, Wh*Ww, 2 relative_coords[:, :, 0] += self.window_size[0] - 1 # shift to start from 0 relative_coords[:, :, 1] += self.window_size[1] - 1 relative_coords[:, :, 0] *= 2 * self.window_size[1] - 1 relative_position_index = relative_coords.sum(-1) # Wh*Ww, Wh*Ww self.register_buffer("relative_position_index", relative_position_index) self.qkv = nn.Linear(dim, dim * 3, bias=qkv_bias) self.attn_drop = nn.Dropout(attn_drop) self.proj = nn.Linear(dim, dim) self.proj_drop = nn.Dropout(proj_drop) trunc_normal_(self.relative_position_bias_table, std=.02) self.softmax = nn.Softmax(dim=-1) def forward(self, x, mask: Optional[torch.Tensor] = None): """ Args: x: input features with shape of (num_windows*B, N, C) mask: (0/-inf) mask with shape of (num_windows, Wh*Ww, Wh*Ww) or None """ B_, N, C = x.shape qkv = self.qkv(x).reshape(B_, N, 3, self.num_heads, C // self.num_heads).permute(2, 0, 3, 1, 4) q, k, v = qkv[0], qkv[1], qkv[2] # make torchscript happy (cannot use tensor as tuple) q = q * self.scale attn = (q @ k.transpose(-2, -1)) relative_position_bias = self.relative_position_bias_table[self.relative_position_index.view(-1)].view( self.window_size[0] * self.window_size[1], self.window_size[0] * self.window_size[1], -1) # Wh*Ww,Wh*Ww,nH relative_position_bias = relative_position_bias.permute(2, 0, 1).contiguous() # nH, Wh*Ww, Wh*Ww attn = attn + relative_position_bias.unsqueeze(0) if mask is not None: nW = mask.shape[0] attn = attn.view(B_ // nW, nW, self.num_heads, N, N) + mask.unsqueeze(1).unsqueeze(0) attn = attn.view(-1, self.num_heads, N, N) attn = self.softmax(attn) else: attn = self.softmax(attn) attn = self.attn_drop(attn) x = (attn @ v).transpose(1, 2).reshape(B_, N, C) x = self.proj(x) x = self.proj_drop(x) return x class SwinTransformerBlock(nn.Module): r""" Swin Transformer Block. Args: dim (int): Number of input channels. input_resolution (tuple[int]): Input resulotion. num_heads (int): Number of attention heads. window_size (int): Window size. shift_size (int): Shift size for SW-MSA. mlp_ratio (float): Ratio of mlp hidden dim to embedding dim. qkv_bias (bool, optional): If True, add a learnable bias to query, key, value. Default: True drop (float, optional): Dropout rate. Default: 0.0 attn_drop (float, optional): Attention dropout rate. Default: 0.0 drop_path (float, optional): Stochastic depth rate. Default: 0.0 act_layer (nn.Module, optional): Activation layer. Default: nn.GELU norm_layer (nn.Module, optional): Normalization layer. Default: nn.LayerNorm """ def __init__(self, dim, input_resolution, num_heads, window_size=7, shift_size=0, mlp_ratio=4., qkv_bias=True, drop=0., attn_drop=0., drop_path=0., act_layer=nn.GELU, norm_layer=nn.LayerNorm): super().__init__() self.dim = dim self.input_resolution = input_resolution self.num_heads = num_heads self.window_size = window_size self.shift_size = shift_size self.mlp_ratio = mlp_ratio if min(self.input_resolution) <= self.window_size: # if window size is larger than input resolution, we don't partition windows self.shift_size = 0 self.window_size = min(self.input_resolution) assert 0 <= self.shift_size < self.window_size, "shift_size must in 0-window_size" self.norm1 = norm_layer(dim) self.attn = WindowAttention( dim, window_size=to_2tuple(self.window_size), num_heads=num_heads, qkv_bias=qkv_bias, attn_drop=attn_drop, proj_drop=drop) self.drop_path = DropPath(drop_path) if drop_path > 0. else nn.Identity() self.norm2 = norm_layer(dim) mlp_hidden_dim = int(dim * mlp_ratio) self.mlp = Mlp(in_features=dim, hidden_features=mlp_hidden_dim, act_layer=act_layer, drop=drop) if self.shift_size > 0: # calculate attention mask for SW-MSA H, W = self.input_resolution img_mask = torch.zeros((1, H, W, 1)) # 1 H W 1 h_slices = (slice(0, -self.window_size), slice(-self.window_size, -self.shift_size), slice(-self.shift_size, None)) w_slices = (slice(0, -self.window_size), slice(-self.window_size, -self.shift_size), slice(-self.shift_size, None)) cnt = 0 for h in h_slices: for w in w_slices: img_mask[:, h, w, :] = cnt cnt += 1 mask_windows = window_partition(img_mask, self.window_size) # nW, window_size, window_size, 1 mask_windows = mask_windows.view(-1, self.window_size * self.window_size) attn_mask = mask_windows.unsqueeze(1) - mask_windows.unsqueeze(2) attn_mask = attn_mask.masked_fill(attn_mask != 0, float(-100.0)).masked_fill(attn_mask == 0, float(0.0)) else: attn_mask = None self.register_buffer("attn_mask", attn_mask) def forward(self, x): H, W = self.input_resolution B, L, C = x.shape assert L == H * W, "input feature has wrong size" shortcut = x x = self.norm1(x) x = x.view(B, H, W, C) # cyclic shift if self.shift_size > 0: shifted_x = torch.roll(x, shifts=(-self.shift_size, -self.shift_size), dims=(1, 2)) else: shifted_x = x # partition windows x_windows = window_partition(shifted_x, self.window_size) # nW*B, window_size, window_size, C x_windows = x_windows.view(-1, self.window_size * self.window_size, C) # nW*B, window_size*window_size, C # W-MSA/SW-MSA attn_windows = self.attn(x_windows, mask=self.attn_mask) # nW*B, window_size*window_size, C # merge windows attn_windows = attn_windows.view(-1, self.window_size, self.window_size, C) shifted_x = window_reverse(attn_windows, self.window_size, H, W) # B H' W' C # reverse cyclic shift if self.shift_size > 0: x = torch.roll(shifted_x, shifts=(self.shift_size, self.shift_size), dims=(1, 2)) else: x = shifted_x x = x.view(B, H * W, C) # FFN x = shortcut + self.drop_path(x) x = x + self.drop_path(self.mlp(self.norm2(x))) return x class PatchMerging(nn.Module): r""" Patch Merging Layer. Args: input_resolution (tuple[int]): Resolution of input feature. dim (int): Number of input channels. norm_layer (nn.Module, optional): Normalization layer. Default: nn.LayerNorm """ def __init__(self, input_resolution, dim, norm_layer=nn.LayerNorm): super().__init__() self.input_resolution = input_resolution self.dim = dim self.reduction = nn.Linear(4 * dim, 2 * dim, bias=False) self.norm = norm_layer(4 * dim) def forward(self, x): """ x: B, H*W, C """ H, W = self.input_resolution B, L, C = x.shape assert L == H * W, "input feature has wrong size" assert H % 2 == 0 and W % 2 == 0, f"x size ({H}*{W}) are not even." x = x.view(B, H, W, C) x0 = x[:, 0::2, 0::2, :] # B H/2 W/2 C x1 = x[:, 1::2, 0::2, :] # B H/2 W/2 C x2 = x[:, 0::2, 1::2, :] # B H/2 W/2 C x3 = x[:, 1::2, 1::2, :] # B H/2 W/2 C x = torch.cat([x0, x1, x2, x3], -1) # B H/2 W/2 4*C x = x.view(B, -1, 4 * C) # B H/2*W/2 4*C x = self.norm(x) x = self.reduction(x) return x def extra_repr(self) -> str: return f"input_resolution={self.input_resolution}, dim={self.dim}" def flops(self): H, W = self.input_resolution flops = H * W * self.dim flops += (H // 2) * (W // 2) * 4 * self.dim * 2 * self.dim return flops class BasicLayer(nn.Module): """ A basic Swin Transformer layer for one stage. Args: dim (int): Number of input channels. input_resolution (tuple[int]): Input resolution. depth (int): Number of blocks. num_heads (int): Number of attention heads. window_size (int): Local window size. mlp_ratio (float): Ratio of mlp hidden dim to embedding dim. qkv_bias (bool, optional): If True, add a learnable bias to query, key, value. Default: True drop (float, optional): Dropout rate. Default: 0.0 attn_drop (float, optional): Attention dropout rate. Default: 0.0 drop_path (float | tuple[float], optional): Stochastic depth rate. Default: 0.0 norm_layer (nn.Module, optional): Normalization layer. Default: nn.LayerNorm downsample (nn.Module | None, optional): Downsample layer at the end of the layer. Default: None use_checkpoint (bool): Whether to use checkpointing to save memory. Default: False. """ def __init__(self, dim, input_resolution, depth, num_heads, window_size, mlp_ratio=4., qkv_bias=True, drop=0., attn_drop=0., drop_path=0., norm_layer=nn.LayerNorm, downsample=None, use_checkpoint=False): super().__init__() self.dim = dim self.input_resolution = input_resolution self.depth = depth self.use_checkpoint = use_checkpoint # build blocks self.blocks = nn.ModuleList([ SwinTransformerBlock( dim=dim, input_resolution=input_resolution, num_heads=num_heads, window_size=window_size, shift_size=0 if (i % 2 == 0) else window_size // 2, mlp_ratio=mlp_ratio, qkv_bias=qkv_bias, drop=drop, attn_drop=attn_drop, drop_path=drop_path[i] if isinstance(drop_path, list) else drop_path, norm_layer=norm_layer) for i in range(depth)]) # patch merging layer if downsample is not None: self.downsample = downsample(input_resolution, dim=dim, norm_layer=norm_layer) else: self.downsample = None def forward(self, x): for blk in self.blocks: if not torch.jit.is_scripting() and self.use_checkpoint: x = checkpoint.checkpoint(blk, x) else: x = blk(x) if self.downsample is not None: x = self.downsample(x) return x def extra_repr(self) -> str: return f"dim={self.dim}, input_resolution={self.input_resolution}, depth={self.depth}" class SwinTransformer(nn.Module): r""" Swin Transformer A PyTorch impl of : `Swin Transformer: Hierarchical Vision Transformer using Shifted Windows` - https://arxiv.org/pdf/2103.14030 Args: img_size (int | tuple(int)): Input image size. Default 224 patch_size (int | tuple(int)): Patch size. Default: 4 in_chans (int): Number of input image channels. Default: 3 num_classes (int): Number of classes for classification head. Default: 1000 embed_dim (int): Patch embedding dimension. Default: 96 depths (tuple(int)): Depth of each Swin Transformer layer. num_heads (tuple(int)): Number of attention heads in different layers. window_size (int): Window size. Default: 7 mlp_ratio (float): Ratio of mlp hidden dim to embedding dim. Default: 4 qkv_bias (bool): If True, add a learnable bias to query, key, value. Default: True drop_rate (float): Dropout rate. Default: 0 attn_drop_rate (float): Attention dropout rate. Default: 0 drop_path_rate (float): Stochastic depth rate. Default: 0.1 norm_layer (nn.Module): Normalization layer. Default: nn.LayerNorm. ape (bool): If True, add absolute position embedding to the patch embedding. Default: False patch_norm (bool): If True, add normalization after patch embedding. Default: True use_checkpoint (bool): Whether to use checkpointing to save memory. Default: False """ def __init__(self, img_size=224, patch_size=4, in_chans=3, num_classes=1000, embed_dim=96, depths=(2, 2, 6, 2), num_heads=(3, 6, 12, 24), window_size=7, mlp_ratio=4., qkv_bias=True, drop_rate=0., attn_drop_rate=0., drop_path_rate=0.1, norm_layer=nn.LayerNorm, ape=False, patch_norm=True, use_checkpoint=False, weight_init='', **kwargs): super().__init__() window_size=int(img_size/32) self.num_classes = num_classes self.num_layers = len(depths) self.embed_dim = embed_dim self.ape = ape self.patch_norm = patch_norm self.num_features = int(embed_dim * 2 ** (self.num_layers - 1)) self.mlp_ratio = mlp_ratio # split image into non-overlapping patches self.patch_embed = PatchEmbed( img_size=img_size, patch_size=patch_size, in_chans=in_chans, embed_dim=embed_dim, norm_layer=norm_layer if self.patch_norm else None) num_patches = self.patch_embed.num_patches self.patch_grid = self.patch_embed.grid_size # absolute position embedding if self.ape: self.absolute_pos_embed = nn.Parameter(torch.zeros(1, num_patches, embed_dim)) trunc_normal_(self.absolute_pos_embed, std=.02) else: self.absolute_pos_embed = None self.pos_drop = nn.Dropout(p=drop_rate) # stochastic depth dpr = [x.item() for x in torch.linspace(0, drop_path_rate, sum(depths))] # stochastic depth decay rule # build layers layers = [] for i_layer in range(self.num_layers): layers += [BasicLayer( dim=int(embed_dim * 2 ** i_layer), input_resolution=(self.patch_grid[0] // (2 ** i_layer), self.patch_grid[1] // (2 ** i_layer)), depth=depths[i_layer], num_heads=num_heads[i_layer], window_size=window_size, mlp_ratio=self.mlp_ratio, qkv_bias=qkv_bias, drop=drop_rate, attn_drop=attn_drop_rate, drop_path=dpr[sum(depths[:i_layer]):sum(depths[:i_layer + 1])], norm_layer=norm_layer, downsample=PatchMerging if (i_layer < self.num_layers - 1) else None, use_checkpoint=use_checkpoint) ] self.layers = nn.Sequential(*layers) self.norm = norm_layer(self.num_features) self.avgpool = nn.AdaptiveAvgPool1d(1) assert weight_init in ('jax', 'jax_nlhb', 'nlhb', '') head_bias = -math.log(self.num_classes) if 'nlhb' in weight_init else 0. if weight_init.startswith('jax'): for n, m in self.named_modules(): _init_vit_weights(m, n, head_bias=head_bias, jax_impl=True) else: self.apply(_init_vit_weights) @torch.jit.ignore def no_weight_decay(self): return {'absolute_pos_embed'} @torch.jit.ignore def no_weight_decay_keywords(self): return {'relative_position_bias_table'} def get_classifier(self): return self.head def reset_classifier(self, num_classes, global_pool=''): self.num_classes = num_classes self.head = nn.Linear(self.num_features, num_classes) if num_classes > 0 else nn.Identity() def forward_features(self, x): x = self.patch_embed(x) if self.absolute_pos_embed is not None: x = x + self.absolute_pos_embed x = self.pos_drop(x) x = self.layers(x) x = self.norm(x) # B L C return x def forward(self, x): x = self.forward_features(x) return x def _create_swin_transformer(variant, pretrained=False, default_cfg=None, **kwargs): if default_cfg is None: default_cfg = deepcopy(default_cfgs[variant]) overlay_external_default_cfg(default_cfg, kwargs) default_num_classes = default_cfg['num_classes'] default_img_size = default_cfg['input_size'][-2:] num_classes = kwargs.pop('num_classes', default_num_classes) img_size = kwargs['config']['image_size'] if kwargs.get('features_only', None): raise RuntimeError('features_only not implemented for Vision Transformer models.') model = swin_build_model_with_cfg( SwinTransformer, variant, pretrained, default_cfg=default_cfg, img_size=img_size, num_classes=num_classes, pretrained_filter_fn=checkpoint_filter_fn, pretrained_strict=False, **kwargs) return model @register_model def swin_base_patch4_window12_384(pretrained=False, **kwargs): """ Swin-B @ 384x384, pretrained ImageNet-22k, fine tune 1k """ model_kwargs = dict( patch_size=4, window_size=12, embed_dim=128, depths=(2, 2, 18, 2), num_heads=(4, 8, 16, 32), **kwargs) return _create_swin_transformer('swin_base_patch4_window12_384', pretrained=pretrained, **model_kwargs) @register_model def swin_base_patch4_window7_224(pretrained=False, **kwargs): """ Swin-B @ 224x224, pretrained ImageNet-22k, fine tune 1k """ model_kwargs = dict( patch_size=4, window_size=7, embed_dim=128, depths=(2, 2, 18, 2), num_heads=(4, 8, 16, 32), **kwargs) return _create_swin_transformer('swin_base_patch4_window7_224', pretrained=pretrained, **model_kwargs) @register_model def swin_large_patch4_window12_384(pretrained=False, **kwargs): """ Swin-L @ 384x384, pretrained ImageNet-22k, fine tune 1k """ model_kwargs = dict( patch_size=4, window_size=12, embed_dim=192, depths=(2, 2, 18, 2), num_heads=(6, 12, 24, 48), **kwargs) return _create_swin_transformer('swin_large_patch4_window12_384', pretrained=pretrained, **model_kwargs) @register_model def swin_large_patch4_window7_224(pretrained=False, **kwargs): """ Swin-L @ 224x224, pretrained ImageNet-22k, fine tune 1k """ model_kwargs = dict( patch_size=4, window_size=7, embed_dim=192, depths=(2, 2, 18, 2), num_heads=(6, 12, 24, 48), **kwargs) return _create_swin_transformer('swin_large_patch4_window7_224', pretrained=pretrained, **model_kwargs) @register_model def swin_small_patch4_window7_224(pretrained=False, **kwargs): """ Swin-S @ 224x224, trained ImageNet-1k """ model_kwargs = dict( patch_size=4, window_size=7, embed_dim=96, depths=(2, 2, 18, 2), num_heads=(3, 6, 12, 24), **kwargs) return _create_swin_transformer('swin_small_patch4_window7_224', pretrained=pretrained, **model_kwargs) @register_model def swin_tiny_patch4_window7_224(pretrained=False, **kwargs): """ Swin-T @ 224x224, trained ImageNet-1k """ model_kwargs = dict( patch_size=4, window_size=7, embed_dim=96, depths=(2, 2, 6, 2), num_heads=(3, 6, 12, 24), **kwargs) return _create_swin_transformer('swin_tiny_patch4_window7_224', pretrained=pretrained, **model_kwargs) @register_model def swin_base_patch4_window12_384_in22k(pretrained=False, **kwargs): """ Swin-B @ 384x384, trained ImageNet-22k """ model_kwargs = dict( patch_size=4, window_size=12, embed_dim=128, depths=(2, 2, 18, 2), num_heads=(4, 8, 16, 32), **kwargs) return _create_swin_transformer('swin_base_patch4_window12_384_in22k', pretrained=pretrained, **model_kwargs) @register_model def swin_base_patch4_window7_224_in22k(pretrained=False, **kwargs): """ Swin-B @ 224x224, trained ImageNet-22k """ model_kwargs = dict( patch_size=4, window_size=7, embed_dim=128, depths=(2, 2, 18, 2), num_heads=(4, 8, 16, 32), **kwargs) return _create_swin_transformer('swin_base_patch4_window7_224_in22k', pretrained=pretrained, **model_kwargs) @register_model def swin_large_patch4_window12_384_in22k(pretrained=False, **kwargs): """ Swin-L @ 384x384, trained ImageNet-22k """ model_kwargs = dict( patch_size=4, window_size=12, embed_dim=192, depths=(2, 2, 18, 2), num_heads=(6, 12, 24, 48), **kwargs) return _create_swin_transformer('swin_large_patch4_window12_384_in22k', pretrained=pretrained, **model_kwargs) @register_model def swin_large_patch4_window7_224_in22k(pretrained=False, **kwargs): """ Swin-L @ 224x224, trained ImageNet-22k """ model_kwargs = dict( patch_size=4, window_size=7, embed_dim=192, depths=(2, 2, 18, 2), num_heads=(6, 12, 24, 48), **kwargs) return _create_swin_transformer('swin_large_patch4_window7_224_in22k', pretrained=pretrained, **model_kwargs)
27,086
41.191589
125
py
Mr.Right
Mr.Right-main/models/METER/meter/modules/bert_model.py
# coding=utf-8 # Copyright 2018 The Google AI Language Team Authors and The HuggingFace Inc. team. # Copyright (c) 2018, NVIDIA CORPORATION. All rights reserved. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. # You may obtain a copy of the License at # # http://www.apache.org/licenses/LICENSE-2.0 # # Unless required by applicable law or agreed to in writing, software # distributed under the License is distributed on an "AS IS" BASIS, # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. """PyTorch BERT model. """ import math import os import warnings from dataclasses import dataclass from typing import Optional, Tuple import torch import torch.utils.checkpoint from torch import nn from torch.nn import BCEWithLogitsLoss, CrossEntropyLoss, MSELoss from transformers.activations import ACT2FN from transformers.file_utils import ( ModelOutput, add_code_sample_docstrings, add_start_docstrings, add_start_docstrings_to_model_forward, replace_return_docstrings, ) from transformers.modeling_outputs import ( BaseModelOutputWithPastAndCrossAttentions, BaseModelOutputWithPoolingAndCrossAttentions, CausalLMOutputWithCrossAttentions, MaskedLMOutput, MultipleChoiceModelOutput, NextSentencePredictorOutput, QuestionAnsweringModelOutput, SequenceClassifierOutput, TokenClassifierOutput, ) from transformers.modeling_utils import ( PreTrainedModel, apply_chunking_to_forward, find_pruneable_heads_and_indices, prune_linear_layer, ) from transformers.utils import logging from transformers.models.bert.configuration_bert import BertConfig logger = logging.get_logger(__name__) _CHECKPOINT_FOR_DOC = "bert-base-uncased" _CONFIG_FOR_DOC = "BertConfig" _TOKENIZER_FOR_DOC = "BertTokenizer" BERT_PRETRAINED_MODEL_ARCHIVE_LIST = [ "bert-base-uncased", "bert-large-uncased", "bert-base-cased", "bert-large-cased", "bert-base-multilingual-uncased", "bert-base-multilingual-cased", "bert-base-chinese", "bert-base-german-cased", "bert-large-uncased-whole-word-masking", "bert-large-cased-whole-word-masking", "bert-large-uncased-whole-word-masking-finetuned-squad", "bert-large-cased-whole-word-masking-finetuned-squad", "bert-base-cased-finetuned-mrpc", "bert-base-german-dbmdz-cased", "bert-base-german-dbmdz-uncased", "cl-tohoku/bert-base-japanese", "cl-tohoku/bert-base-japanese-whole-word-masking", "cl-tohoku/bert-base-japanese-char", "cl-tohoku/bert-base-japanese-char-whole-word-masking", "TurkuNLP/bert-base-finnish-cased-v1", "TurkuNLP/bert-base-finnish-uncased-v1", "wietsedv/bert-base-dutch-cased", # See all BERT models at https://huggingface.co/models?filter=bert ] def load_tf_weights_in_bert(model, config, tf_checkpoint_path): """Load tf checkpoints in a pytorch model.""" try: import re import numpy as np import tensorflow as tf except ImportError: logger.error( "Loading a TensorFlow model in PyTorch, requires TensorFlow to be installed. Please see " "https://www.tensorflow.org/install/ for installation instructions." ) raise tf_path = os.path.abspath(tf_checkpoint_path) logger.info(f"Converting TensorFlow checkpoint from {tf_path}") # Load weights from TF model init_vars = tf.train.list_variables(tf_path) names = [] arrays = [] for name, shape in init_vars: logger.info(f"Loading TF weight {name} with shape {shape}") array = tf.train.load_variable(tf_path, name) names.append(name) arrays.append(array) for name, array in zip(names, arrays): name = name.split("/") # adam_v and adam_m are variables used in AdamWeightDecayOptimizer to calculated m and v # which are not required for using pretrained model if any( n in ["adam_v", "adam_m", "AdamWeightDecayOptimizer", "AdamWeightDecayOptimizer_1", "global_step"] for n in name ): logger.info(f"Skipping {'/'.join(name)}") continue pointer = model for m_name in name: if re.fullmatch(r"[A-Za-z]+_\d+", m_name): scope_names = re.split(r"_(\d+)", m_name) else: scope_names = [m_name] if scope_names[0] == "kernel" or scope_names[0] == "gamma": pointer = getattr(pointer, "weight") elif scope_names[0] == "output_bias" or scope_names[0] == "beta": pointer = getattr(pointer, "bias") elif scope_names[0] == "output_weights": pointer = getattr(pointer, "weight") elif scope_names[0] == "squad": pointer = getattr(pointer, "classifier") else: try: pointer = getattr(pointer, scope_names[0]) except AttributeError: logger.info(f"Skipping {'/'.join(name)}") continue if len(scope_names) >= 2: num = int(scope_names[1]) pointer = pointer[num] if m_name[-11:] == "_embeddings": pointer = getattr(pointer, "weight") elif m_name == "kernel": array = np.transpose(array) try: assert ( pointer.shape == array.shape ), f"Pointer shape {pointer.shape} and array shape {array.shape} mismatched" except AssertionError as e: e.args += (pointer.shape, array.shape) raise logger.info(f"Initialize PyTorch weight {name}") pointer.data = torch.from_numpy(array) return model class BertEmbeddings(nn.Module): """Construct the embeddings from word, position and token_type embeddings.""" def __init__(self, config): super().__init__() self.word_embeddings = nn.Embedding(config.vocab_size, config.hidden_size, padding_idx=config.pad_token_id) self.position_embeddings = nn.Embedding(config.max_position_embeddings, config.hidden_size) self.token_type_embeddings = nn.Embedding(config.type_vocab_size, config.hidden_size) # self.LayerNorm is not snake-cased to stick with TensorFlow model variable name and be able to load # any TensorFlow checkpoint file self.LayerNorm = nn.LayerNorm(config.hidden_size, eps=config.layer_norm_eps) self.dropout = nn.Dropout(config.hidden_dropout_prob) # position_ids (1, len position emb) is contiguous in memory and exported when serialized self.register_buffer("position_ids", torch.arange(config.max_position_embeddings).expand((1, -1))) self.position_embedding_type = getattr(config, "position_embedding_type", "absolute") def forward( self, input_ids=None, token_type_ids=None, position_ids=None, inputs_embeds=None, past_key_values_length=0 ): if input_ids is not None: input_shape = input_ids.size() else: input_shape = inputs_embeds.size()[:-1] seq_length = input_shape[1] if position_ids is None: position_ids = self.position_ids[:, past_key_values_length : seq_length + past_key_values_length] if token_type_ids is None: token_type_ids = torch.zeros(input_shape, dtype=torch.long, device=self.position_ids.device) if inputs_embeds is None: inputs_embeds = self.word_embeddings(input_ids) token_type_embeddings = self.token_type_embeddings(token_type_ids) embeddings = inputs_embeds + token_type_embeddings if self.position_embedding_type == "absolute": position_embeddings = self.position_embeddings(position_ids) embeddings += position_embeddings embeddings = self.LayerNorm(embeddings) embeddings = self.dropout(embeddings) return embeddings class BertSelfAttention(nn.Module): def __init__(self, config): super().__init__() if config.hidden_size % config.num_attention_heads != 0 and not hasattr(config, "embedding_size"): raise ValueError( f"The hidden size ({config.hidden_size}) is not a multiple of the number of attention " f"heads ({config.num_attention_heads})" ) self.num_attention_heads = config.num_attention_heads self.attention_head_size = int(config.hidden_size / config.num_attention_heads) self.all_head_size = self.num_attention_heads * self.attention_head_size self.query = nn.Linear(config.hidden_size, self.all_head_size) self.key = nn.Linear(config.hidden_size, self.all_head_size) self.value = nn.Linear(config.hidden_size, self.all_head_size) self.dropout = nn.Dropout(config.attention_probs_dropout_prob) self.position_embedding_type = getattr(config, "position_embedding_type", "absolute") if self.position_embedding_type == "relative_key" or self.position_embedding_type == "relative_key_query": self.max_position_embeddings = config.max_position_embeddings self.distance_embedding = nn.Embedding(2 * config.max_position_embeddings - 1, self.attention_head_size) self.is_decoder = config.is_decoder self.save_attention = False def save_attn_gradients(self, attn_gradients): self.attn_gradients = attn_gradients def get_attn_gradients(self): return self.attn_gradients def save_attention_map(self, attention_map): self.attention_map = attention_map def get_attention_map(self): return self.attention_map def transpose_for_scores(self, x): new_x_shape = x.size()[:-1] + (self.num_attention_heads, self.attention_head_size) x = x.view(*new_x_shape) return x.permute(0, 2, 1, 3) def forward( self, hidden_states, attention_mask=None, head_mask=None, encoder_hidden_states=None, encoder_attention_mask=None, past_key_value=None, output_attentions=False, ): mixed_query_layer = self.query(hidden_states) # If this is instantiated as a cross-attention module, the keys # and values come from an encoder; the attention mask needs to be # such that the encoder's padding tokens are not attended to. is_cross_attention = encoder_hidden_states is not None if is_cross_attention and past_key_value is not None: # reuse k,v, cross_attentions key_layer = past_key_value[0] value_layer = past_key_value[1] attention_mask = encoder_attention_mask elif is_cross_attention: key_layer = self.transpose_for_scores(self.key(encoder_hidden_states)) value_layer = self.transpose_for_scores(self.value(encoder_hidden_states)) attention_mask = encoder_attention_mask elif past_key_value is not None: key_layer = self.transpose_for_scores(self.key(hidden_states)) value_layer = self.transpose_for_scores(self.value(hidden_states)) key_layer = torch.cat([past_key_value[0], key_layer], dim=2) value_layer = torch.cat([past_key_value[1], value_layer], dim=2) else: key_layer = self.transpose_for_scores(self.key(hidden_states)) value_layer = self.transpose_for_scores(self.value(hidden_states)) query_layer = self.transpose_for_scores(mixed_query_layer) if self.is_decoder: # if cross_attention save Tuple(torch.Tensor, torch.Tensor) of all cross attention key/value_states. # Further calls to cross_attention layer can then reuse all cross-attention # key/value_states (first "if" case) # if uni-directional self-attention (decoder) save Tuple(torch.Tensor, torch.Tensor) of # all previous decoder key/value_states. Further calls to uni-directional self-attention # can concat previous decoder key/value_states to current projected key/value_states (third "elif" case) # if encoder bi-directional self-attention `past_key_value` is always `None` past_key_value = (key_layer, value_layer) # Take the dot product between "query" and "key" to get the raw attention scores. attention_scores = torch.matmul(query_layer, key_layer.transpose(-1, -2)) if self.position_embedding_type == "relative_key" or self.position_embedding_type == "relative_key_query": seq_length = hidden_states.size()[1] position_ids_l = torch.arange(seq_length, dtype=torch.long, device=hidden_states.device).view(-1, 1) position_ids_r = torch.arange(seq_length, dtype=torch.long, device=hidden_states.device).view(1, -1) distance = position_ids_l - position_ids_r positional_embedding = self.distance_embedding(distance + self.max_position_embeddings - 1) positional_embedding = positional_embedding.to(dtype=query_layer.dtype) # fp16 compatibility if self.position_embedding_type == "relative_key": relative_position_scores = torch.einsum("bhld,lrd->bhlr", query_layer, positional_embedding) attention_scores = attention_scores + relative_position_scores elif self.position_embedding_type == "relative_key_query": relative_position_scores_query = torch.einsum("bhld,lrd->bhlr", query_layer, positional_embedding) relative_position_scores_key = torch.einsum("bhrd,lrd->bhlr", key_layer, positional_embedding) attention_scores = attention_scores + relative_position_scores_query + relative_position_scores_key attention_scores = attention_scores / math.sqrt(self.attention_head_size) if attention_mask is not None: # Apply the attention mask is (precomputed for all layers in BertModel forward() function) attention_scores = attention_scores + attention_mask # Normalize the attention scores to probabilities. attention_probs = nn.Softmax(dim=-1)(attention_scores) # if True: if self.save_attention: self.save_attention_map(attention_probs) attention_probs.register_hook(self.save_attn_gradients) # This is actually dropping out entire tokens to attend to, which might # seem a bit unusual, but is taken from the original Transformer paper. attention_probs = self.dropout(attention_probs) # Mask heads if we want to if head_mask is not None: attention_probs = attention_probs * head_mask context_layer = torch.matmul(attention_probs, value_layer) context_layer = context_layer.permute(0, 2, 1, 3).contiguous() new_context_layer_shape = context_layer.size()[:-2] + (self.all_head_size,) context_layer = context_layer.view(*new_context_layer_shape) outputs = (context_layer, attention_probs) if output_attentions else (context_layer,) if self.is_decoder: outputs = outputs + (past_key_value,) return outputs class BertSelfOutput(nn.Module): def __init__(self, config): super().__init__() self.dense = nn.Linear(config.hidden_size, config.hidden_size) self.LayerNorm = nn.LayerNorm(config.hidden_size, eps=config.layer_norm_eps) self.dropout = nn.Dropout(config.hidden_dropout_prob) def forward(self, hidden_states, input_tensor): hidden_states = self.dense(hidden_states) hidden_states = self.dropout(hidden_states) hidden_states = self.LayerNorm(hidden_states + input_tensor) return hidden_states class BertAttention(nn.Module): def __init__(self, config): super().__init__() self.self = BertSelfAttention(config) self.output = BertSelfOutput(config) self.pruned_heads = set() def prune_heads(self, heads): if len(heads) == 0: return heads, index = find_pruneable_heads_and_indices( heads, self.self.num_attention_heads, self.self.attention_head_size, self.pruned_heads ) # Prune linear layers self.self.query = prune_linear_layer(self.self.query, index) self.self.key = prune_linear_layer(self.self.key, index) self.self.value = prune_linear_layer(self.self.value, index) self.output.dense = prune_linear_layer(self.output.dense, index, dim=1) # Update hyper params and store pruned heads self.self.num_attention_heads = self.self.num_attention_heads - len(heads) self.self.all_head_size = self.self.attention_head_size * self.self.num_attention_heads self.pruned_heads = self.pruned_heads.union(heads) def forward( self, hidden_states, attention_mask=None, head_mask=None, encoder_hidden_states=None, encoder_attention_mask=None, past_key_value=None, output_attentions=False, ): self_outputs = self.self( hidden_states, attention_mask, head_mask, encoder_hidden_states, encoder_attention_mask, past_key_value, output_attentions, ) attention_output = self.output(self_outputs[0], hidden_states) outputs = (attention_output,) + self_outputs[1:] # add attentions if we output them return outputs class BertIntermediate(nn.Module): def __init__(self, config): super().__init__() self.dense = nn.Linear(config.hidden_size, config.intermediate_size) if isinstance(config.hidden_act, str): self.intermediate_act_fn = ACT2FN[config.hidden_act] else: self.intermediate_act_fn = config.hidden_act def forward(self, hidden_states): hidden_states = self.dense(hidden_states) hidden_states = self.intermediate_act_fn(hidden_states) return hidden_states class BertOutput(nn.Module): def __init__(self, config): super().__init__() self.dense = nn.Linear(config.intermediate_size, config.hidden_size) self.LayerNorm = nn.LayerNorm(config.hidden_size, eps=config.layer_norm_eps) self.dropout = nn.Dropout(config.hidden_dropout_prob) def forward(self, hidden_states, input_tensor): hidden_states = self.dense(hidden_states) hidden_states = self.dropout(hidden_states) hidden_states = self.LayerNorm(hidden_states + input_tensor) return hidden_states class BertCrossLayer(nn.Module): def __init__(self, config): super().__init__() self.chunk_size_feed_forward = config.chunk_size_feed_forward self.seq_len_dim = 1 self.attention = BertAttention(config) self.is_decoder = config.is_decoder self.add_cross_attention = config.add_cross_attention self.crossattention = BertAttention(config) self.intermediate = BertIntermediate(config) self.output = BertOutput(config) def forward( self, hidden_states, encoder_hidden_states, attention_mask=None, encoder_attention_mask=None, output_attentions=False, ): # decoder uni-directional self-attention cached key/values tuple is at positions 1,2 self_attn_past_key_value = None #past_key_value[:2] if past_key_value is not None else None self_attention_outputs = self.attention( hidden_states, attention_mask, head_mask=None, output_attentions=output_attentions, past_key_value=None, ) attention_output = self_attention_outputs[0] # if decoder, the last output is tuple of self-attn cache outputs = self_attention_outputs[1:] # add self attentions if we output attention weights cross_attn_present_key_value = None cross_attention_outputs = self.crossattention( attention_output, attention_mask, None, encoder_hidden_states, encoder_attention_mask, None, output_attentions, ) attention_output = cross_attention_outputs[0] outputs = outputs + cross_attention_outputs[1:-1] # add cross attentions if we output attention weights layer_output = apply_chunking_to_forward( self.feed_forward_chunk, self.chunk_size_feed_forward, self.seq_len_dim, attention_output ) outputs = (layer_output,) + outputs return outputs def feed_forward_chunk(self, attention_output): intermediate_output = self.intermediate(attention_output) layer_output = self.output(intermediate_output, attention_output) return layer_output class BertEncoder(nn.Module): def __init__(self, config): super().__init__() self.config = config self.layer = nn.ModuleList([BertLayer(config) for _ in range(config.num_hidden_layers)]) def forward( self, hidden_states, attention_mask=None, head_mask=None, encoder_hidden_states=None, encoder_attention_mask=None, past_key_values=None, use_cache=None, output_attentions=False, output_hidden_states=False, return_dict=True, ): all_hidden_states = () if output_hidden_states else None all_self_attentions = () if output_attentions else None all_cross_attentions = () if output_attentions and self.config.add_cross_attention else None next_decoder_cache = () if use_cache else None for i, layer_module in enumerate(self.layer): if output_hidden_states: all_hidden_states = all_hidden_states + (hidden_states,) layer_head_mask = head_mask[i] if head_mask is not None else None past_key_value = past_key_values[i] if past_key_values is not None else None if getattr(self.config, "gradient_checkpointing", False) and self.training: if use_cache: logger.warning( "`use_cache=True` is incompatible with `config.gradient_checkpointing=True`. Setting " "`use_cache=False`..." ) use_cache = False def create_custom_forward(module): def custom_forward(*inputs): return module(*inputs, past_key_value, output_attentions) return custom_forward layer_outputs = torch.utils.checkpoint.checkpoint( create_custom_forward(layer_module), hidden_states, attention_mask, layer_head_mask, encoder_hidden_states, encoder_attention_mask, ) else: layer_outputs = layer_module( hidden_states, attention_mask, layer_head_mask, encoder_hidden_states, encoder_attention_mask, past_key_value, output_attentions, ) hidden_states = layer_outputs[0] if use_cache: next_decoder_cache += (layer_outputs[-1],) if output_attentions: all_self_attentions = all_self_attentions + (layer_outputs[1],) if self.config.add_cross_attention: all_cross_attentions = all_cross_attentions + (layer_outputs[2],) if output_hidden_states: all_hidden_states = all_hidden_states + (hidden_states,) if not return_dict: return tuple( v for v in [ hidden_states, next_decoder_cache, all_hidden_states, all_self_attentions, all_cross_attentions, ] if v is not None ) return BaseModelOutputWithPastAndCrossAttentions( last_hidden_state=hidden_states, past_key_values=next_decoder_cache, hidden_states=all_hidden_states, attentions=all_self_attentions, cross_attentions=all_cross_attentions, ) class BertPooler(nn.Module): def __init__(self, config): super().__init__() self.dense = nn.Linear(config.hidden_size, config.hidden_size) self.activation = nn.Tanh() def forward(self, hidden_states): # We "pool" the model by simply taking the hidden state corresponding # to the first token. first_token_tensor = hidden_states[:, 0] pooled_output = self.dense(first_token_tensor) pooled_output = self.activation(pooled_output) return pooled_output class BertPredictionHeadTransform(nn.Module): def __init__(self, config): super().__init__() self.dense = nn.Linear(config.hidden_size, config.hidden_size) if isinstance(config.hidden_act, str): self.transform_act_fn = ACT2FN[config.hidden_act] else: self.transform_act_fn = config.hidden_act self.LayerNorm = nn.LayerNorm(config.hidden_size, eps=config.layer_norm_eps) def forward(self, hidden_states): hidden_states = self.dense(hidden_states) hidden_states = self.transform_act_fn(hidden_states) hidden_states = self.LayerNorm(hidden_states) return hidden_states class BertLMPredictionHead(nn.Module): def __init__(self, config): super().__init__() self.transform = BertPredictionHeadTransform(config) # The output weights are the same as the input embeddings, but there is # an output-only bias for each token. self.decoder = nn.Linear(config.hidden_size, config.vocab_size, bias=False) self.bias = nn.Parameter(torch.zeros(config.vocab_size)) # Need a link between the two variables so that the bias is correctly resized with `resize_token_embeddings` self.decoder.bias = self.bias def forward(self, hidden_states): hidden_states = self.transform(hidden_states) hidden_states = self.decoder(hidden_states) return hidden_states class BertOnlyMLMHead(nn.Module): def __init__(self, config): super().__init__() self.predictions = BertLMPredictionHead(config) def forward(self, sequence_output): prediction_scores = self.predictions(sequence_output) return prediction_scores class BertOnlyNSPHead(nn.Module): def __init__(self, config): super().__init__() self.seq_relationship = nn.Linear(config.hidden_size, 2) def forward(self, pooled_output): seq_relationship_score = self.seq_relationship(pooled_output) return seq_relationship_score class BertPreTrainingHeads(nn.Module): def __init__(self, config): super().__init__() self.predictions = BertLMPredictionHead(config) self.seq_relationship = nn.Linear(config.hidden_size, 2) def forward(self, sequence_output, pooled_output): prediction_scores = self.predictions(sequence_output) seq_relationship_score = self.seq_relationship(pooled_output) return prediction_scores, seq_relationship_score class BertPreTrainedModel(PreTrainedModel): """ An abstract class to handle weights initialization and a simple interface for downloading and loading pretrained models. """ config_class = BertConfig load_tf_weights = load_tf_weights_in_bert base_model_prefix = "bert" _keys_to_ignore_on_load_missing = [r"position_ids"] def _init_weights(self, module): """Initialize the weights""" if isinstance(module, nn.Linear): # Slightly different from the TF version which uses truncated_normal for initialization # cf https://github.com/pytorch/pytorch/pull/5617 module.weight.data.normal_(mean=0.0, std=self.config.initializer_range) if module.bias is not None: module.bias.data.zero_() elif isinstance(module, nn.Embedding): module.weight.data.normal_(mean=0.0, std=self.config.initializer_range) if module.padding_idx is not None: module.weight.data[module.padding_idx].zero_() elif isinstance(module, nn.LayerNorm): module.bias.data.zero_() module.weight.data.fill_(1.0) @dataclass class BertForPreTrainingOutput(ModelOutput): """ Output type of :class:`~transformers.BertForPreTraining`. Args: loss (`optional`, returned when ``labels`` is provided, ``torch.FloatTensor`` of shape :obj:`(1,)`): Total loss as the sum of the masked language modeling loss and the next sequence prediction (classification) loss. prediction_logits (:obj:`torch.FloatTensor` of shape :obj:`(batch_size, sequence_length, config.vocab_size)`): Prediction scores of the language modeling head (scores for each vocabulary token before SoftMax). seq_relationship_logits (:obj:`torch.FloatTensor` of shape :obj:`(batch_size, 2)`): Prediction scores of the next sequence prediction (classification) head (scores of True/False continuation before SoftMax). hidden_states (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_hidden_states=True`` is passed or when ``config.output_hidden_states=True``): Tuple of :obj:`torch.FloatTensor` (one for the output of the embeddings + one for the output of each layer) of shape :obj:`(batch_size, sequence_length, hidden_size)`. Hidden-states of the model at the output of each layer plus the initial embedding outputs. attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True`` is passed or when ``config.output_attentions=True``): Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape :obj:`(batch_size, num_heads, sequence_length, sequence_length)`. Attentions weights after the attention softmax, used to compute the weighted average in the self-attention heads. """ loss: Optional[torch.FloatTensor] = None prediction_logits: torch.FloatTensor = None seq_relationship_logits: torch.FloatTensor = None hidden_states: Optional[Tuple[torch.FloatTensor]] = None attentions: Optional[Tuple[torch.FloatTensor]] = None BERT_START_DOCSTRING = r""" This model inherits from :class:`~transformers.PreTrainedModel`. Check the superclass documentation for the generic methods the library implements for all its model (such as downloading or saving, resizing the input embeddings, pruning heads etc.) This model is also a PyTorch `torch.nn.Module <https://pytorch.org/docs/stable/nn.html#torch.nn.Module>`__ subclass. Use it as a regular PyTorch Module and refer to the PyTorch documentation for all matter related to general usage and behavior. Parameters: config (:class:`~transformers.BertConfig`): Model configuration class with all the parameters of the model. Initializing with a config file does not load the weights associated with the model, only the configuration. Check out the :meth:`~transformers.PreTrainedModel.from_pretrained` method to load the model weights. """ BERT_INPUTS_DOCSTRING = r""" Args: input_ids (:obj:`torch.LongTensor` of shape :obj:`({0})`): Indices of input sequence tokens in the vocabulary. Indices can be obtained using :class:`~transformers.BertTokenizer`. See :meth:`transformers.PreTrainedTokenizer.encode` and :meth:`transformers.PreTrainedTokenizer.__call__` for details. `What are input IDs? <../glossary.html#input-ids>`__ attention_mask (:obj:`torch.FloatTensor` of shape :obj:`({0})`, `optional`): Mask to avoid performing attention on padding token indices. Mask values selected in ``[0, 1]``: - 1 for tokens that are **not masked**, - 0 for tokens that are **masked**. `What are attention masks? <../glossary.html#attention-mask>`__ token_type_ids (:obj:`torch.LongTensor` of shape :obj:`({0})`, `optional`): Segment token indices to indicate first and second portions of the inputs. Indices are selected in ``[0, 1]``: - 0 corresponds to a `sentence A` token, - 1 corresponds to a `sentence B` token. `What are token type IDs? <../glossary.html#token-type-ids>`_ position_ids (:obj:`torch.LongTensor` of shape :obj:`({0})`, `optional`): Indices of positions of each input sequence tokens in the position embeddings. Selected in the range ``[0, config.max_position_embeddings - 1]``. `What are position IDs? <../glossary.html#position-ids>`_ head_mask (:obj:`torch.FloatTensor` of shape :obj:`(num_heads,)` or :obj:`(num_layers, num_heads)`, `optional`): Mask to nullify selected heads of the self-attention modules. Mask values selected in ``[0, 1]``: - 1 indicates the head is **not masked**, - 0 indicates the head is **masked**. inputs_embeds (:obj:`torch.FloatTensor` of shape :obj:`({0}, hidden_size)`, `optional`): Optionally, instead of passing :obj:`input_ids` you can choose to directly pass an embedded representation. This is useful if you want more control over how to convert :obj:`input_ids` indices into associated vectors than the model's internal embedding lookup matrix. output_attentions (:obj:`bool`, `optional`): Whether or not to return the attentions tensors of all attention layers. See ``attentions`` under returned tensors for more detail. output_hidden_states (:obj:`bool`, `optional`): Whether or not to return the hidden states of all layers. See ``hidden_states`` under returned tensors for more detail. return_dict (:obj:`bool`, `optional`): Whether or not to return a :class:`~transformers.file_utils.ModelOutput` instead of a plain tuple. """ @add_start_docstrings( "The bare Bert Model transformer outputting raw hidden-states without any specific head on top.", BERT_START_DOCSTRING, ) class BertModel(BertPreTrainedModel): """ The model can behave as an encoder (with only self-attention) as well as a decoder, in which case a layer of cross-attention is added between the self-attention layers, following the architecture described in `Attention is all you need <https://arxiv.org/abs/1706.03762>`__ by Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N. Gomez, Lukasz Kaiser and Illia Polosukhin. To behave as an decoder the model needs to be initialized with the :obj:`is_decoder` argument of the configuration set to :obj:`True`. To be used in a Seq2Seq model, the model needs to initialized with both :obj:`is_decoder` argument and :obj:`add_cross_attention` set to :obj:`True`; an :obj:`encoder_hidden_states` is then expected as an input to the forward pass. """ def __init__(self, config, add_pooling_layer=True): super().__init__(config) self.config = config self.embeddings = BertEmbeddings(config) self.encoder = BertEncoder(config) self.pooler = BertPooler(config) if add_pooling_layer else None self.init_weights() def get_input_embeddings(self): return self.embeddings.word_embeddings def set_input_embeddings(self, value): self.embeddings.word_embeddings = value def _prune_heads(self, heads_to_prune): """ Prunes heads of the model. heads_to_prune: dict of {layer_num: list of heads to prune in this layer} See base class PreTrainedModel """ for layer, heads in heads_to_prune.items(): self.encoder.layer[layer].attention.prune_heads(heads) @add_start_docstrings_to_model_forward(BERT_INPUTS_DOCSTRING.format("batch_size, sequence_length")) @add_code_sample_docstrings( tokenizer_class=_TOKENIZER_FOR_DOC, checkpoint=_CHECKPOINT_FOR_DOC, output_type=BaseModelOutputWithPoolingAndCrossAttentions, config_class=_CONFIG_FOR_DOC, ) def forward( self, input_ids=None, attention_mask=None, token_type_ids=None, position_ids=None, head_mask=None, inputs_embeds=None, encoder_hidden_states=None, encoder_attention_mask=None, past_key_values=None, use_cache=None, output_attentions=None, output_hidden_states=None, return_dict=None, ): r""" encoder_hidden_states (:obj:`torch.FloatTensor` of shape :obj:`(batch_size, sequence_length, hidden_size)`, `optional`): Sequence of hidden-states at the output of the last layer of the encoder. Used in the cross-attention if the model is configured as a decoder. encoder_attention_mask (:obj:`torch.FloatTensor` of shape :obj:`(batch_size, sequence_length)`, `optional`): Mask to avoid performing attention on the padding token indices of the encoder input. This mask is used in the cross-attention if the model is configured as a decoder. Mask values selected in ``[0, 1]``: - 1 for tokens that are **not masked**, - 0 for tokens that are **masked**. past_key_values (:obj:`tuple(tuple(torch.FloatTensor))` of length :obj:`config.n_layers` with each tuple having 4 tensors of shape :obj:`(batch_size, num_heads, sequence_length - 1, embed_size_per_head)`): Contains precomputed key and value hidden states of the attention blocks. Can be used to speed up decoding. If :obj:`past_key_values` are used, the user can optionally input only the last :obj:`decoder_input_ids` (those that don't have their past key value states given to this model) of shape :obj:`(batch_size, 1)` instead of all :obj:`decoder_input_ids` of shape :obj:`(batch_size, sequence_length)`. use_cache (:obj:`bool`, `optional`): If set to :obj:`True`, :obj:`past_key_values` key value states are returned and can be used to speed up decoding (see :obj:`past_key_values`). """ output_attentions = output_attentions if output_attentions is not None else self.config.output_attentions output_hidden_states = ( output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states ) return_dict = return_dict if return_dict is not None else self.config.use_return_dict if self.config.is_decoder: use_cache = use_cache if use_cache is not None else self.config.use_cache else: use_cache = False if input_ids is not None and inputs_embeds is not None: raise ValueError("You cannot specify both input_ids and inputs_embeds at the same time") elif input_ids is not None: input_shape = input_ids.size() batch_size, seq_length = input_shape elif inputs_embeds is not None: input_shape = inputs_embeds.size()[:-1] batch_size, seq_length = input_shape else: raise ValueError("You have to specify either input_ids or inputs_embeds") device = input_ids.device if input_ids is not None else inputs_embeds.device # past_key_values_length past_key_values_length = past_key_values[0][0].shape[2] if past_key_values is not None else 0 if attention_mask is None: attention_mask = torch.ones(((batch_size, seq_length + past_key_values_length)), device=device) if token_type_ids is None: token_type_ids = torch.zeros(input_shape, dtype=torch.long, device=device) # We can provide a self-attention mask of dimensions [batch_size, from_seq_length, to_seq_length] # ourselves in which case we just need to make it broadcastable to all heads. extended_attention_mask: torch.Tensor = self.get_extended_attention_mask(attention_mask, input_shape, device) # If a 2D or 3D attention mask is provided for the cross-attention # we need to make broadcastable to [batch_size, num_heads, seq_length, seq_length] if self.config.is_decoder and encoder_hidden_states is not None: encoder_batch_size, encoder_sequence_length, _ = encoder_hidden_states.size() encoder_hidden_shape = (encoder_batch_size, encoder_sequence_length) if encoder_attention_mask is None: encoder_attention_mask = torch.ones(encoder_hidden_shape, device=device) encoder_extended_attention_mask = self.invert_attention_mask(encoder_attention_mask) else: encoder_extended_attention_mask = None # Prepare head mask if needed # 1.0 in head_mask indicate we keep the head # attention_probs has shape bsz x n_heads x N x N # input head_mask has shape [num_heads] or [num_hidden_layers x num_heads] # and head_mask is converted to shape [num_hidden_layers x batch x num_heads x seq_length x seq_length] head_mask = self.get_head_mask(head_mask, self.config.num_hidden_layers) embedding_output = self.embeddings( input_ids=input_ids, position_ids=position_ids, token_type_ids=token_type_ids, inputs_embeds=inputs_embeds, past_key_values_length=past_key_values_length, ) encoder_outputs = self.encoder( embedding_output, attention_mask=extended_attention_mask, head_mask=head_mask, encoder_hidden_states=encoder_hidden_states, encoder_attention_mask=encoder_extended_attention_mask, past_key_values=past_key_values, use_cache=use_cache, output_attentions=output_attentions, output_hidden_states=output_hidden_states, return_dict=return_dict, ) sequence_output = encoder_outputs[0] pooled_output = self.pooler(sequence_output) if self.pooler is not None else None if not return_dict: return (sequence_output, pooled_output) + encoder_outputs[1:] return BaseModelOutputWithPoolingAndCrossAttentions( last_hidden_state=sequence_output, pooler_output=pooled_output, past_key_values=encoder_outputs.past_key_values, hidden_states=encoder_outputs.hidden_states, attentions=encoder_outputs.attentions, cross_attentions=encoder_outputs.cross_attentions, ) @add_start_docstrings( """ Bert Model with two heads on top as done during the pretraining: a `masked language modeling` head and a `next sentence prediction (classification)` head. """, BERT_START_DOCSTRING, ) class BertForPreTraining(BertPreTrainedModel): def __init__(self, config): super().__init__(config) self.bert = BertModel(config) self.cls = BertPreTrainingHeads(config) self.init_weights() def get_output_embeddings(self): return self.cls.predictions.decoder def set_output_embeddings(self, new_embeddings): self.cls.predictions.decoder = new_embeddings @add_start_docstrings_to_model_forward(BERT_INPUTS_DOCSTRING.format("batch_size, sequence_length")) @replace_return_docstrings(output_type=BertForPreTrainingOutput, config_class=_CONFIG_FOR_DOC) def forward( self, input_ids=None, attention_mask=None, token_type_ids=None, position_ids=None, head_mask=None, inputs_embeds=None, labels=None, next_sentence_label=None, output_attentions=None, output_hidden_states=None, return_dict=None, ): r""" labels (:obj:`torch.LongTensor` of shape ``(batch_size, sequence_length)``, `optional`): Labels for computing the masked language modeling loss. Indices should be in ``[-100, 0, ..., config.vocab_size]`` (see ``input_ids`` docstring) Tokens with indices set to ``-100`` are ignored (masked), the loss is only computed for the tokens with labels in ``[0, ..., config.vocab_size]`` next_sentence_label (``torch.LongTensor`` of shape ``(batch_size,)``, `optional`): Labels for computing the next sequence prediction (classification) loss. Input should be a sequence pair (see :obj:`input_ids` docstring) Indices should be in ``[0, 1]``: - 0 indicates sequence B is a continuation of sequence A, - 1 indicates sequence B is a random sequence. kwargs (:obj:`Dict[str, any]`, optional, defaults to `{}`): Used to hide legacy arguments that have been deprecated. Returns: Example:: >>> from transformers import BertTokenizer, BertForPreTraining >>> import torch >>> tokenizer = BertTokenizer.from_pretrained('bert-base-uncased') >>> model = BertForPreTraining.from_pretrained('bert-base-uncased') >>> inputs = tokenizer("Hello, my dog is cute", return_tensors="pt") >>> outputs = model(**inputs) >>> prediction_logits = outputs.prediction_logits >>> seq_relationship_logits = outputs.seq_relationship_logits """ return_dict = return_dict if return_dict is not None else self.config.use_return_dict outputs = self.bert( input_ids, attention_mask=attention_mask, token_type_ids=token_type_ids, position_ids=position_ids, head_mask=head_mask, inputs_embeds=inputs_embeds, output_attentions=output_attentions, output_hidden_states=output_hidden_states, return_dict=return_dict, ) sequence_output, pooled_output = outputs[:2] prediction_scores, seq_relationship_score = self.cls(sequence_output, pooled_output) total_loss = None if labels is not None and next_sentence_label is not None: loss_fct = CrossEntropyLoss() masked_lm_loss = loss_fct(prediction_scores.view(-1, self.config.vocab_size), labels.view(-1)) next_sentence_loss = loss_fct(seq_relationship_score.view(-1, 2), next_sentence_label.view(-1)) total_loss = masked_lm_loss + next_sentence_loss if not return_dict: output = (prediction_scores, seq_relationship_score) + outputs[2:] return ((total_loss,) + output) if total_loss is not None else output return BertForPreTrainingOutput( loss=total_loss, prediction_logits=prediction_scores, seq_relationship_logits=seq_relationship_score, hidden_states=outputs.hidden_states, attentions=outputs.attentions, ) @add_start_docstrings( """Bert Model with a `language modeling` head on top for CLM fine-tuning. """, BERT_START_DOCSTRING ) class BertLMHeadModel(BertPreTrainedModel): _keys_to_ignore_on_load_unexpected = [r"pooler"] _keys_to_ignore_on_load_missing = [r"position_ids", r"predictions.decoder.bias"] def __init__(self, config): super().__init__(config) if not config.is_decoder: logger.warning("If you want to use `BertLMHeadModel` as a standalone, add `is_decoder=True.`") self.bert = BertModel(config, add_pooling_layer=False) self.cls = BertOnlyMLMHead(config) self.init_weights() def get_output_embeddings(self): return self.cls.predictions.decoder def set_output_embeddings(self, new_embeddings): self.cls.predictions.decoder = new_embeddings @add_start_docstrings_to_model_forward(BERT_INPUTS_DOCSTRING.format("batch_size, sequence_length")) @replace_return_docstrings(output_type=CausalLMOutputWithCrossAttentions, config_class=_CONFIG_FOR_DOC) def forward( self, input_ids=None, attention_mask=None, token_type_ids=None, position_ids=None, head_mask=None, inputs_embeds=None, encoder_hidden_states=None, encoder_attention_mask=None, labels=None, past_key_values=None, use_cache=None, output_attentions=None, output_hidden_states=None, return_dict=None, ): r""" encoder_hidden_states (:obj:`torch.FloatTensor` of shape :obj:`(batch_size, sequence_length, hidden_size)`, `optional`): Sequence of hidden-states at the output of the last layer of the encoder. Used in the cross-attention if the model is configured as a decoder. encoder_attention_mask (:obj:`torch.FloatTensor` of shape :obj:`(batch_size, sequence_length)`, `optional`): Mask to avoid performing attention on the padding token indices of the encoder input. This mask is used in the cross-attention if the model is configured as a decoder. Mask values selected in ``[0, 1]``: - 1 for tokens that are **not masked**, - 0 for tokens that are **masked**. labels (:obj:`torch.LongTensor` of shape :obj:`(batch_size, sequence_length)`, `optional`): Labels for computing the left-to-right language modeling loss (next word prediction). Indices should be in ``[-100, 0, ..., config.vocab_size]`` (see ``input_ids`` docstring) Tokens with indices set to ``-100`` are ignored (masked), the loss is only computed for the tokens with labels n ``[0, ..., config.vocab_size]`` past_key_values (:obj:`tuple(tuple(torch.FloatTensor))` of length :obj:`config.n_layers` with each tuple having 4 tensors of shape :obj:`(batch_size, num_heads, sequence_length - 1, embed_size_per_head)`): Contains precomputed key and value hidden states of the attention blocks. Can be used to speed up decoding. If :obj:`past_key_values` are used, the user can optionally input only the last :obj:`decoder_input_ids` (those that don't have their past key value states given to this model) of shape :obj:`(batch_size, 1)` instead of all :obj:`decoder_input_ids` of shape :obj:`(batch_size, sequence_length)`. use_cache (:obj:`bool`, `optional`): If set to :obj:`True`, :obj:`past_key_values` key value states are returned and can be used to speed up decoding (see :obj:`past_key_values`). Returns: Example:: >>> from transformers import BertTokenizer, BertLMHeadModel, BertConfig >>> import torch >>> tokenizer = BertTokenizer.from_pretrained('bert-base-cased') >>> config = BertConfig.from_pretrained("bert-base-cased") >>> config.is_decoder = True >>> model = BertLMHeadModel.from_pretrained('bert-base-cased', config=config) >>> inputs = tokenizer("Hello, my dog is cute", return_tensors="pt") >>> outputs = model(**inputs) >>> prediction_logits = outputs.logits """ return_dict = return_dict if return_dict is not None else self.config.use_return_dict if labels is not None: use_cache = False outputs = self.bert( input_ids, attention_mask=attention_mask, token_type_ids=token_type_ids, position_ids=position_ids, head_mask=head_mask, inputs_embeds=inputs_embeds, encoder_hidden_states=encoder_hidden_states, encoder_attention_mask=encoder_attention_mask, past_key_values=past_key_values, use_cache=use_cache, output_attentions=output_attentions, output_hidden_states=output_hidden_states, return_dict=return_dict, ) sequence_output = outputs[0] prediction_scores = self.cls(sequence_output) lm_loss = None if labels is not None: # we are doing next-token prediction; shift prediction scores and input ids by one shifted_prediction_scores = prediction_scores[:, :-1, :].contiguous() labels = labels[:, 1:].contiguous() loss_fct = CrossEntropyLoss() lm_loss = loss_fct(shifted_prediction_scores.view(-1, self.config.vocab_size), labels.view(-1)) if not return_dict: output = (prediction_scores,) + outputs[2:] return ((lm_loss,) + output) if lm_loss is not None else output return CausalLMOutputWithCrossAttentions( loss=lm_loss, logits=prediction_scores, past_key_values=outputs.past_key_values, hidden_states=outputs.hidden_states, attentions=outputs.attentions, cross_attentions=outputs.cross_attentions, ) def prepare_inputs_for_generation(self, input_ids, past=None, attention_mask=None, **model_kwargs): input_shape = input_ids.shape # if model is used as a decoder in encoder-decoder model, the decoder attention mask is created on the fly if attention_mask is None: attention_mask = input_ids.new_ones(input_shape) # cut decoder_input_ids if past is used if past is not None: input_ids = input_ids[:, -1:] return {"input_ids": input_ids, "attention_mask": attention_mask, "past_key_values": past} def _reorder_cache(self, past, beam_idx): reordered_past = () for layer_past in past: reordered_past += (tuple(past_state.index_select(0, beam_idx) for past_state in layer_past),) return reordered_past @add_start_docstrings("""Bert Model with a `language modeling` head on top. """, BERT_START_DOCSTRING) class BertForMaskedLM(BertPreTrainedModel): _keys_to_ignore_on_load_unexpected = [r"pooler"] _keys_to_ignore_on_load_missing = [r"position_ids", r"predictions.decoder.bias"] def __init__(self, config): super().__init__(config) if config.is_decoder: logger.warning( "If you want to use `BertForMaskedLM` make sure `config.is_decoder=False` for " "bi-directional self-attention." ) self.bert = BertModel(config, add_pooling_layer=False) self.cls = BertOnlyMLMHead(config) self.init_weights() def get_output_embeddings(self): return self.cls.predictions.decoder def set_output_embeddings(self, new_embeddings): self.cls.predictions.decoder = new_embeddings @add_start_docstrings_to_model_forward(BERT_INPUTS_DOCSTRING.format("batch_size, sequence_length")) @add_code_sample_docstrings( tokenizer_class=_TOKENIZER_FOR_DOC, checkpoint=_CHECKPOINT_FOR_DOC, output_type=MaskedLMOutput, config_class=_CONFIG_FOR_DOC, ) def forward( self, input_ids=None, attention_mask=None, token_type_ids=None, position_ids=None, head_mask=None, inputs_embeds=None, encoder_hidden_states=None, encoder_attention_mask=None, labels=None, output_attentions=None, output_hidden_states=None, return_dict=None, ): r""" labels (:obj:`torch.LongTensor` of shape :obj:`(batch_size, sequence_length)`, `optional`): Labels for computing the masked language modeling loss. Indices should be in ``[-100, 0, ..., config.vocab_size]`` (see ``input_ids`` docstring) Tokens with indices set to ``-100`` are ignored (masked), the loss is only computed for the tokens with labels in ``[0, ..., config.vocab_size]`` """ return_dict = return_dict if return_dict is not None else self.config.use_return_dict outputs = self.bert( input_ids, attention_mask=attention_mask, token_type_ids=token_type_ids, position_ids=position_ids, head_mask=head_mask, inputs_embeds=inputs_embeds, encoder_hidden_states=encoder_hidden_states, encoder_attention_mask=encoder_attention_mask, output_attentions=output_attentions, output_hidden_states=output_hidden_states, return_dict=return_dict, ) sequence_output = outputs[0] prediction_scores = self.cls(sequence_output) masked_lm_loss = None if labels is not None: loss_fct = CrossEntropyLoss() # -100 index = padding token masked_lm_loss = loss_fct(prediction_scores.view(-1, self.config.vocab_size), labels.view(-1)) if not return_dict: output = (prediction_scores,) + outputs[2:] return ((masked_lm_loss,) + output) if masked_lm_loss is not None else output return MaskedLMOutput( loss=masked_lm_loss, logits=prediction_scores, hidden_states=outputs.hidden_states, attentions=outputs.attentions, ) def prepare_inputs_for_generation(self, input_ids, attention_mask=None, **model_kwargs): input_shape = input_ids.shape effective_batch_size = input_shape[0] # add a dummy token assert self.config.pad_token_id is not None, "The PAD token should be defined for generation" attention_mask = torch.cat([attention_mask, attention_mask.new_zeros((attention_mask.shape[0], 1))], dim=-1) dummy_token = torch.full( (effective_batch_size, 1), self.config.pad_token_id, dtype=torch.long, device=input_ids.device ) input_ids = torch.cat([input_ids, dummy_token], dim=1) return {"input_ids": input_ids, "attention_mask": attention_mask} @add_start_docstrings( """Bert Model with a `next sentence prediction (classification)` head on top. """, BERT_START_DOCSTRING, ) class BertForNextSentencePrediction(BertPreTrainedModel): def __init__(self, config): super().__init__(config) self.bert = BertModel(config) self.cls = BertOnlyNSPHead(config) self.init_weights() @add_start_docstrings_to_model_forward(BERT_INPUTS_DOCSTRING.format("batch_size, sequence_length")) @replace_return_docstrings(output_type=NextSentencePredictorOutput, config_class=_CONFIG_FOR_DOC) def forward( self, input_ids=None, attention_mask=None, token_type_ids=None, position_ids=None, head_mask=None, inputs_embeds=None, labels=None, output_attentions=None, output_hidden_states=None, return_dict=None, **kwargs, ): r""" labels (:obj:`torch.LongTensor` of shape :obj:`(batch_size,)`, `optional`): Labels for computing the next sequence prediction (classification) loss. Input should be a sequence pair (see ``input_ids`` docstring). Indices should be in ``[0, 1]``: - 0 indicates sequence B is a continuation of sequence A, - 1 indicates sequence B is a random sequence. Returns: Example:: >>> from transformers import BertTokenizer, BertForNextSentencePrediction >>> import torch >>> tokenizer = BertTokenizer.from_pretrained('bert-base-uncased') >>> model = BertForNextSentencePrediction.from_pretrained('bert-base-uncased') >>> prompt = "In Italy, pizza served in formal settings, such as at a restaurant, is presented unsliced." >>> next_sentence = "The sky is blue due to the shorter wavelength of blue light." >>> encoding = tokenizer(prompt, next_sentence, return_tensors='pt') >>> outputs = model(**encoding, labels=torch.LongTensor([1])) >>> logits = outputs.logits >>> assert logits[0, 0] < logits[0, 1] # next sentence was random """ if "next_sentence_label" in kwargs: warnings.warn( "The `next_sentence_label` argument is deprecated and will be removed in a future version, use `labels` instead.", FutureWarning, ) labels = kwargs.pop("next_sentence_label") return_dict = return_dict if return_dict is not None else self.config.use_return_dict outputs = self.bert( input_ids, attention_mask=attention_mask, token_type_ids=token_type_ids, position_ids=position_ids, head_mask=head_mask, inputs_embeds=inputs_embeds, output_attentions=output_attentions, output_hidden_states=output_hidden_states, return_dict=return_dict, ) pooled_output = outputs[1] seq_relationship_scores = self.cls(pooled_output) next_sentence_loss = None if labels is not None: loss_fct = CrossEntropyLoss() next_sentence_loss = loss_fct(seq_relationship_scores.view(-1, 2), labels.view(-1)) if not return_dict: output = (seq_relationship_scores,) + outputs[2:] return ((next_sentence_loss,) + output) if next_sentence_loss is not None else output return NextSentencePredictorOutput( loss=next_sentence_loss, logits=seq_relationship_scores, hidden_states=outputs.hidden_states, attentions=outputs.attentions, ) @add_start_docstrings( """ Bert Model transformer with a sequence classification/regression head on top (a linear layer on top of the pooled output) e.g. for GLUE tasks. """, BERT_START_DOCSTRING, ) class BertForSequenceClassification(BertPreTrainedModel): def __init__(self, config): super().__init__(config) self.num_labels = config.num_labels self.config = config self.bert = BertModel(config) self.dropout = nn.Dropout(config.hidden_dropout_prob) self.classifier = nn.Linear(config.hidden_size, config.num_labels) self.init_weights() @add_start_docstrings_to_model_forward(BERT_INPUTS_DOCSTRING.format("batch_size, sequence_length")) @add_code_sample_docstrings( tokenizer_class=_TOKENIZER_FOR_DOC, checkpoint=_CHECKPOINT_FOR_DOC, output_type=SequenceClassifierOutput, config_class=_CONFIG_FOR_DOC, ) def forward( self, input_ids=None, attention_mask=None, token_type_ids=None, position_ids=None, head_mask=None, inputs_embeds=None, labels=None, output_attentions=None, output_hidden_states=None, return_dict=None, ): r""" labels (:obj:`torch.LongTensor` of shape :obj:`(batch_size,)`, `optional`): Labels for computing the sequence classification/regression loss. Indices should be in :obj:`[0, ..., config.num_labels - 1]`. If :obj:`config.num_labels == 1` a regression loss is computed (Mean-Square loss), If :obj:`config.num_labels > 1` a classification loss is computed (Cross-Entropy). """ return_dict = return_dict if return_dict is not None else self.config.use_return_dict outputs = self.bert( input_ids, attention_mask=attention_mask, token_type_ids=token_type_ids, position_ids=position_ids, head_mask=head_mask, inputs_embeds=inputs_embeds, output_attentions=output_attentions, output_hidden_states=output_hidden_states, return_dict=return_dict, ) pooled_output = outputs[1] pooled_output = self.dropout(pooled_output) logits = self.classifier(pooled_output) loss = None if labels is not None: if self.config.problem_type is None: if self.num_labels == 1: self.config.problem_type = "regression" elif self.num_labels > 1 and (labels.dtype == torch.long or labels.dtype == torch.int): self.config.problem_type = "single_label_classification" else: self.config.problem_type = "multi_label_classification" if self.config.problem_type == "regression": loss_fct = MSELoss() if self.num_labels == 1: loss = loss_fct(logits.squeeze(), labels.squeeze()) else: loss = loss_fct(logits, labels) elif self.config.problem_type == "single_label_classification": loss_fct = CrossEntropyLoss() loss = loss_fct(logits.view(-1, self.num_labels), labels.view(-1)) elif self.config.problem_type == "multi_label_classification": loss_fct = BCEWithLogitsLoss() loss = loss_fct(logits, labels) if not return_dict: output = (logits,) + outputs[2:] return ((loss,) + output) if loss is not None else output return SequenceClassifierOutput( loss=loss, logits=logits, hidden_states=outputs.hidden_states, attentions=outputs.attentions, ) @add_start_docstrings( """ Bert Model with a multiple choice classification head on top (a linear layer on top of the pooled output and a softmax) e.g. for RocStories/SWAG tasks. """, BERT_START_DOCSTRING, ) class BertForMultipleChoice(BertPreTrainedModel): def __init__(self, config): super().__init__(config) self.bert = BertModel(config) self.dropout = nn.Dropout(config.hidden_dropout_prob) self.classifier = nn.Linear(config.hidden_size, 1) self.init_weights() @add_start_docstrings_to_model_forward(BERT_INPUTS_DOCSTRING.format("batch_size, num_choices, sequence_length")) @add_code_sample_docstrings( tokenizer_class=_TOKENIZER_FOR_DOC, checkpoint=_CHECKPOINT_FOR_DOC, output_type=MultipleChoiceModelOutput, config_class=_CONFIG_FOR_DOC, ) def forward( self, input_ids=None, attention_mask=None, token_type_ids=None, position_ids=None, head_mask=None, inputs_embeds=None, labels=None, output_attentions=None, output_hidden_states=None, return_dict=None, ): r""" labels (:obj:`torch.LongTensor` of shape :obj:`(batch_size,)`, `optional`): Labels for computing the multiple choice classification loss. Indices should be in ``[0, ..., num_choices-1]`` where :obj:`num_choices` is the size of the second dimension of the input tensors. (See :obj:`input_ids` above) """ return_dict = return_dict if return_dict is not None else self.config.use_return_dict num_choices = input_ids.shape[1] if input_ids is not None else inputs_embeds.shape[1] input_ids = input_ids.view(-1, input_ids.size(-1)) if input_ids is not None else None attention_mask = attention_mask.view(-1, attention_mask.size(-1)) if attention_mask is not None else None token_type_ids = token_type_ids.view(-1, token_type_ids.size(-1)) if token_type_ids is not None else None position_ids = position_ids.view(-1, position_ids.size(-1)) if position_ids is not None else None inputs_embeds = ( inputs_embeds.view(-1, inputs_embeds.size(-2), inputs_embeds.size(-1)) if inputs_embeds is not None else None ) outputs = self.bert( input_ids, attention_mask=attention_mask, token_type_ids=token_type_ids, position_ids=position_ids, head_mask=head_mask, inputs_embeds=inputs_embeds, output_attentions=output_attentions, output_hidden_states=output_hidden_states, return_dict=return_dict, ) pooled_output = outputs[1] pooled_output = self.dropout(pooled_output) logits = self.classifier(pooled_output) reshaped_logits = logits.view(-1, num_choices) loss = None if labels is not None: loss_fct = CrossEntropyLoss() loss = loss_fct(reshaped_logits, labels) if not return_dict: output = (reshaped_logits,) + outputs[2:] return ((loss,) + output) if loss is not None else output return MultipleChoiceModelOutput( loss=loss, logits=reshaped_logits, hidden_states=outputs.hidden_states, attentions=outputs.attentions, ) @add_start_docstrings( """ Bert Model with a token classification head on top (a linear layer on top of the hidden-states output) e.g. for Named-Entity-Recognition (NER) tasks. """, BERT_START_DOCSTRING, ) class BertForTokenClassification(BertPreTrainedModel): _keys_to_ignore_on_load_unexpected = [r"pooler"] def __init__(self, config): super().__init__(config) self.num_labels = config.num_labels self.bert = BertModel(config, add_pooling_layer=False) self.dropout = nn.Dropout(config.hidden_dropout_prob) self.classifier = nn.Linear(config.hidden_size, config.num_labels) self.init_weights() @add_start_docstrings_to_model_forward(BERT_INPUTS_DOCSTRING.format("batch_size, sequence_length")) @add_code_sample_docstrings( tokenizer_class=_TOKENIZER_FOR_DOC, checkpoint=_CHECKPOINT_FOR_DOC, output_type=TokenClassifierOutput, config_class=_CONFIG_FOR_DOC, ) def forward( self, input_ids=None, attention_mask=None, token_type_ids=None, position_ids=None, head_mask=None, inputs_embeds=None, labels=None, output_attentions=None, output_hidden_states=None, return_dict=None, ): r""" labels (:obj:`torch.LongTensor` of shape :obj:`(batch_size, sequence_length)`, `optional`): Labels for computing the token classification loss. Indices should be in ``[0, ..., config.num_labels - 1]``. """ return_dict = return_dict if return_dict is not None else self.config.use_return_dict outputs = self.bert( input_ids, attention_mask=attention_mask, token_type_ids=token_type_ids, position_ids=position_ids, head_mask=head_mask, inputs_embeds=inputs_embeds, output_attentions=output_attentions, output_hidden_states=output_hidden_states, return_dict=return_dict, ) sequence_output = outputs[0] sequence_output = self.dropout(sequence_output) logits = self.classifier(sequence_output) loss = None if labels is not None: loss_fct = CrossEntropyLoss() # Only keep active parts of the loss if attention_mask is not None: active_loss = attention_mask.view(-1) == 1 active_logits = logits.view(-1, self.num_labels) active_labels = torch.where( active_loss, labels.view(-1), torch.tensor(loss_fct.ignore_index).type_as(labels) ) loss = loss_fct(active_logits, active_labels) else: loss = loss_fct(logits.view(-1, self.num_labels), labels.view(-1)) if not return_dict: output = (logits,) + outputs[2:] return ((loss,) + output) if loss is not None else output return TokenClassifierOutput( loss=loss, logits=logits, hidden_states=outputs.hidden_states, attentions=outputs.attentions, ) @add_start_docstrings( """ Bert Model with a span classification head on top for extractive question-answering tasks like SQuAD (a linear layers on top of the hidden-states output to compute `span start logits` and `span end logits`). """, BERT_START_DOCSTRING, ) class BertForQuestionAnswering(BertPreTrainedModel): _keys_to_ignore_on_load_unexpected = [r"pooler"] def __init__(self, config): super().__init__(config) self.num_labels = config.num_labels self.bert = BertModel(config, add_pooling_layer=False) self.qa_outputs = nn.Linear(config.hidden_size, config.num_labels) self.init_weights() @add_start_docstrings_to_model_forward(BERT_INPUTS_DOCSTRING.format("batch_size, sequence_length")) @add_code_sample_docstrings( tokenizer_class=_TOKENIZER_FOR_DOC, checkpoint=_CHECKPOINT_FOR_DOC, output_type=QuestionAnsweringModelOutput, config_class=_CONFIG_FOR_DOC, ) def forward( self, input_ids=None, attention_mask=None, token_type_ids=None, position_ids=None, head_mask=None, inputs_embeds=None, start_positions=None, end_positions=None, output_attentions=None, output_hidden_states=None, return_dict=None, ): r""" start_positions (:obj:`torch.LongTensor` of shape :obj:`(batch_size,)`, `optional`): Labels for position (index) of the start of the labelled span for computing the token classification loss. Positions are clamped to the length of the sequence (:obj:`sequence_length`). Position outside of the sequence are not taken into account for computing the loss. end_positions (:obj:`torch.LongTensor` of shape :obj:`(batch_size,)`, `optional`): Labels for position (index) of the end of the labelled span for computing the token classification loss. Positions are clamped to the length of the sequence (:obj:`sequence_length`). Position outside of the sequence are not taken into account for computing the loss. """ return_dict = return_dict if return_dict is not None else self.config.use_return_dict outputs = self.bert( input_ids, attention_mask=attention_mask, token_type_ids=token_type_ids, position_ids=position_ids, head_mask=head_mask, inputs_embeds=inputs_embeds, output_attentions=output_attentions, output_hidden_states=output_hidden_states, return_dict=return_dict, ) sequence_output = outputs[0] logits = self.qa_outputs(sequence_output) start_logits, end_logits = logits.split(1, dim=-1) start_logits = start_logits.squeeze(-1) end_logits = end_logits.squeeze(-1) total_loss = None if start_positions is not None and end_positions is not None: # If we are on multi-GPU, split add a dimension if len(start_positions.size()) > 1: start_positions = start_positions.squeeze(-1) if len(end_positions.size()) > 1: end_positions = end_positions.squeeze(-1) # sometimes the start/end positions are outside our model inputs, we ignore these terms ignored_index = start_logits.size(1) start_positions.clamp_(0, ignored_index) end_positions.clamp_(0, ignored_index) loss_fct = CrossEntropyLoss(ignore_index=ignored_index) start_loss = loss_fct(start_logits, start_positions) end_loss = loss_fct(end_logits, end_positions) total_loss = (start_loss + end_loss) / 2 if not return_dict: output = (start_logits, end_logits) + outputs[2:] return ((total_loss,) + output) if total_loss is not None else output return QuestionAnsweringModelOutput( loss=total_loss, start_logits=start_logits, end_logits=end_logits, hidden_states=outputs.hidden_states, attentions=outputs.attentions, )
76,774
41.915036
213
py
Mr.Right
Mr.Right-main/models/METER/meter/modules/meter_module.py
import torch import torch.nn as nn import pytorch_lightning as pl import numpy as np import pdb from transformers.models.bert.modeling_bert import BertConfig, BertEmbeddings, BertModel, BertEncoder, BertLayer from .bert_model import BertCrossLayer, BertAttention from . import swin_transformer as swin from . import heads, objectives # meter_utils from .clip_model import build_model, adapt_position_encoding from .swin_helpers import swin_adapt_position_encoding from transformers import RobertaConfig, RobertaModel class METERTransformerSS(pl.LightningModule): def __init__(self, config): super().__init__() self.save_hyperparameters() self.is_clip= (not 'swin' in config['vit']) if 'roberta' in config['tokenizer']: bert_config = RobertaConfig( vocab_size=config["vocab_size"], hidden_size=config["hidden_size"], num_hidden_layers=config["num_layers"], num_attention_heads=config["num_heads"], intermediate_size=config["hidden_size"] * config["mlp_ratio"], max_position_embeddings=config["max_text_len"], hidden_dropout_prob=config["drop_rate"], attention_probs_dropout_prob=config["drop_rate"], ) else: bert_config = BertConfig( vocab_size=config["vocab_size"], hidden_size=config["hidden_size"], num_hidden_layers=config["num_layers"], num_attention_heads=config["num_heads"], intermediate_size=config["hidden_size"] * config["mlp_ratio"], max_position_embeddings=config["max_text_len"], hidden_dropout_prob=config["drop_rate"], attention_probs_dropout_prob=config["drop_rate"], ) resolution_after=config['image_size'] self.cross_modal_text_transform = nn.Linear(config['input_text_embed_size'], config['hidden_size']) self.cross_modal_text_transform.apply(objectives.init_weights) self.cross_modal_image_transform = nn.Linear(config['input_image_embed_size'], config['hidden_size']) self.cross_modal_image_transform.apply(objectives.init_weights) self.token_type_embeddings = nn.Embedding(2, config["hidden_size"]) self.token_type_embeddings.apply(objectives.init_weights) if torch.distributed.is_initialized(): if torch.distributed.get_rank() == 0: if self.is_clip: build_model(config['vit'], resolution_after=resolution_after) else: getattr(swin, self.hparams.config["vit"])( pretrained=True, config=self.hparams.config, ) if 'roberta' in config['tokenizer']: RobertaModel.from_pretrained(config['tokenizer']) else: BertModel.from_pretrained(config['tokenizer']) torch.distributed.barrier() if self.is_clip: self.vit_model = build_model(config['vit'], resolution_after=resolution_after) else: self.vit_model = getattr(swin, self.hparams.config["vit"])( pretrained=True, config=self.hparams.config, ) self.avgpool = nn.AdaptiveAvgPool1d(1) if 'roberta' in config['tokenizer']: self.text_transformer = RobertaModel.from_pretrained(config['tokenizer']) else: self.text_transformer = BertModel.from_pretrained(config['tokenizer']) self.cross_modal_image_layers = nn.ModuleList([BertCrossLayer(bert_config) for _ in range(config['num_top_layer'])]) self.cross_modal_image_layers.apply(objectives.init_weights) self.cross_modal_text_layers = nn.ModuleList([BertCrossLayer(bert_config) for _ in range(config['num_top_layer'])]) self.cross_modal_text_layers.apply(objectives.init_weights) self.cross_modal_image_pooler = heads.Pooler(config["hidden_size"]) self.cross_modal_image_pooler.apply(objectives.init_weights) self.cross_modal_text_pooler = heads.Pooler(config["hidden_size"]) self.cross_modal_text_pooler.apply(objectives.init_weights) if config["loss_names"]["mlm"] > 0: self.mlm_score = heads.MLMHead(bert_config) self.mlm_score.apply(objectives.init_weights) if config["loss_names"]["itm"] > 0: self.itm_score = heads.ITMHead(config["hidden_size"]*2) self.itm_score.apply(objectives.init_weights) hs = self.hparams.config["hidden_size"] if self.hparams.config["loss_names"]["vqa"] > 0: vs = self.hparams.config["vqav2_label_size"] self.vqa_classifier = nn.Sequential( nn.Linear(hs * 2, hs * 2), nn.LayerNorm(hs * 2), nn.GELU(), nn.Linear(hs * 2, vs), ) self.vqa_classifier.apply(objectives.init_weights) # ===================== Downstream ===================== # if self.hparams.config["loss_names"]["nlvr2"] > 0: self.nlvr2_classifier = nn.Sequential( nn.Linear(hs * 4, hs * 2), nn.LayerNorm(hs * 2), nn.GELU(), nn.Linear(hs * 2, 2), ) self.nlvr2_classifier.apply(objectives.init_weights) emb_data = self.token_type_embeddings.weight.data self.token_type_embeddings = nn.Embedding(3, hs) self.token_type_embeddings.apply(objectives.init_weights) self.token_type_embeddings.weight.data[0, :] = emb_data[0, :] self.token_type_embeddings.weight.data[1, :] = emb_data[1, :] self.token_type_embeddings.weight.data[2, :] = emb_data[1, :] if self.hparams.config["loss_names"]["snli"] > 0: self.snli_classifier = nn.Sequential( nn.Linear(hs * 2, hs * 2), nn.LayerNorm(hs * 2), nn.GELU(), nn.Linear(hs * 2, 3), ) self.snli_classifier.apply(objectives.init_weights) if self.hparams.config["loss_names"]["irtr"] > 0: self.rank_output = nn.Linear(hs, 1) self.rank_output.weight.data = self.itm_score.fc.weight.data[1:, :] self.rank_output.bias.data = self.itm_score.fc.bias.data[1:] self.margin = 0.2 for p in self.itm_score.parameters(): p.requires_grad = False # meter_utils.set_metrics(self) self.current_tasks = list() # ===================== load downstream (test_only) ====================== if self.hparams.config["checkpoint"] != "": ckpt = torch.load(self.hparams.config["checkpoint"], map_location="cpu") state_dict = ckpt["state_dict"] if self.is_clip: state_dict = adapt_position_encoding(state_dict, after=resolution_after, patch_size=self.hparams.config['patch_size']) else: state_dict = swin_adapt_position_encoding(state_dict, after=resolution_after, before=config['resolution_before']) msg = self.load_state_dict(state_dict, strict=False) print(msg) def infer( self, batch, mask_text=False, mask_image=False, image_token_type_idx=1, img=None, ): do_mlm = "_mlm" if mask_text else "" # query text query_ids = batch[f"query_ids"] query_masks = batch[f"query_masks"] query_embeds = self.text_transformer.embeddings(input_ids=query_ids) device = query_embeds.device input_shape = query_masks.size() extend_query_masks = self.text_transformer.get_extended_attention_mask(query_masks, input_shape, device) for layer in self.text_transformer.encoder.layer: query_embeds = layer(query_embeds, extend_query_masks)[0] query_embeds = self.cross_modal_text_transform(query_embeds) query_cls = self.cross_modal_text_pooler(query_embeds) # doc text text_ids = batch[f"text_ids{do_mlm}"] # text_labels = batch[f"text_labels{do_mlm}"] text_masks = batch[f"text_masks"] text_embeds = self.text_transformer.embeddings(input_ids=text_ids) device = text_embeds.device input_shape = text_masks.size() extend_text_masks = self.text_transformer.get_extended_attention_mask(text_masks, input_shape, device) for layer in self.text_transformer.encoder.layer: text_embeds = layer(text_embeds, extend_text_masks)[0] text_embeds = self.cross_modal_text_transform(text_embeds) img = batch["image"] image_embeds = self.vit_model(img) image_embeds = self.cross_modal_image_transform(image_embeds) image_masks = torch.ones((image_embeds.size(0), image_embeds.size(1)), dtype=torch.long, device=device) extend_image_masks = self.text_transformer.get_extended_attention_mask(image_masks, image_masks.size(), device) text_embeds, image_embeds = ( text_embeds + self.token_type_embeddings(torch.zeros_like(text_masks)), image_embeds + self.token_type_embeddings( torch.full_like(image_masks, image_token_type_idx) ), ) x, y = text_embeds, image_embeds for text_layer, image_layer in zip(self.cross_modal_text_layers, self.cross_modal_image_layers): x1 = text_layer(x, y, extend_text_masks, extend_image_masks) y1 = image_layer(y, x, extend_image_masks, extend_text_masks) x, y = x1[0], y1[0] text_feats, image_feats = x, y cls_feats_text = self.cross_modal_text_pooler(x) if self.is_clip: cls_feats_image = self.cross_modal_image_pooler(y) else: avg_image_feats = self.avgpool(image_feats.transpose(1, 2)).view(image_feats.size(0), 1, -1) cls_feats_image = self.cross_modal_image_pooler(avg_image_feats) cls_feats = torch.cat([cls_feats_text, cls_feats_image], dim=-1) ret = { "query_embeds": query_embeds, "query_atts": query_masks, "query_cls":query_cls, "text_feats": text_feats, "text_masks": text_masks, "text_cls":cls_feats_text, "multi_embeds": torch.cat([image_feats,text_feats],dim=1), "multi_atts":torch.cat([image_masks,text_masks],dim=-1), "cls_feats_text": cls_feats_text, "image_feats": image_feats, "image_masks": image_masks, "cls_feats_image": cls_feats_image, "img_cls":cls_feats_image, "cls_feats": cls_feats, # "text_labels": text_labels, # "text_ids": text_ids, } return ret @torch.no_grad() def output_multi( self, batch, mask_text=False, mask_image=False, image_token_type_idx=1, img=None,): img = batch["image"] do_mlm = "_mlm" if mask_text else "" # doc text text_ids = batch[f"text_ids{do_mlm}"] text_masks = batch[f"text_masks"] text_embeds = self.text_transformer.embeddings(input_ids=text_ids) device = text_embeds.device input_shape = text_masks.size() extend_text_masks = self.text_transformer.get_extended_attention_mask(text_masks, input_shape, device) for layer in self.text_transformer.encoder.layer: text_embeds = layer(text_embeds, extend_text_masks)[0] text_embeds = self.cross_modal_text_transform(text_embeds) img = batch["image"] image_embeds = self.vit_model(img) image_embeds = self.cross_modal_image_transform(image_embeds) image_masks = torch.ones((image_embeds.size(0), image_embeds.size(1)), dtype=torch.long, device=device) extend_image_masks = self.text_transformer.get_extended_attention_mask(image_masks, image_masks.size(), device) text_embeds, image_embeds = ( text_embeds + self.token_type_embeddings(torch.zeros_like(text_masks)), image_embeds + self.token_type_embeddings( torch.full_like(image_masks, image_token_type_idx) ), ) x, y = text_embeds, image_embeds for text_layer, image_layer in zip(self.cross_modal_text_layers, self.cross_modal_image_layers): x1 = text_layer(x, y, extend_text_masks, extend_image_masks) y1 = image_layer(y, x, extend_image_masks, extend_text_masks) x, y = x1[0], y1[0] text_feats, image_feats = x, y cls_feats_text = self.cross_modal_text_pooler(x) if self.is_clip: cls_feats_image = self.cross_modal_image_pooler(y) else: avg_image_feats = self.avgpool(image_feats.transpose(1, 2)).view(image_feats.size(0), 1, -1) cls_feats_image = self.cross_modal_image_pooler(avg_image_feats) cls_feats = torch.cat([cls_feats_text, cls_feats_image], dim=-1) ret = { "text_feats": text_feats, "text_masks": text_masks, "text_cls":cls_feats_text, "cls_feats_text": cls_feats_text, "image_feats": image_feats, "image_masks": image_masks, "cls_feats_image": cls_feats_image, "cls_feats": cls_feats, "img_cls":cls_feats_image, "multi_embeds": torch.cat([image_feats,text_feats],dim=1), "multi_atts":torch.cat([image_masks,text_masks],dim=-1), } return ret def forward(self, batch): ret = dict() if len(self.current_tasks) == 0: ret.update(self.infer(batch)) return ret # Masked Language Modeling if "mlm" in self.current_tasks: ret.update(objectives.compute_mlm(self, batch)) # Image Text Matching if "itm" in self.current_tasks: ret.update(objectives.compute_itm(self, batch)) # Visual Question Answering if "vqa" in self.current_tasks: ret.update(objectives.compute_vqa(self, batch)) # Natural Language for Visual Reasoning 2 if "nlvr2" in self.current_tasks: ret.update(objectives.compute_nlvr2(self, batch)) # SNLI Visual Entailment if "snli" in self.current_tasks: ret.update(objectives.compute_snli(self, batch)) # Image Retrieval and Text Retrieval if "irtr" in self.current_tasks: ret.update(objectives.compute_irtr(self, batch)) return ret def training_step(self, batch, batch_idx): # meter_utils.set_task(self) output = self(batch) total_loss = sum([v for k, v in output.items() if "loss" in k]) return total_loss def training_epoch_end(self, outs): # meter_utils.epoch_wrapup(self) pass def validation_step(self, batch, batch_idx): # meter_utils.set_task(self) pass output = self(batch) def validation_epoch_end(self, outs): pass # meter_utils.epoch_wrapup(self) def test_step(self, batch, batch_idx): pass meter_utils.set_task(self) output = self(batch) ret = dict() if self.hparams.config["loss_names"]["vqa"] > 0: ret.update(objectives.vqa_test_step(self, batch, output)) return ret def test_epoch_end(self, outs): model_name = self.hparams.config["load_path"].split("/")[-1][:-5] if self.hparams.config["loss_names"]["vqa"] > 0: objectives.vqa_test_wrapup(outs, model_name) # meter_utils.epoch_wrapup(self) def configure_optimizers(self): pass # return meter_utils.set_schedule(self)
15,962
40.141753
134
py
Mr.Right
Mr.Right-main/models/METER/meter/modules/dist_utils.py
# Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved """ This file contains primitives for multi-gpu communication. This is useful when doing distributed training. """ import functools import logging import numpy as np import pickle import torch import torch.distributed as dist import torch _LOCAL_PROCESS_GROUP = None """ A torch process group which only includes processes that on the same machine as the current process. This variable is set when processes are spawned by `launch()` in "engine/launch.py". """ def get_world_size() -> int: if not dist.is_available(): return 1 if not dist.is_initialized(): return 1 return dist.get_world_size() def get_rank() -> int: if not dist.is_available(): return 0 if not dist.is_initialized(): return 0 return dist.get_rank() def get_local_rank() -> int: """ Returns: The rank of the current process within the local (per-machine) process group. """ if not dist.is_available(): return 0 if not dist.is_initialized(): return 0 assert _LOCAL_PROCESS_GROUP is not None return dist.get_rank(group=_LOCAL_PROCESS_GROUP) def get_local_size() -> int: """ Returns: The size of the per-machine process group, i.e. the number of processes per machine. """ if not dist.is_available(): return 1 if not dist.is_initialized(): return 1 return dist.get_world_size(group=_LOCAL_PROCESS_GROUP) def is_main_process() -> bool: return get_rank() == 0 def synchronize(): """ Helper function to synchronize (barrier) among all processes when using distributed training """ if not dist.is_available(): return if not dist.is_initialized(): return world_size = dist.get_world_size() if world_size == 1: return dist.barrier() @functools.lru_cache() def _get_global_gloo_group(): """ Return a process group based on gloo backend, containing all the ranks The result is cached. """ if dist.get_backend() == "nccl": return dist.new_group(backend="gloo") else: return dist.group.WORLD def _serialize_to_tensor(data, group): backend = dist.get_backend(group) assert backend in ["gloo", "nccl"] device = torch.device("cpu" if backend == "gloo" else "cuda") buffer = pickle.dumps(data) if len(buffer) > 1024 ** 3: logger = logging.getLogger(__name__) logger.warning( "Rank {} trying to all-gather {:.2f} GB of data on device {}".format( get_rank(), len(buffer) / (1024 ** 3), device ) ) storage = torch.ByteStorage.from_buffer(buffer) tensor = torch.ByteTensor(storage).to(device=device) return tensor def _pad_to_largest_tensor(tensor, group): """ Returns: list[int]: size of the tensor, on each rank Tensor: padded tensor that has the max size """ world_size = dist.get_world_size(group=group) assert ( world_size >= 1 ), "comm.gather/all_gather must be called from ranks within the given group!" local_size = torch.tensor([tensor.numel()], dtype=torch.int64, device=tensor.device) size_list = [ torch.zeros([1], dtype=torch.int64, device=tensor.device) for _ in range(world_size) ] dist.all_gather(size_list, local_size, group=group) size_list = [int(size.item()) for size in size_list] max_size = max(size_list) # we pad the tensor because torch all_gather does not support # gathering tensors of different shapes if local_size != max_size: padding = torch.zeros( (max_size - local_size,), dtype=torch.uint8, device=tensor.device ) tensor = torch.cat((tensor, padding), dim=0) return size_list, tensor def all_gather(data, group=None): """ Run all_gather on arbitrary picklable data (not necessarily tensors). Args: data: any picklable object group: a torch process group. By default, will use a group which contains all ranks on gloo backend. Returns: list[data]: list of data gathered from each rank """ if get_world_size() == 1: return [data] if group is None: group = _get_global_gloo_group() if dist.get_world_size(group) == 1: return [data] tensor = _serialize_to_tensor(data, group) size_list, tensor = _pad_to_largest_tensor(tensor, group) max_size = max(size_list) # receiving Tensor from all ranks tensor_list = [ torch.empty((max_size,), dtype=torch.uint8, device=tensor.device) for _ in size_list ] dist.all_gather(tensor_list, tensor, group=group) data_list = [] for size, tensor in zip(size_list, tensor_list): buffer = tensor.cpu().numpy().tobytes()[:size] data_list.append(pickle.loads(buffer)) return data_list def gather(data, dst=0, group=None): """ Run gather on arbitrary picklable data (not necessarily tensors). Args: data: any picklable object dst (int): destination rank group: a torch process group. By default, will use a group which contains all ranks on gloo backend. Returns: list[data]: on dst, a list of data gathered from each rank. Otherwise, an empty list. """ if get_world_size() == 1: return [data] if group is None: group = _get_global_gloo_group() if dist.get_world_size(group=group) == 1: return [data] rank = dist.get_rank(group=group) tensor = _serialize_to_tensor(data, group) size_list, tensor = _pad_to_largest_tensor(tensor, group) # receiving Tensor from all ranks if rank == dst: max_size = max(size_list) tensor_list = [ torch.empty((max_size,), dtype=torch.uint8, device=tensor.device) for _ in size_list ] dist.gather(tensor, tensor_list, dst=dst, group=group) data_list = [] for size, tensor in zip(size_list, tensor_list): buffer = tensor.cpu().numpy().tobytes()[:size] data_list.append(pickle.loads(buffer)) return data_list else: dist.gather(tensor, [], dst=dst, group=group) return [] def shared_random_seed(): """ Returns: int: a random number that is the same across all workers. If workers need a shared RNG, they can use this shared seed to create one. All workers must call this function, otherwise it will deadlock. """ ints = np.random.randint(2 ** 31) all_ints = all_gather(ints) return all_ints[0] def reduce_dict(input_dict, average=True): """ Reduce the values in the dictionary from all processes so that process with rank 0 has the reduced results. Args: input_dict (dict): inputs to be reduced. All the values must be scalar CUDA Tensor. average (bool): whether to do average or sum Returns: a dict with the same keys as input_dict, after reduction. """ world_size = get_world_size() if world_size < 2: return input_dict with torch.no_grad(): names = [] values = [] # sort the keys so that they are consistent across processes for k in sorted(input_dict.keys()): names.append(k) values.append(input_dict[k]) values = torch.stack(values, dim=0) dist.reduce(values, dst=0) if dist.get_rank() == 0 and average: # only main process gets accumulated, so only divide by # world_size in this case values /= world_size reduced_dict = {k: v for k, v in zip(names, values)} return reduced_dict
7,814
27.837638
100
py
Mr.Right
Mr.Right-main/models/METER/meter/modules/objectives.py
import torch import torch.nn as nn import torch.nn.functional as F import os import glob import json import tqdm import functools from torch.utils.data.distributed import DistributedSampler from einops import rearrange from .dist_utils import all_gather def compute_mlm(pl_module, batch): infer = pl_module.infer(batch, mask_text=True, mask_image=False) mlm_logits = pl_module.mlm_score(infer["text_feats"]) mlm_labels = infer["text_labels"] mlm_loss = F.cross_entropy( mlm_logits.view(-1, pl_module.hparams.config["vocab_size"]), mlm_labels.view(-1), ignore_index=-100, ) ret = { "mlm_loss": mlm_loss, "mlm_logits": mlm_logits, "mlm_labels": mlm_labels, "mlm_ids": infer["text_ids"], } phase = "train" if pl_module.training else "val" loss = getattr(pl_module, f"{phase}_mlm_loss")(ret["mlm_loss"]) acc = getattr(pl_module, f"{phase}_mlm_accuracy")( ret["mlm_logits"], ret["mlm_labels"] ) pl_module.log(f"mlm/{phase}/loss", loss) pl_module.log(f"mlm/{phase}/accuracy", acc) return ret def compute_itm(pl_module, batch): pos_len = len(batch["text"]) // 2 neg_len = len(batch["text"]) - pos_len itm_labels = torch.cat([torch.ones(pos_len), torch.zeros(neg_len)]).to( pl_module.device ) itm_labels = itm_labels[torch.randperm(itm_labels.size(0))] itm_images = [ torch.stack( [ ti if itm_labels[i] == 1 else fi for i, (ti, fi) in enumerate(zip(bti, bfi)) ] ) for bti, bfi in zip(batch["image"], batch["false_image_0"]) ] batch = {k: v for k, v in batch.items()} batch["image"] = itm_images infer = pl_module.infer(batch, mask_text=False, mask_image=False) itm_logits = pl_module.itm_score(infer["cls_feats"]) itm_loss = F.cross_entropy(itm_logits, itm_labels.long()) ret = { "itm_loss": itm_loss, "itm_logits": itm_logits, "itm_labels": itm_labels, } phase = "train" if pl_module.training else "val" loss = getattr(pl_module, f"{phase}_itm_loss")(ret["itm_loss"]) acc = getattr(pl_module, f"{phase}_itm_accuracy")( ret["itm_logits"], ret["itm_labels"] ) pl_module.log(f"itm/{phase}/loss", loss) pl_module.log(f"itm/{phase}/accuracy", acc) return ret def compute_snli(pl_module, batch): infer = pl_module.infer( batch, mask_text=False, mask_image=False, ) snli_logits = pl_module.snli_classifier(infer["cls_feats"]) snli_labels = batch["labels"] snli_labels = torch.tensor(snli_labels).to(pl_module.device).long() snli_loss = F.cross_entropy(snli_logits, snli_labels.view(-1)) ret = { "snli_loss": snli_loss, "snli_logits": snli_logits, "snli_labels": snli_labels, } phase = "train" if pl_module.training else "val" if phase == "train": loss = getattr(pl_module, f"{phase}_snli_loss")(ret["snli_loss"]) acc = getattr(pl_module, f"{phase}_snli_accuracy")( ret["snli_logits"], ret["snli_labels"] ) pl_module.log(f"snli/{phase}/loss", loss) pl_module.log(f"snli/{phase}/accuracy", acc) else: dev_batches = [i for i, n in enumerate(batch["table_name"]) if "dev" in n] test_batches = [i for i, n in enumerate(batch["table_name"]) if "test" in n] if dev_batches: dev_loss = getattr(pl_module, f"dev_snli_loss")( F.cross_entropy( ret["snli_logits"][dev_batches], ret["snli_labels"][dev_batches] ) ) dev_acc = getattr(pl_module, f"dev_snli_accuracy")( ret["snli_logits"][dev_batches], ret["snli_labels"][dev_batches] ) pl_module.log(f"snli/dev/loss", dev_loss) pl_module.log(f"snli/dev/accuracy", dev_acc) if test_batches: test_loss = getattr(pl_module, f"test_snli_loss")( F.cross_entropy( ret["snli_logits"][test_batches], ret["snli_labels"][test_batches] ) ) test_acc = getattr(pl_module, f"test_snli_accuracy")( ret["snli_logits"][test_batches], ret["snli_labels"][test_batches] ) pl_module.log(f"snli/test/loss", test_loss) pl_module.log(f"snli/test/accuracy", test_acc) return ret def compute_vqa(pl_module, batch): infer = pl_module.infer(batch, mask_text=False, mask_image=False) vqa_logits = pl_module.vqa_classifier(infer["cls_feats"]) vqa_targets = torch.zeros( len(vqa_logits), pl_module.hparams.config["vqav2_label_size"] ).to(pl_module.device) vqa_labels = batch["vqa_labels"] vqa_scores = batch["vqa_scores"] for i, (_label, _score) in enumerate(zip(vqa_labels, vqa_scores)): for l, s in zip(_label, _score): vqa_targets[i, l] = s vqa_loss = ( F.binary_cross_entropy_with_logits(vqa_logits, vqa_targets) * vqa_targets.shape[1] ) # https://github.com/jnhwkim/ban-vqa/blob/master/train.py#L19 ret = { "vqa_loss": vqa_loss, "vqa_logits": vqa_logits, "vqa_targets": vqa_targets, "vqa_labels": vqa_labels, "vqa_scores": vqa_scores, } phase = "train" if pl_module.training else "val" loss = getattr(pl_module, f"{phase}_vqa_loss")(ret["vqa_loss"]) score = getattr(pl_module, f"{phase}_vqa_score")( ret["vqa_logits"], ret["vqa_targets"] ) pl_module.log(f"vqa/{phase}/loss", loss) pl_module.log(f"vqa/{phase}/score", score) return ret def compute_nlvr2(pl_module, batch): infer1 = pl_module.infer( batch, mask_text=False, mask_image=False, image_token_type_idx=1 ) infer2 = pl_module.infer( batch, mask_text=False, mask_image=False, image_token_type_idx=2 ) cls_feats = torch.cat([infer1["cls_feats"], infer2["cls_feats"]], dim=-1) nlvr2_logits = pl_module.nlvr2_classifier(cls_feats) nlvr2_labels = batch["answers"] nlvr2_labels = torch.tensor(nlvr2_labels).to(pl_module.device).long() nlvr2_loss = F.cross_entropy(nlvr2_logits, nlvr2_labels.view(-1)) ret = { "nlvr2_loss": nlvr2_loss, "nlvr2_logits": nlvr2_logits, "nlvr2_labels": nlvr2_labels, } phase = "train" if pl_module.training else "val" if phase == "train": loss = getattr(pl_module, f"{phase}_nlvr2_loss")(ret["nlvr2_loss"]) acc = getattr(pl_module, f"{phase}_nlvr2_accuracy")( ret["nlvr2_logits"], ret["nlvr2_labels"] ) pl_module.log(f"nlvr2/{phase}/loss", loss) pl_module.log(f"nlvr2/{phase}/accuracy", acc) else: dev_batches = [i for i, n in enumerate(batch["table_name"]) if "dev" in n] test_batches = [i for i, n in enumerate(batch["table_name"]) if "test" in n] if dev_batches: dev_loss = getattr(pl_module, f"dev_nlvr2_loss")( F.cross_entropy( ret["nlvr2_logits"][dev_batches], ret["nlvr2_labels"][dev_batches] ) ) dev_acc = getattr(pl_module, f"dev_nlvr2_accuracy")( ret["nlvr2_logits"][dev_batches], ret["nlvr2_labels"][dev_batches] ) pl_module.log(f"nlvr2/dev/loss", dev_loss) pl_module.log(f"nlvr2/dev/accuracy", dev_acc) if test_batches: test_loss = getattr(pl_module, f"test_nlvr2_loss")( F.cross_entropy( ret["nlvr2_logits"][test_batches], ret["nlvr2_labels"][test_batches] ) ) test_acc = getattr(pl_module, f"test_nlvr2_accuracy")( ret["nlvr2_logits"][test_batches], ret["nlvr2_labels"][test_batches] ) pl_module.log(f"nlvr2/test/loss", test_loss) pl_module.log(f"nlvr2/test/accuracy", test_acc) return ret def compute_irtr(pl_module, batch): is_training_phase = pl_module.training _bs, _c, _h, _w = batch["image"][0].shape false_len = pl_module.hparams.config["draw_false_text"] text_ids = torch.stack( [batch[f"false_text_{i}_ids"] for i in range(false_len)], dim=1 ) text_masks = torch.stack( [batch[f"false_text_{i}_masks"] for i in range(false_len)], dim=1 ) text_labels = torch.stack( [batch[f"false_text_{i}_labels"] for i in range(false_len)], dim=1 ) text_ids = torch.cat([batch["text_ids"].unsqueeze(1), text_ids], dim=1) text_masks = torch.cat([batch["text_masks"].unsqueeze(1), text_masks], dim=1) text_labels = torch.cat([batch["text_labels"].unsqueeze(1), text_labels], dim=1) images = batch["image"][0].unsqueeze(1).expand(_bs, false_len + 1, _c, _h, _w) infer = pl_module.infer( { "image": [rearrange(images, "bs fs c h w -> (bs fs) c h w")], "text_ids": rearrange(text_ids, "bs fs tl -> (bs fs) tl"), "text_masks": rearrange(text_masks, "bs fs tl -> (bs fs) tl"), "text_labels": rearrange(text_labels, "bs fs tl -> (bs fs) tl"), } ) score = pl_module.rank_output(infer["cls_feats"])[:, 0] score = rearrange(score, "(bs fs) -> bs fs", bs=_bs, fs=false_len + 1) answer = torch.zeros(_bs).to(score).long() irtr_loss = F.cross_entropy(score, answer) ret = { "irtr_loss": irtr_loss, } phase = "train" if pl_module.training else "val" irtr_loss = getattr(pl_module, f"{phase}_irtr_loss")(ret["irtr_loss"]) pl_module.log(f"irtr/{phase}/irtr_loss", irtr_loss) return ret @torch.no_grad() def compute_irtr_recall(pl_module): text_dset = pl_module.trainer.datamodule.dms[0].make_no_false_val_dset() text_dset.tokenizer = pl_module.trainer.datamodule.dms[0].tokenizer text_loader = torch.utils.data.DataLoader( text_dset, batch_size=64, num_workers=pl_module.hparams.config["num_workers"], pin_memory=True, collate_fn=functools.partial( text_dset.collate, mlm_collator=pl_module.trainer.datamodule.dms[0].mlm_collator, ), ) image_dset = pl_module.trainer.datamodule.dms[0].make_no_false_val_dset( image_only=True ) image_dset.tokenizer = pl_module.trainer.datamodule.dms[0].tokenizer dist_sampler = DistributedSampler(image_dset, shuffle=False) image_loader = torch.utils.data.DataLoader( image_dset, batch_size=1, num_workers=pl_module.hparams.config["num_workers"], sampler=dist_sampler, pin_memory=True, collate_fn=functools.partial( image_dset.collate, mlm_collator=pl_module.trainer.datamodule.dms[0].mlm_collator, ), ) #TODO: speed up the process by caching text/image features text_preload = list() for _b in tqdm.tqdm(text_loader, desc="text prefetch loop"): text_preload.append( { "text_ids": _b["text_ids"].to(pl_module.device), "text_masks": _b["text_masks"].to(pl_module.device), "text_labels": _b["text_labels"].to(pl_module.device), "img_index": _b["img_index"], } ) tiids = list() for pre in text_preload: tiids += pre["img_index"] tiids = torch.tensor(tiids) image_preload = list() for _b in tqdm.tqdm(image_loader, desc="image prefetch loop"): image_preload.append((_b['image'][0], _b["img_index"][0])) rank_scores = list() rank_iids = list() for img_batch in tqdm.tqdm(image_preload, desc="rank loop"): _im, _iid = img_batch img_batch_score = list() for txt_batch in text_preload: fblen = len(txt_batch["text_ids"]) im = _im.repeat(fblen, 1, 1, 1).to(device=txt_batch['text_ids'].device) with torch.cuda.amp.autocast(): score = pl_module.rank_output( pl_module.infer( { "text_ids": txt_batch["text_ids"], "text_masks": txt_batch["text_masks"], "text_labels": txt_batch["text_labels"], }, img=im, )["cls_feats"] )[:, 0] img_batch_score.append(score) img_batch_score = torch.cat(img_batch_score) rank_scores.append(img_batch_score.cpu().tolist()) rank_iids.append(_iid) torch.distributed.barrier() gather_rank_scores = all_gather(rank_scores) gather_rank_iids = all_gather(rank_iids) iids = torch.tensor(gather_rank_iids) iids = iids.view(-1) scores = torch.tensor(gather_rank_scores) scores = scores.view(len(iids), -1) topk10 = scores.topk(10, dim=1) topk5 = scores.topk(5, dim=1) topk1 = scores.topk(1, dim=1) topk10_iids = tiids[topk10.indices] topk5_iids = tiids[topk5.indices] topk1_iids = tiids[topk1.indices] tr_r10 = (iids.unsqueeze(1) == topk10_iids).float().max(dim=1)[0].mean() tr_r5 = (iids.unsqueeze(1) == topk5_iids).float().max(dim=1)[0].mean() tr_r1 = (iids.unsqueeze(1) == topk1_iids).float().max(dim=1)[0].mean() topk10 = scores.topk(10, dim=0) topk5 = scores.topk(5, dim=0) topk1 = scores.topk(1, dim=0) topk10_iids = iids[topk10.indices] topk5_iids = iids[topk5.indices] topk1_iids = iids[topk1.indices] ir_r10 = (tiids.unsqueeze(0) == topk10_iids).float().max(dim=0)[0].mean() ir_r5 = (tiids.unsqueeze(0) == topk5_iids).float().max(dim=0)[0].mean() ir_r1 = (tiids.unsqueeze(0) == topk1_iids).float().max(dim=0)[0].mean() return (ir_r1, ir_r5, ir_r10, tr_r1, tr_r5, tr_r10) def init_weights(module): if isinstance(module, (nn.Linear, nn.Embedding)): module.weight.data.normal_(mean=0.0, std=0.02) elif isinstance(module, nn.LayerNorm): module.bias.data.zero_() module.weight.data.fill_(1.0) if isinstance(module, nn.Linear) and module.bias is not None: module.bias.data.zero_() def vqa_test_step(pl_module, batch, output): try: id2answer = ( pl_module.trainer.datamodule.dm_dicts["vqa_trainval"].id2answer if "vqa_trainval" in pl_module.trainer.datamodule.dm_dicts else pl_module.trainer.datamodule.dm_dicts["vqa"].id2answer ) except: id2answer = ( pl_module.trainer.datamodule.dm_dicts["gqa_test"].id2answer if "gqa_test" in pl_module.trainer.datamodule.dm_dicts else pl_module.trainer.datamodule.dm_dicts["gqa"].id2answer ) vqa_logits = output["vqa_logits"] vqa_preds = vqa_logits.argmax(dim=-1) vqa_preds = [id2answer[pred.item()] for pred in vqa_preds] questions = batch["text"] qids = batch["qid"] return {"qids": qids, "preds": vqa_preds, "gqa": True} vqa_logits = output["vqa_logits"] vqa_preds = vqa_logits.argmax(dim=-1) vqa_preds = [id2answer[pred.item()] for pred in vqa_preds] questions = batch["text"] qids = batch["qid"] return {"qids": qids, "preds": vqa_preds, "gqa": False} def arc_test_step(pl_module, batch, output): return output def vqa_test_wrapup(outs, model_name): rank = torch.distributed.get_rank() qids, preds = list(), list() gqa = False for out in outs: qids += out["qids"] preds += out["preds"] gqa = out['gqa'] rets = list() for qid, pred in zip(qids, preds): if gqa: rets.append({"questionId": qid, "prediction": pred}) else: rets.append({"question_id": qid, "answer": pred}) with open(f"vqa_submit_{rank}.json", "w") as fp: json.dump(rets, fp, indent=4) torch.distributed.barrier() if rank == 0: jsons = list() paths = list(glob.glob("vqa_submit_*.json")) for path in paths: with open(path, "r") as fp: jsons += json.load(fp) os.makedirs("result", exist_ok=True) with open(f"result/vqa_submit_{model_name}.json", "w") as fp: json.dump(jsons, fp, indent=4) torch.distributed.barrier() os.remove(f"vqa_submit_{rank}.json") def arc_test_wrapup(outs, caplen, model_name): rank = torch.distributed.get_rank() iids, captions = list(), list() for out in outs: iids += out["iid"] captions += out["captions"] rets = list() for iid, caption in zip(iids, captions): rets.append({"image_id": iid, "caption": caption}) with open(f"coco_cap_len{caplen}_{rank}.json", "w") as fp: json.dump(rets, fp, indent=4) torch.distributed.barrier() if rank == 0: jsons = list() paths = list(glob.glob(f"coco_cap_len{caplen}_*.json")) for path in paths: with open(path, "r") as fp: jsons += json.load(fp) os.makedirs("result/arc", exist_ok=True) jsons = sorted(jsons, key=lambda x: x["image_id"]) with open(f"result/arc/coco_cap_{model_name}_len{caplen}.json", "w") as fp: json.dump(jsons, fp, indent=4) torch.distributed.barrier() os.remove(f"coco_cap_len{caplen}_{rank}.json")
17,360
33.514911
88
py
Mr.Right
Mr.Right-main/models/METER/meter/modules/swin_helpers.py
""" Model creation / weight loading / state_dict helpers Hacked together by / Copyright 2020 Ross Wightman """ import logging import os import math from collections import OrderedDict from copy import deepcopy from typing import Any, Callable, Optional, Tuple import torch import torch.nn as nn from timm.models.features import FeatureListNet, FeatureDictNet, FeatureHookNet from timm.models.hub import has_hf_hub, download_cached_file, load_state_dict_from_hf, load_state_dict_from_url from timm.models.layers import Conv2dSame, Linear def swin_adapt_position_encoding(model, before=384, patch_size=32, after=384, suffix='relative_position_bias_table'): if after == before: return model grid_before = int(before/32) grid_after = int(after/32) #after // patch_size before = (2*grid_before-1) import math after = (2*grid_after-1) keys = [k for k in model if k.endswith(suffix)] assert len(keys) > 0 for key in keys: pos_embed = model[key] pos_embed = pos_embed.transpose(0, 1).view(-1, before, before) pos_embed = torch.nn.functional.interpolate(pos_embed.unsqueeze(0), size=(after, after), mode='bicubic') pos_embed = pos_embed.squeeze(0).permute((1, 2, 0)) pos_embed = pos_embed.contiguous().view(-1, pos_embed.size(-1)) model[key] = pos_embed keys = [k for k in model if k.endswith('attn_mask')] for key in keys: model.pop(key) keys = [k for k in model if k.endswith('relative_position_index')] for key in keys: model.pop(key) return model _logger = logging.getLogger(__name__) def load_state_dict(checkpoint_path, use_ema=False): if checkpoint_path and os.path.isfile(checkpoint_path): checkpoint = torch.load(checkpoint_path, map_location='cpu') state_dict_key = 'state_dict' if isinstance(checkpoint, dict): if use_ema and 'state_dict_ema' in checkpoint: state_dict_key = 'state_dict_ema' if state_dict_key and state_dict_key in checkpoint: new_state_dict = OrderedDict() for k, v in checkpoint[state_dict_key].items(): # strip `module.` prefix name = k[7:] if k.startswith('module') else k new_state_dict[name] = v state_dict = new_state_dict else: state_dict = checkpoint _logger.info("Loaded {} from checkpoint '{}'".format(state_dict_key, checkpoint_path)) return state_dict else: _logger.error("No checkpoint found at '{}'".format(checkpoint_path)) raise FileNotFoundError() def load_checkpoint(model, checkpoint_path, use_ema=False, strict=True): if os.path.splitext(checkpoint_path)[-1].lower() in ('.npz', '.npy'): # numpy checkpoint, try to load via model specific load_pretrained fn if hasattr(model, 'load_pretrained'): model.load_pretrained(checkpoint_path) else: raise NotImplementedError('Model cannot load numpy checkpoint') return state_dict = load_state_dict(checkpoint_path, use_ema) model.load_state_dict(state_dict, strict=strict) def resume_checkpoint(model, checkpoint_path, optimizer=None, loss_scaler=None, log_info=True): resume_epoch = None if os.path.isfile(checkpoint_path): checkpoint = torch.load(checkpoint_path, map_location='cpu') if isinstance(checkpoint, dict) and 'state_dict' in checkpoint: if log_info: _logger.info('Restoring model state from checkpoint...') new_state_dict = OrderedDict() for k, v in checkpoint['state_dict'].items(): name = k[7:] if k.startswith('module') else k new_state_dict[name] = v model.load_state_dict(new_state_dict) if optimizer is not None and 'optimizer' in checkpoint: if log_info: _logger.info('Restoring optimizer state from checkpoint...') optimizer.load_state_dict(checkpoint['optimizer']) if loss_scaler is not None and loss_scaler.state_dict_key in checkpoint: if log_info: _logger.info('Restoring AMP loss scaler state from checkpoint...') loss_scaler.load_state_dict(checkpoint[loss_scaler.state_dict_key]) if 'epoch' in checkpoint: resume_epoch = checkpoint['epoch'] if 'version' in checkpoint and checkpoint['version'] > 1: resume_epoch += 1 # start at the next epoch, old checkpoints incremented before save if log_info: _logger.info("Loaded checkpoint '{}' (epoch {})".format(checkpoint_path, checkpoint['epoch'])) else: model.load_state_dict(checkpoint) if log_info: _logger.info("Loaded checkpoint '{}'".format(checkpoint_path)) return resume_epoch else: _logger.error("No checkpoint found at '{}'".format(checkpoint_path)) raise FileNotFoundError() def load_custom_pretrained(model, default_cfg=None, load_fn=None, progress=False, check_hash=False): r"""Loads a custom (read non .pth) weight file Downloads checkpoint file into cache-dir like torch.hub based loaders, but calls a passed in custom load fun, or the `load_pretrained` model member fn. If the object is already present in `model_dir`, it's deserialized and returned. The default value of `model_dir` is ``<hub_dir>/checkpoints`` where `hub_dir` is the directory returned by :func:`~torch.hub.get_dir`. Args: model: The instantiated model to load weights into default_cfg (dict): Default pretrained model cfg load_fn: An external stand alone fn that loads weights into provided model, otherwise a fn named 'laod_pretrained' on the model will be called if it exists progress (bool, optional): whether or not to display a progress bar to stderr. Default: False check_hash(bool, optional): If True, the filename part of the URL should follow the naming convention ``filename-<sha256>.ext`` where ``<sha256>`` is the first eight or more digits of the SHA256 hash of the contents of the file. The hash is used to ensure unique names and to verify the contents of the file. Default: False """ default_cfg = default_cfg or getattr(model, 'default_cfg', None) or {} pretrained_url = default_cfg.get('url', None) if not pretrained_url: _logger.warning("No pretrained weights exist for this model. Using random initialization.") return cached_file = download_cached_file(default_cfg['url'], check_hash=check_hash, progress=progress) if load_fn is not None: load_fn(model, cached_file) elif hasattr(model, 'load_pretrained'): model.load_pretrained(cached_file) else: _logger.warning("Valid function to load pretrained weights is not available, using random initialization.") def adapt_input_conv(in_chans, conv_weight): conv_type = conv_weight.dtype conv_weight = conv_weight.float() # Some weights are in torch.half, ensure it's float for sum on CPU O, I, J, K = conv_weight.shape if in_chans == 1: if I > 3: assert conv_weight.shape[1] % 3 == 0 # For models with space2depth stems conv_weight = conv_weight.reshape(O, I // 3, 3, J, K) conv_weight = conv_weight.sum(dim=2, keepdim=False) else: conv_weight = conv_weight.sum(dim=1, keepdim=True) elif in_chans != 3: if I != 3: raise NotImplementedError('Weight format not supported by conversion.') else: # NOTE this strategy should be better than random init, but there could be other combinations of # the original RGB input layer weights that'd work better for specific cases. repeat = int(math.ceil(in_chans / 3)) conv_weight = conv_weight.repeat(1, repeat, 1, 1)[:, :in_chans, :, :] conv_weight *= (3 / float(in_chans)) conv_weight = conv_weight.to(conv_type) return conv_weight def load_pretrained(model, img_size, default_cfg=None, num_classes=1000, in_chans=3, filter_fn=None, strict=True, progress=False, resolution_before=384): """ Load pretrained checkpoint Args: model (nn.Module) : PyTorch model module default_cfg (Optional[Dict]): default configuration for pretrained weights / target dataset num_classes (int): num_classes for model in_chans (int): in_chans for model filter_fn (Optional[Callable]): state_dict filter fn for load (takes state_dict, model as args) strict (bool): strict load of checkpoint progress (bool): enable progress bar for weight download """ default_cfg = default_cfg or getattr(model, 'default_cfg', None) or {} pretrained_url = default_cfg.get('url', None) hf_hub_id = default_cfg.get('hf_hub', None) if not pretrained_url and not hf_hub_id: _logger.warning("No pretrained weights exist for this model. Using random initialization.") return if hf_hub_id and has_hf_hub(necessary=not pretrained_url): _logger.info(f'Loading pretrained weights from Hugging Face hub ({hf_hub_id})') state_dict = load_state_dict_from_hf(hf_hub_id) else: _logger.info(f'Loading pretrained weights from url ({pretrained_url})') state_dict = load_state_dict_from_url(pretrained_url, progress=progress, map_location='cpu') swin_adapt_position_encoding(state_dict['model'], before=resolution_before, after=img_size) if filter_fn is not None: # for backwards compat with filter fn that take one arg, try one first, the two try: state_dict = filter_fn(state_dict) except TypeError: state_dict = filter_fn(state_dict, model) input_convs = default_cfg.get('first_conv', None) if input_convs is not None and in_chans != 3: if isinstance(input_convs, str): input_convs = (input_convs,) for input_conv_name in input_convs: weight_name = input_conv_name + '.weight' try: state_dict[weight_name] = adapt_input_conv(in_chans, state_dict[weight_name]) _logger.info( f'Converted input conv {input_conv_name} pretrained weights from 3 to {in_chans} channel(s)') except NotImplementedError as e: del state_dict[weight_name] strict = False _logger.warning( f'Unable to convert pretrained {input_conv_name} weights, using random init for this layer.') classifiers = default_cfg.get('classifier', None) label_offset = default_cfg.get('label_offset', 0) if classifiers is not None: if isinstance(classifiers, str): classifiers = (classifiers,) if num_classes != default_cfg['num_classes']: for classifier_name in classifiers: # completely discard fully connected if model num_classes doesn't match pretrained weights del state_dict[classifier_name + '.weight'] del state_dict[classifier_name + '.bias'] strict = False elif label_offset > 0: for classifier_name in classifiers: # special case for pretrained weights with an extra background class in pretrained weights classifier_weight = state_dict[classifier_name + '.weight'] state_dict[classifier_name + '.weight'] = classifier_weight[label_offset:] classifier_bias = state_dict[classifier_name + '.bias'] state_dict[classifier_name + '.bias'] = classifier_bias[label_offset:] model.load_state_dict(state_dict, strict=strict) def extract_layer(model, layer): layer = layer.split('.') module = model if hasattr(model, 'module') and layer[0] != 'module': module = model.module if not hasattr(model, 'module') and layer[0] == 'module': layer = layer[1:] for l in layer: if hasattr(module, l): if not l.isdigit(): module = getattr(module, l) else: module = module[int(l)] else: return module return module def set_layer(model, layer, val): layer = layer.split('.') module = model if hasattr(model, 'module') and layer[0] != 'module': module = model.module lst_index = 0 module2 = module for l in layer: if hasattr(module2, l): if not l.isdigit(): module2 = getattr(module2, l) else: module2 = module2[int(l)] lst_index += 1 lst_index -= 1 for l in layer[:lst_index]: if not l.isdigit(): module = getattr(module, l) else: module = module[int(l)] l = layer[lst_index] setattr(module, l, val) def adapt_model_from_string(parent_module, model_string): separator = '***' state_dict = {} lst_shape = model_string.split(separator) for k in lst_shape: k = k.split(':') key = k[0] shape = k[1][1:-1].split(',') if shape[0] != '': state_dict[key] = [int(i) for i in shape] new_module = deepcopy(parent_module) for n, m in parent_module.named_modules(): old_module = extract_layer(parent_module, n) if isinstance(old_module, nn.Conv2d) or isinstance(old_module, Conv2dSame): if isinstance(old_module, Conv2dSame): conv = Conv2dSame else: conv = nn.Conv2d s = state_dict[n + '.weight'] in_channels = s[1] out_channels = s[0] g = 1 if old_module.groups > 1: in_channels = out_channels g = in_channels new_conv = conv( in_channels=in_channels, out_channels=out_channels, kernel_size=old_module.kernel_size, bias=old_module.bias is not None, padding=old_module.padding, dilation=old_module.dilation, groups=g, stride=old_module.stride) set_layer(new_module, n, new_conv) if isinstance(old_module, nn.BatchNorm2d): new_bn = nn.BatchNorm2d( num_features=state_dict[n + '.weight'][0], eps=old_module.eps, momentum=old_module.momentum, affine=old_module.affine, track_running_stats=True) set_layer(new_module, n, new_bn) if isinstance(old_module, nn.Linear): # FIXME extra checks to ensure this is actually the FC classifier layer and not a diff Linear layer? num_features = state_dict[n + '.weight'][1] new_fc = Linear( in_features=num_features, out_features=old_module.out_features, bias=old_module.bias is not None) set_layer(new_module, n, new_fc) if hasattr(new_module, 'num_features'): new_module.num_features = num_features new_module.eval() parent_module.eval() return new_module def adapt_model_from_file(parent_module, model_variant): adapt_file = os.path.join(os.path.dirname(__file__), 'pruned', model_variant + '.txt') with open(adapt_file, 'r') as f: return adapt_model_from_string(parent_module, f.read().strip()) def default_cfg_for_features(default_cfg): default_cfg = deepcopy(default_cfg) # remove default pretrained cfg fields that don't have much relevance for feature backbone to_remove = ('num_classes', 'crop_pct', 'classifier', 'global_pool') # add default final pool size? for tr in to_remove: default_cfg.pop(tr, None) return default_cfg def overlay_external_default_cfg(default_cfg, kwargs): """ Overlay 'external_default_cfg' in kwargs on top of default_cfg arg. """ external_default_cfg = kwargs.pop('external_default_cfg', None) if external_default_cfg: default_cfg.pop('url', None) # url should come from external cfg default_cfg.pop('hf_hub', None) # hf hub id should come from external cfg default_cfg.update(external_default_cfg) def set_default_kwargs(kwargs, names, default_cfg): for n in names: # for legacy reasons, model __init__args uses img_size + in_chans as separate args while # default_cfg has one input_size=(C, H ,W) entry if n == 'img_size': input_size = default_cfg.get('input_size', None) if input_size is not None: assert len(input_size) == 3 kwargs.setdefault(n, input_size[-2:]) elif n == 'in_chans': input_size = default_cfg.get('input_size', None) if input_size is not None: assert len(input_size) == 3 kwargs.setdefault(n, input_size[0]) else: default_val = default_cfg.get(n, None) if default_val is not None: kwargs.setdefault(n, default_cfg[n]) def filter_kwargs(kwargs, names): if not kwargs or not names: return for n in names: kwargs.pop(n, None) def update_default_cfg_and_kwargs(default_cfg, kwargs, kwargs_filter): """ Update the default_cfg and kwargs before passing to model FIXME this sequence of overlay default_cfg, set default kwargs, filter kwargs could/should be replaced by an improved configuration mechanism Args: default_cfg: input default_cfg (updated in-place) kwargs: keyword args passed to model build fn (updated in-place) kwargs_filter: keyword arg keys that must be removed before model __init__ """ # Overlay default cfg values from `external_default_cfg` if it exists in kwargs overlay_external_default_cfg(default_cfg, kwargs) # Set model __init__ args that can be determined by default_cfg (if not already passed as kwargs) default_kwarg_names = ('num_classes', 'global_pool', 'in_chans') if default_cfg.get('fixed_input_size', False): # if fixed_input_size exists and is True, model takes an img_size arg that fixes its input size default_kwarg_names += ('img_size',) set_default_kwargs(kwargs, names=default_kwarg_names, default_cfg=default_cfg) # Filter keyword args for task specific model variants (some 'features only' models, etc.) filter_kwargs(kwargs, names=kwargs_filter) def swin_build_model_with_cfg( model_cls: Callable, variant: str, pretrained: bool, default_cfg: dict, model_cfg: Optional[Any] = None, feature_cfg: Optional[dict] = None, pretrained_strict: bool = True, pretrained_filter_fn: Optional[Callable] = None, pretrained_custom_load: bool = False, kwargs_filter: Optional[Tuple[str]] = None, **kwargs): """ Build model with specified default_cfg and optional model_cfg This helper fn aids in the construction of a model including: * handling default_cfg and associated pretained weight loading * passing through optional model_cfg for models with config based arch spec * features_only model adaptation * pruning config / model adaptation Args: model_cls (nn.Module): model class variant (str): model variant name pretrained (bool): load pretrained weights default_cfg (dict): model's default pretrained/task config model_cfg (Optional[Dict]): model's architecture config feature_cfg (Optional[Dict]: feature extraction adapter config pretrained_strict (bool): load pretrained weights strictly pretrained_filter_fn (Optional[Callable]): filter callable for pretrained weights pretrained_custom_load (bool): use custom load fn, to load numpy or other non PyTorch weights kwargs_filter (Optional[Tuple]): kwargs to filter before passing to model **kwargs: model args passed through to model __init__ """ pruned = kwargs.pop('pruned', False) features = False feature_cfg = feature_cfg or {} default_cfg = deepcopy(default_cfg) if default_cfg else {} update_default_cfg_and_kwargs(default_cfg, kwargs, kwargs_filter) default_cfg.setdefault('architecture', variant) # Setup for feature extraction wrapper done at end of this fn if kwargs.pop('features_only', False): features = True feature_cfg.setdefault('out_indices', (0, 1, 2, 3, 4)) if 'out_indices' in kwargs: feature_cfg['out_indices'] = kwargs.pop('out_indices') # Build the model model = model_cls(**kwargs) if model_cfg is None else model_cls(cfg=model_cfg, **kwargs) model.default_cfg = default_cfg if pruned: model = adapt_model_from_file(model, variant) # For classification models, check class attr, then kwargs, then default to 1k, otherwise 0 for feats num_classes_pretrained = 0 if features else getattr(model, 'num_classes', kwargs.get('num_classes', 1000)) if pretrained: if pretrained_custom_load: load_custom_pretrained(model) else: load_pretrained( model, num_classes=num_classes_pretrained, in_chans=kwargs.get('in_chans', 3), filter_fn=pretrained_filter_fn, img_size=kwargs['img_size'], strict=pretrained_strict, resolution_before=kwargs['config']['resolution_before']) # Wrap the model in a feature extraction module if enabled if features: feature_cls = FeatureListNet if 'feature_cls' in feature_cfg: feature_cls = feature_cfg.pop('feature_cls') if isinstance(feature_cls, str): feature_cls = feature_cls.lower() if 'hook' in feature_cls: feature_cls = FeatureHookNet else: assert False, f'Unknown feature class {feature_cls}' model = feature_cls(model, **feature_cfg) model.default_cfg = default_cfg_for_features(default_cfg) # add back default_cfg return model def model_parameters(model, exclude_head=False): if exclude_head: # FIXME this a bit of a quick and dirty hack to skip classifier head params based on ordering return [p for p in model.parameters()][:-2] else: return model.parameters() def named_apply(fn: Callable, module: nn.Module, name='', depth_first=True, include_root=False) -> nn.Module: if not depth_first and include_root: fn(module=module, name=name) for child_name, child_module in module.named_children(): child_name = '.'.join((name, child_name)) if name else child_name named_apply(fn=fn, module=child_module, name=child_name, depth_first=depth_first, include_root=True) if depth_first and include_root: fn(module=module, name=name) return module def named_modules(module: nn.Module, name='', depth_first=True, include_root=False): if not depth_first and include_root: yield name, module for child_name, child_module in module.named_children(): child_name = '.'.join((name, child_name)) if name else child_name yield from named_modules( module=child_module, name=child_name, depth_first=depth_first, include_root=True) if depth_first and include_root: yield name, module
23,550
43.519849
153
py
Mr.Right
Mr.Right-main/models/METER/meter/modules/heads.py
import torch import torch.nn as nn import torch.nn.functional as F from transformers.models.bert.modeling_bert import BertPredictionHeadTransform class Pooler(nn.Module): def __init__(self, hidden_size): super().__init__() self.dense = nn.Linear(hidden_size, hidden_size) self.activation = nn.Tanh() def forward(self, hidden_states): first_token_tensor = hidden_states[:, 0] pooled_output = self.dense(first_token_tensor) pooled_output = self.activation(pooled_output) return pooled_output class ITMHead(nn.Module): def __init__(self, hidden_size): super().__init__() self.fc = nn.Linear(hidden_size, 2) def forward(self, x): x = self.fc(x) return x class MLMHead(nn.Module): def __init__(self, config, weight=None): super().__init__() self.transform = BertPredictionHeadTransform(config) self.decoder = nn.Linear(config.hidden_size, config.vocab_size, bias=False) self.bias = nn.Parameter(torch.zeros(config.vocab_size)) if weight is not None: self.decoder.weight = weight def forward(self, x): x = self.transform(x) x = self.decoder(x) + self.bias return x
1,257
27.590909
83
py
Mr.Right
Mr.Right-main/models/METER/meter/modules/__init__.py
from .meter_module import METERTransformerSS
45
22
44
py
Mr.Right
Mr.Right-main/models/METER/meter/utils/write_vg.py
import json import pandas as pd import pyarrow as pa import random import os from tqdm import tqdm from glob import glob from collections import defaultdict def path2rest(path, iid2captions): name = path.split("/")[-1] iid = int(name[:-4]) with open(path, "rb") as fp: binary = fp.read() cdicts = iid2captions[iid] captions = [c["phrase"] for c in cdicts] widths = [c["width"] for c in cdicts] heights = [c["height"] for c in cdicts] xs = [c["x"] for c in cdicts] ys = [c["y"] for c in cdicts] return [ binary, captions, widths, heights, xs, ys, str(iid), ] def make_arrow(root, dataset_root): with open(f"{root}/annotations/region_descriptions.json", "r") as fp: captions = json.load(fp) iid2captions = defaultdict(list) for cap in tqdm(captions): cap = cap["regions"] for c in cap: iid2captions[c["image_id"]].append(c) paths = list(glob(f"{root}/images/VG_100K/*.jpg")) + list( glob(f"{root}/images/VG_100K_2/*.jpg") ) random.shuffle(paths) caption_paths = [ path for path in paths if int(path.split("/")[-1][:-4]) in iid2captions ] if len(paths) == len(caption_paths): print("all images have caption annotations") else: print("not all images have caption annotations") print( len(paths), len(caption_paths), len(iid2captions), ) bs = [path2rest(path, iid2captions) for path in tqdm(caption_paths)] dataframe = pd.DataFrame( bs, columns=["image", "caption", "width", "height", "x", "y", "image_id"], ) table = pa.Table.from_pandas(dataframe) os.makedirs(dataset_root, exist_ok=True) with pa.OSFile(f"{dataset_root}/vg.arrow", "wb") as sink: with pa.RecordBatchFileWriter(sink, table.schema) as writer: writer.write_table(table)
1,928
25.424658
82
py
Mr.Right
Mr.Right-main/models/METER/meter/utils/glossary.py
import re contractions = { "aint": "ain't", "arent": "aren't", "cant": "can't", "couldve": "could've", "couldnt": "couldn't", "couldn'tve": "couldn't've", "couldnt've": "couldn't've", "didnt": "didn't", "doesnt": "doesn't", "dont": "don't", "hadnt": "hadn't", "hadnt've": "hadn't've", "hadn'tve": "hadn't've", "hasnt": "hasn't", "havent": "haven't", "hed": "he'd", "hed've": "he'd've", "he'dve": "he'd've", "hes": "he's", "howd": "how'd", "howll": "how'll", "hows": "how's", "Id've": "I'd've", "I'dve": "I'd've", "Im": "I'm", "Ive": "I've", "isnt": "isn't", "itd": "it'd", "itd've": "it'd've", "it'dve": "it'd've", "itll": "it'll", "let's": "let's", "maam": "ma'am", "mightnt": "mightn't", "mightnt've": "mightn't've", "mightn'tve": "mightn't've", "mightve": "might've", "mustnt": "mustn't", "mustve": "must've", "neednt": "needn't", "notve": "not've", "oclock": "o'clock", "oughtnt": "oughtn't", "ow's'at": "'ow's'at", "'ows'at": "'ow's'at", "'ow'sat": "'ow's'at", "shant": "shan't", "shed've": "she'd've", "she'dve": "she'd've", "she's": "she's", "shouldve": "should've", "shouldnt": "shouldn't", "shouldnt've": "shouldn't've", "shouldn'tve": "shouldn't've", "somebody'd": "somebodyd", "somebodyd've": "somebody'd've", "somebody'dve": "somebody'd've", "somebodyll": "somebody'll", "somebodys": "somebody's", "someoned": "someone'd", "someoned've": "someone'd've", "someone'dve": "someone'd've", "someonell": "someone'll", "someones": "someone's", "somethingd": "something'd", "somethingd've": "something'd've", "something'dve": "something'd've", "somethingll": "something'll", "thats": "that's", "thered": "there'd", "thered've": "there'd've", "there'dve": "there'd've", "therere": "there're", "theres": "there's", "theyd": "they'd", "theyd've": "they'd've", "they'dve": "they'd've", "theyll": "they'll", "theyre": "they're", "theyve": "they've", "twas": "'twas", "wasnt": "wasn't", "wed've": "we'd've", "we'dve": "we'd've", "weve": "we've", "werent": "weren't", "whatll": "what'll", "whatre": "what're", "whats": "what's", "whatve": "what've", "whens": "when's", "whered": "where'd", "wheres": "where's", "whereve": "where've", "whod": "who'd", "whod've": "who'd've", "who'dve": "who'd've", "wholl": "who'll", "whos": "who's", "whove": "who've", "whyll": "why'll", "whyre": "why're", "whys": "why's", "wont": "won't", "wouldve": "would've", "wouldnt": "wouldn't", "wouldnt've": "wouldn't've", "wouldn'tve": "wouldn't've", "yall": "y'all", "yall'll": "y'all'll", "y'allll": "y'all'll", "yall'd've": "y'all'd've", "y'alld've": "y'all'd've", "y'all'dve": "y'all'd've", "youd": "you'd", "youd've": "you'd've", "you'dve": "you'd've", "youll": "you'll", "youre": "you're", "youve": "you've", } manual_map = { "none": "0", "zero": "0", "one": "1", "two": "2", "three": "3", "four": "4", "five": "5", "six": "6", "seven": "7", "eight": "8", "nine": "9", "ten": "10", } articles = ["a", "an", "the"] period_strip = re.compile("(?!<=\d)(\.)(?!\d)") comma_strip = re.compile("(\d)(\,)(\d)") punct = [ ";", r"/", "[", "]", '"', "{", "}", "(", ")", "=", "+", "\\", "_", "-", ">", "<", "@", "`", ",", "?", "!", ] def normalize_word(token): _token = token for p in punct: if (p + " " in token or " " + p in token) or ( re.search(comma_strip, token) != None ): _token = _token.replace(p, "") else: _token = _token.replace(p, " ") token = period_strip.sub("", _token, re.UNICODE) _token = [] temp = token.lower().split() for word in temp: word = manual_map.setdefault(word, word) if word not in articles: _token.append(word) for i, word in enumerate(_token): if word in contractions: _token[i] = contractions[word] token = " ".join(_token) token = token.replace(",", "") return token
4,435
22.225131
54
py
Mr.Right
Mr.Right-main/models/METER/meter/utils/write_vqa.py
import json import pandas as pd import pyarrow as pa import random import os from tqdm import tqdm from glob import glob from collections import defaultdict, Counter from .glossary import normalize_word def get_score(occurences): if occurences == 0: return 0.0 elif occurences == 1: return 0.3 elif occurences == 2: return 0.6 elif occurences == 3: return 0.9 else: return 1.0 def path2rest(path, split, annotations, label2ans): iid = int(path.split("/")[-1].split("_")[-1][:-4]) with open(path, "rb") as fp: binary = fp.read() _annot = annotations[split][iid] _annot = list(_annot.items()) qids, qas = [a[0] for a in _annot], [a[1] for a in _annot] questions = [qa[0] for qa in qas] answers = [qa[1] for qa in qas] if "test" not in split else list(list()) answer_labels = ( [a["labels"] for a in answers] if "test" not in split else list(list()) ) answer_scores = ( [a["scores"] for a in answers] if "test" not in split else list(list()) ) answers = ( [[label2ans[l] for l in al] for al in answer_labels] if "test" not in split else list(list()) ) return [binary, questions, answers, answer_labels, answer_scores, iid, qids, split] def make_arrow(root, dataset_root): with open(f"{root}/v2_OpenEnded_mscoco_train2014_questions.json", "r") as fp: questions_train2014 = json.load(fp)["questions"] with open(f"{root}/v2_OpenEnded_mscoco_val2014_questions.json", "r") as fp: questions_val2014 = json.load(fp)["questions"] with open(f"{root}/v2_OpenEnded_mscoco_test2015_questions.json", "r") as fp: questions_test2015 = json.load(fp)["questions"] with open(f"{root}/v2_OpenEnded_mscoco_test-dev2015_questions.json", "r") as fp: questions_test_dev2015 = json.load(fp)["questions"] with open(f"{root}/v2_mscoco_train2014_annotations.json", "r") as fp: annotations_train2014 = json.load(fp)["annotations"] with open(f"{root}/v2_mscoco_val2014_annotations.json", "r") as fp: annotations_val2014 = json.load(fp)["annotations"] annotations = dict() for split, questions in zip( ["train", "val", "test", "test-dev"], [ questions_train2014, questions_val2014, questions_test2015, questions_test_dev2015, ], ): _annot = defaultdict(dict) for q in tqdm(questions): _annot[q["image_id"]][q["question_id"]] = [q["question"]] annotations[split] = _annot all_major_answers = list() for split, annots in zip( ["train", "val"], [annotations_train2014, annotations_val2014], ): _annot = annotations[split] for q in tqdm(annots): all_major_answers.append(q["multiple_choice_answer"]) all_major_answers = [normalize_word(word) for word in tqdm(all_major_answers)] counter = {k: v for k, v in Counter(all_major_answers).items() if v >= 9} ans2label = {k: i for i, k in enumerate(counter.keys())} label2ans = list(counter.keys()) for split, annots in zip( ["train", "val"], [annotations_train2014, annotations_val2014], ): _annot = annotations[split] for q in tqdm(annots): answers = q["answers"] answer_count = {} for answer in answers: answer_ = answer["answer"] answer_count[answer_] = answer_count.get(answer_, 0) + 1 labels = [] scores = [] for answer in answer_count: if answer not in ans2label: continue labels.append(ans2label[answer]) score = get_score(answer_count[answer]) scores.append(score) _annot[q["image_id"]][q["question_id"]].append( {"labels": labels, "scores": scores,} ) for split in ["train", "val"]: filtered_annot = dict() for ik, iv in annotations[split].items(): new_q = dict() for qk, qv in iv.items(): if len(qv[1]["labels"]) != 0: new_q[qk] = qv if len(new_q) != 0: filtered_annot[ik] = new_q annotations[split] = filtered_annot for split in [ "train", "val", "test", "test-dev", ]: annot = annotations[split] split_name = { "train": "train2014", "val": "val2014", "test": "test2015", "test-dev": "test2015", }[split] paths = list(glob(f"{root}/{split_name}/*.jpg")) random.shuffle(paths) annot_paths = [ path for path in paths if int(path.split("/")[-1].split("_")[-1][:-4]) in annot ] if len(paths) == len(annot_paths): print("all images have caption annotations") else: print("not all images have caption annotations") print( len(paths), len(annot_paths), len(annot), ) bs = [ path2rest(path, split, annotations, label2ans) for path in tqdm(annot_paths) ] dataframe = pd.DataFrame( bs, columns=[ "image", "questions", "answers", "answer_labels", "answer_scores", "image_id", "question_id", "split", ], ) table = pa.Table.from_pandas(dataframe) os.makedirs(dataset_root, exist_ok=True) with pa.OSFile(f"{dataset_root}/vqav2_{split}.arrow", "wb") as sink: with pa.RecordBatchFileWriter(sink, table.schema) as writer: writer.write_table(table) table = pa.ipc.RecordBatchFileReader( pa.memory_map(f"{dataset_root}/vqav2_val.arrow", "r") ).read_all() pdtable = table.to_pandas() df1 = pdtable[:-1000] df2 = pdtable[-1000:] df1 = pa.Table.from_pandas(df1) df2 = pa.Table.from_pandas(df2) with pa.OSFile(f"{dataset_root}/vqav2_trainable_val.arrow", "wb") as sink: with pa.RecordBatchFileWriter(sink, df1.schema) as writer: writer.write_table(df1) with pa.OSFile(f"{dataset_root}/vqav2_rest_val.arrow", "wb") as sink: with pa.RecordBatchFileWriter(sink, df2.schema) as writer: writer.write_table(df2)
6,523
30.669903
88
py
Mr.Right
Mr.Right-main/models/METER/meter/utils/write_conceptual_caption.py
import json import pandas as pd import pyarrow as pa import gc import random import os from tqdm import tqdm from glob import glob def path2rest(path, iid2captions): split, _, name = path.split("/")[-3:] split = split.split("_")[-1] iid = name with open(path, "rb") as fp: binary = fp.read() captions = iid2captions[iid] return [ binary, captions, iid, split, ] def make_arrow(root, dataset_root): for split in ["val", "train"]: with open(f"{root}/{split}_annot.json", "r") as fp: captions = json.load(fp) iid2captions = dict() for cap in tqdm(captions): iid = cap[0].split("/")[-1] iid2captions[iid] = [cap[1]] paths = list(glob(f"{root}/images_{split}/*/*")) random.shuffle(paths) caption_paths = [path for path in paths if path.split("/")[-1] in iid2captions] if len(paths) == len(caption_paths): print("all images have caption annotations") else: print("not all images have caption annotations") print( len(paths), len(caption_paths), len(iid2captions), ) sub_len = int(len(caption_paths) // 100000) subs = list(range(sub_len + 1)) for sub in subs: sub_paths = caption_paths[sub * 100000 : (sub + 1) * 100000] bs = [path2rest(path, iid2captions) for path in tqdm(sub_paths)] dataframe = pd.DataFrame( bs, columns=["image", "caption", "image_id", "split"], ) table = pa.Table.from_pandas(dataframe) os.makedirs(dataset_root, exist_ok=True) with pa.OSFile( f"{dataset_root}/conceptual_caption_{split}_{sub}.arrow", "wb" ) as sink: with pa.RecordBatchFileWriter(sink, table.schema) as writer: writer.write_table(table) del dataframe del table del bs gc.collect()
2,037
27.305556
87
py
Mr.Right
Mr.Right-main/models/METER/meter/utils/write_nlvr2.py
import json import pandas as pd import pyarrow as pa import os from tqdm import tqdm from collections import defaultdict def process(root, iden, row): texts = [r["sentence"] for r in row] labels = [r["label"] for r in row] split = iden.split("-")[0] if iden.startswith("train"): directory = row[0]["directory"] path = f"{root}/images/train/{directory}/{iden}" else: path = f"{root}/{split}/{iden}" with open(f"{path}-img0.png", "rb") as fp: img0 = fp.read() with open(f"{path}-img1.png", "rb") as fp: img1 = fp.read() return [img0, img1, texts, labels, iden] def make_arrow(root, dataset_root): train_data = list( map(json.loads, open(f"{root}/nlvr2/data/train.json").readlines()) ) test1_data = list( map(json.loads, open(f"{root}/nlvr2/data/test1.json").readlines()) ) dev_data = list(map(json.loads, open(f"{root}/nlvr2/data/dev.json").readlines())) balanced_test1_data = list( map( json.loads, open(f"{root}/nlvr2/data/balanced/balanced_test1.json").readlines(), ) ) balanced_dev_data = list( map( json.loads, open(f"{root}/nlvr2/data/balanced/balanced_dev.json").readlines(), ) ) unbalanced_test1_data = list( map( json.loads, open(f"{root}/nlvr2/data/unbalanced/unbalanced_test1.json").readlines(), ) ) unbalanced_dev_data = list( map( json.loads, open(f"{root}/nlvr2/data/unbalanced/unbalanced_dev.json").readlines(), ) ) splits = [ "train", "dev", "test1", "balanced_dev", "balanced_test1", "unbalanced_dev", "unbalanced_test1", ] datas = [ train_data, dev_data, test1_data, balanced_dev_data, balanced_test1_data, unbalanced_dev_data, unbalanced_test1_data, ] annotations = dict() for split, data in zip(splits, datas): _annot = defaultdict(list) for row in tqdm(data): _annot["-".join(row["identifier"].split("-")[:-1])].append(row) annotations[split] = _annot for split in splits: bs = [ process(root, iden, row) for iden, row in tqdm(annotations[split].items()) ] dataframe = pd.DataFrame( bs, columns=["image_0", "image_1", "questions", "answers", "identifier"], ) table = pa.Table.from_pandas(dataframe) os.makedirs(dataset_root, exist_ok=True) with pa.OSFile(f"{dataset_root}/nlvr2_{split}.arrow", "wb") as sink: with pa.RecordBatchFileWriter(sink, table.schema) as writer: writer.write_table(table)
2,818
25.101852
86
py
Mr.Right
Mr.Right-main/models/METER/meter/utils/write_sbu.py
import json import pandas as pd import pyarrow as pa import gc import random import os from tqdm import tqdm from glob import glob def path2rest(path, iid2captions): split, _, name = path.split("/")[-3:] split = split.split("_")[-1] iid = name with open(path, "rb") as fp: binary = fp.read() captions = iid2captions[iid] return [ binary, captions, iid, split, ] def make_arrow(root, dataset_root): with open(f"{root}/annot.json", "r") as fp: captions = json.load(fp) iid2captions = dict() for cap in tqdm(captions): iid = cap[0].split("/")[-1] iid2captions[iid] = [cap[1]] paths = list(glob(f"{root}/images_train/*/*")) random.shuffle(paths) caption_paths = [path for path in paths if path.split("/")[-1] in iid2captions] if len(paths) == len(caption_paths): print("all images have caption annotations") else: print("not all images have caption annotations") print( len(paths), len(caption_paths), len(iid2captions), ) sub_len = int(len(caption_paths) // 100000) subs = list(range(sub_len + 1)) for sub in subs: sub_paths = caption_paths[sub * 100000 : (sub + 1) * 100000] bs = [path2rest(path, iid2captions) for path in tqdm(sub_paths)] dataframe = pd.DataFrame(bs, columns=["image", "caption", "image_id", "split"],) table = pa.Table.from_pandas(dataframe) os.makedirs(dataset_root, exist_ok=True) with pa.OSFile(f"{dataset_root}/sbu_{sub}.arrow", "wb") as sink: with pa.RecordBatchFileWriter(sink, table.schema) as writer: writer.write_table(table) del dataframe del table del bs gc.collect()
1,785
25.656716
88
py
Mr.Right
Mr.Right-main/models/METER/meter/utils/write_f30k_karpathy.py
import json import pandas as pd import pyarrow as pa import random import os from tqdm import tqdm from glob import glob from collections import defaultdict def path2rest(path, iid2captions, iid2split): name = path.split("/")[-1] with open(path, "rb") as fp: binary = fp.read() captions = iid2captions[name] split = iid2split[name] return [binary, captions, name, split] def make_arrow(root, dataset_root): with open(f"{root}/karpathy/dataset_flickr30k.json", "r") as fp: captions = json.load(fp) captions = captions["images"] iid2captions = defaultdict(list) iid2split = dict() for cap in tqdm(captions): filename = cap["filename"] iid2split[filename] = cap["split"] for c in cap["sentences"]: iid2captions[filename].append(c["raw"]) paths = list(glob(f"{root}/flickr30k-images/*.jpg")) random.shuffle(paths) caption_paths = [path for path in paths if path.split("/")[-1] in iid2captions] if len(paths) == len(caption_paths): print("all images have caption annotations") else: print("not all images have caption annotations") print( len(paths), len(caption_paths), len(iid2captions), ) bs = [path2rest(path, iid2captions, iid2split) for path in tqdm(caption_paths)] for split in ["train", "val", "test"]: batches = [b for b in bs if b[-1] == split] dataframe = pd.DataFrame( batches, columns=["image", "caption", "image_id", "split"], ) table = pa.Table.from_pandas(dataframe) os.makedirs(dataset_root, exist_ok=True) with pa.OSFile( f"{dataset_root}/f30k_caption_karpathy_{split}.arrow", "wb" ) as sink: with pa.RecordBatchFileWriter(sink, table.schema) as writer: writer.write_table(table)
1,871
26.529412
83
py
Mr.Right
Mr.Right-main/models/METER/meter/utils/write_snli.py
import json import pandas as pd import pyarrow as pa import os from tqdm import tqdm from collections import defaultdict label2id = {'contradiction': 0, 'neutral': 1, 'entailment': 2} def process(root, imgid, ann): with open(f"{root}/Flickr30K/images/{imgid}.jpg", "rb") as fp: img = fp.read() sentences = ann['sentences'] labels = ann['labels'] return [img, sentences, labels] def make_arrow(root, dataset_root): train_data = list( map(json.loads, open(f"{root}/snli_ve_train.jsonl").readlines()) ) test_data = list( map(json.loads, open(f"{root}/snli_ve_test.jsonl").readlines()) ) dev_data = list( map(json.loads, open(f"{root}/snli_ve_dev.jsonl").readlines()) ) splits = [ "train", "dev", "test", ] annotations = dict() annotations['train'] = train_data annotations['dev'] = dev_data annotations['test'] = test_data annots = dict() for split in splits: annots[split] = {} for line in annotations[split]: imgid = line['Flickr30K_ID'] if not imgid in annots[split]: annots[split][imgid] = {} annots[split][imgid]['sentences'] = [] annots[split][imgid]['labels'] = [] annots[split][imgid]['sentences'].append( [line['sentence1'], line['sentence2']] ) annots[split][imgid]['labels'].append( label2id[line['gold_label']] ) for split in splits: bs = [process(root, imgid, annots[split][imgid]) for imgid in tqdm(annots[split])] dataframe = pd.DataFrame( bs, columns=["image", "sentences", "labels"] ) table = pa.Table.from_pandas(dataframe) os.makedirs(dataset_root, exist_ok=True) with pa.OSFile(f"{dataset_root}/snli_{split}.arrow", "wb") as sink: with pa.RecordBatchFileWriter(sink, table.schema) as writer: writer.write_table(table)
2,006
26.493151
94
py
Mr.Right
Mr.Right-main/models/METER/meter/utils/__init__.py
0
0
0
py
Mr.Right
Mr.Right-main/models/METER/meter/utils/write_coco_karpathy.py
import json import os import pandas as pd import pyarrow as pa import random from tqdm import tqdm from glob import glob from collections import defaultdict def path2rest(path, iid2captions, iid2split): name = path.split("/")[-1] with open(path, "rb") as fp: binary = fp.read() captions = iid2captions[name] split = iid2split[name] return [binary, captions, name, split] def make_arrow(root, dataset_root): with open(f"{root}/karpathy/dataset_coco.json", "r") as fp: captions = json.load(fp) captions = captions["images"] iid2captions = defaultdict(list) iid2split = dict() for cap in tqdm(captions): filename = cap["filename"] iid2split[filename] = cap["split"] for c in cap["sentences"]: iid2captions[filename].append(c["raw"]) paths = list(glob(f"{root}/train2014/*.jpg")) + list(glob(f"{root}/val2014/*.jpg")) random.shuffle(paths) caption_paths = [path for path in paths if path.split("/")[-1] in iid2captions] if len(paths) == len(caption_paths): print("all images have caption annotations") else: print("not all images have caption annotations") print( len(paths), len(caption_paths), len(iid2captions), ) bs = [path2rest(path, iid2captions, iid2split) for path in tqdm(caption_paths)] for split in ["train", "val", "restval", "test"]: batches = [b for b in bs if b[-1] == split] dataframe = pd.DataFrame( batches, columns=["image", "caption", "image_id", "split"], ) table = pa.Table.from_pandas(dataframe) os.makedirs(dataset_root, exist_ok=True) with pa.OSFile( f"{dataset_root}/coco_caption_karpathy_{split}.arrow", "wb" ) as sink: with pa.RecordBatchFileWriter(sink, table.schema) as writer: writer.write_table(table)
1,904
28.765625
87
py
Mr.Right
Mr.Right-main/models/METER/meter/transforms/transform.py
from .utils import ( inception_normalize, imagenet_normalize, MinMaxResize, ) from PIL import Image from torchvision import transforms from torchvision.transforms import Compose, Resize, CenterCrop, ToTensor, Normalize from .randaug import RandAugment def pixelbert_transform(size=800): longer = int((1333 / 800) * size) return transforms.Compose( [ MinMaxResize(shorter=size, longer=longer), transforms.ToTensor(), inception_normalize, ] ) def pixelbert_transform_randaug(size=800): longer = int((1333 / 800) * size) trs = transforms.Compose( [ MinMaxResize(shorter=size, longer=longer), transforms.ToTensor(), inception_normalize, ] ) trs.transforms.insert(0, RandAugment(2, 9)) return trs def imagenet_transform(size=800): return transforms.Compose( [ Resize(size, interpolation=Image.BICUBIC), CenterCrop(size), transforms.ToTensor(), imagenet_normalize, ] ) def imagenet_transform_randaug(size=800): trs = transforms.Compose( [ Resize(size, interpolation=Image.BICUBIC), CenterCrop(size), transforms.ToTensor(), imagenet_normalize, ] ) trs.transforms.insert(0, RandAugment(2, 9)) return trs def vit_transform(size=800): return transforms.Compose( [ Resize(size, interpolation=Image.BICUBIC), CenterCrop(size), transforms.ToTensor(), inception_normalize, ] ) def vit_transform_randaug(size=800): trs = transforms.Compose( [ Resize(size, interpolation=Image.BICUBIC), CenterCrop(size), transforms.ToTensor(), inception_normalize, ] ) trs.transforms.insert(0, RandAugment(2, 9)) return trs def clip_transform(size): return Compose([ Resize(size, interpolation=Image.BICUBIC), CenterCrop(size), lambda image: image.convert("RGB"), ToTensor(), Normalize((0.48145466, 0.4578275, 0.40821073), (0.26862954, 0.26130258, 0.27577711)), ]) def clip_transform_randaug(size): trs = Compose([ Resize(size, interpolation=Image.BICUBIC), CenterCrop(size), lambda image: image.convert("RGB"), ToTensor(), Normalize((0.48145466, 0.4578275, 0.40821073), (0.26862954, 0.26130258, 0.27577711)), ]) trs.transforms.insert(0, lambda image: image.convert('RGBA')) trs.transforms.insert(0, RandAugment(2, 9)) trs.transforms.insert(0, lambda image: image.convert('RGB')) return trs
2,733
26.34
93
py
Mr.Right
Mr.Right-main/models/METER/meter/transforms/utils.py
from torchvision import transforms from PIL import Image class MinMaxResize: def __init__(self, shorter=800, longer=1333): self.min = shorter self.max = longer def __call__(self, x): w, h = x.size scale = self.min / min(w, h) if h < w: newh, neww = self.min, scale * w else: newh, neww = scale * h, self.min if max(newh, neww) > self.max: scale = self.max / max(newh, neww) newh = newh * scale neww = neww * scale newh, neww = int(newh + 0.5), int(neww + 0.5) newh, neww = newh // 32 * 32, neww // 32 * 32 return x.resize((neww, newh), resample=Image.BICUBIC) class UnNormalize(object): def __init__(self, mean, std): self.mean = mean self.std = std def __call__(self, tensor): """ Args: tensor (Tensor): Tensor image of size (C, H, W) to be normalized. Returns: Tensor: Normalized image. """ for t, m, s in zip(tensor, self.mean, self.std): t.mul_(s).add_(m) # The normalize code -> t.sub_(m).div_(s) return tensor # This is simple maximum entropy normalization performed in Inception paper inception_normalize = transforms.Compose( [transforms.Normalize(mean=[0.5, 0.5, 0.5], std=[0.5, 0.5, 0.5])] ) # ViT uses simple non-biased inception normalization # https://github.com/google-research/vision_transformer/blob/master/vit_jax/input_pipeline.py#L132 inception_unnormalize = transforms.Compose( [UnNormalize(mean=[0.5, 0.5, 0.5], std=[0.5, 0.5, 0.5])] ) # ImageNet normalize imagenet_normalize = transforms.Compose( [transforms.Normalize(mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225])] )
1,792
27.919355
98
py
Mr.Right
Mr.Right-main/models/METER/meter/transforms/randaug.py
# code in this file is adpated from rpmcruz/autoaugment # https://github.com/rpmcruz/autoaugment/blob/master/transformations.py import random import PIL, PIL.ImageOps, PIL.ImageEnhance, PIL.ImageDraw import numpy as np import torch from PIL import Image def ShearX(img, v): # [-0.3, 0.3] assert -0.3 <= v <= 0.3 if random.random() > 0.5: v = -v return img.transform(img.size, PIL.Image.AFFINE, (1, v, 0, 0, 1, 0)) def ShearY(img, v): # [-0.3, 0.3] assert -0.3 <= v <= 0.3 if random.random() > 0.5: v = -v return img.transform(img.size, PIL.Image.AFFINE, (1, 0, 0, v, 1, 0)) def TranslateX(img, v): # [-150, 150] => percentage: [-0.45, 0.45] assert -0.45 <= v <= 0.45 if random.random() > 0.5: v = -v v = v * img.size[0] return img.transform(img.size, PIL.Image.AFFINE, (1, 0, v, 0, 1, 0)) def TranslateXabs(img, v): # [-150, 150] => percentage: [-0.45, 0.45] assert 0 <= v if random.random() > 0.5: v = -v return img.transform(img.size, PIL.Image.AFFINE, (1, 0, v, 0, 1, 0)) def TranslateY(img, v): # [-150, 150] => percentage: [-0.45, 0.45] assert -0.45 <= v <= 0.45 if random.random() > 0.5: v = -v v = v * img.size[1] return img.transform(img.size, PIL.Image.AFFINE, (1, 0, 0, 0, 1, v)) def TranslateYabs(img, v): # [-150, 150] => percentage: [-0.45, 0.45] assert 0 <= v if random.random() > 0.5: v = -v return img.transform(img.size, PIL.Image.AFFINE, (1, 0, 0, 0, 1, v)) def Rotate(img, v): # [-30, 30] assert -30 <= v <= 30 if random.random() > 0.5: v = -v return img.rotate(v) def AutoContrast(img, _): return PIL.ImageOps.autocontrast(img) def Invert(img, _): return PIL.ImageOps.invert(img) def Equalize(img, _): return PIL.ImageOps.equalize(img) def Flip(img, _): # not from the paper return PIL.ImageOps.mirror(img) def Solarize(img, v): # [0, 256] assert 0 <= v <= 256 return PIL.ImageOps.solarize(img, v) def SolarizeAdd(img, addition=0, threshold=128): img_np = np.array(img).astype(np.int) img_np = img_np + addition img_np = np.clip(img_np, 0, 255) img_np = img_np.astype(np.uint8) img = Image.fromarray(img_np) return PIL.ImageOps.solarize(img, threshold) def Posterize(img, v): # [4, 8] v = int(v) v = max(1, v) return PIL.ImageOps.posterize(img, v) def Contrast(img, v): # [0.1,1.9] assert 0.1 <= v <= 1.9 return PIL.ImageEnhance.Contrast(img).enhance(v) def Color(img, v): # [0.1,1.9] assert 0.1 <= v <= 1.9 return PIL.ImageEnhance.Color(img).enhance(v) def Brightness(img, v): # [0.1,1.9] assert 0.1 <= v <= 1.9 return PIL.ImageEnhance.Brightness(img).enhance(v) def Sharpness(img, v): # [0.1,1.9] assert 0.1 <= v <= 1.9 return PIL.ImageEnhance.Sharpness(img).enhance(v) def Cutout(img, v): # [0, 60] => percentage: [0, 0.2] assert 0.0 <= v <= 0.2 if v <= 0.0: return img v = v * img.size[0] return CutoutAbs(img, v) def CutoutAbs(img, v): # [0, 60] => percentage: [0, 0.2] # assert 0 <= v <= 20 if v < 0: return img w, h = img.size x0 = np.random.uniform(w) y0 = np.random.uniform(h) x0 = int(max(0, x0 - v / 2.0)) y0 = int(max(0, y0 - v / 2.0)) x1 = min(w, x0 + v) y1 = min(h, y0 + v) xy = (x0, y0, x1, y1) color = (125, 123, 114) # color = (0, 0, 0) img = img.copy() PIL.ImageDraw.Draw(img).rectangle(xy, color) return img def SamplePairing(imgs): # [0, 0.4] def f(img1, v): i = np.random.choice(len(imgs)) img2 = PIL.Image.fromarray(imgs[i]) return PIL.Image.blend(img1, img2, v) return f def Identity(img, v): return img def augment_list(): # 16 oeprations and their ranges # https://github.com/google-research/uda/blob/master/image/randaugment/policies.py#L57 # l = [ # (Identity, 0., 1.0), # (ShearX, 0., 0.3), # 0 # (ShearY, 0., 0.3), # 1 # (TranslateX, 0., 0.33), # 2 # (TranslateY, 0., 0.33), # 3 # (Rotate, 0, 30), # 4 # (AutoContrast, 0, 1), # 5 # (Invert, 0, 1), # 6 # (Equalize, 0, 1), # 7 # (Solarize, 0, 110), # 8 # (Posterize, 4, 8), # 9 # # (Contrast, 0.1, 1.9), # 10 # (Color, 0.1, 1.9), # 11 # (Brightness, 0.1, 1.9), # 12 # (Sharpness, 0.1, 1.9), # 13 # # (Cutout, 0, 0.2), # 14 # # (SamplePairing(imgs), 0, 0.4), # 15 # ] # https://github.com/tensorflow/tpu/blob/8462d083dd89489a79e3200bcc8d4063bf362186/models/official/efficientnet/autoaugment.py#L505 l = [ (AutoContrast, 0, 1), (Equalize, 0, 1), # (Invert, 0, 1), (Rotate, 0, 30), (Posterize, 0, 4), (Solarize, 0, 256), (SolarizeAdd, 0, 110), (Color, 0.1, 1.9), (Contrast, 0.1, 1.9), (Brightness, 0.1, 1.9), (Sharpness, 0.1, 1.9), (ShearX, 0.0, 0.3), (ShearY, 0.0, 0.3), # (CutoutAbs, 0, 40), (TranslateXabs, 0.0, 100), (TranslateYabs, 0.0, 100), ] return l class Lighting(object): """Lighting noise(AlexNet - style PCA - based noise)""" def __init__(self, alphastd, eigval, eigvec): self.alphastd = alphastd self.eigval = torch.Tensor(eigval) self.eigvec = torch.Tensor(eigvec) def __call__(self, img): if self.alphastd == 0: return img alpha = img.new().resize_(3).normal_(0, self.alphastd) rgb = ( self.eigvec.type_as(img) .clone() .mul(alpha.view(1, 3).expand(3, 3)) .mul(self.eigval.view(1, 3).expand(3, 3)) .sum(1) .squeeze() ) return img.add(rgb.view(3, 1, 1).expand_as(img)) class CutoutDefault(object): """ Reference : https://github.com/quark0/darts/blob/master/cnn/utils.py """ def __init__(self, length): self.length = length def __call__(self, img): h, w = img.size(1), img.size(2) mask = np.ones((h, w), np.float32) y = np.random.randint(h) x = np.random.randint(w) y1 = np.clip(y - self.length // 2, 0, h) y2 = np.clip(y + self.length // 2, 0, h) x1 = np.clip(x - self.length // 2, 0, w) x2 = np.clip(x + self.length // 2, 0, w) mask[y1:y2, x1:x2] = 0.0 mask = torch.from_numpy(mask) mask = mask.expand_as(img) img *= mask return img class RandAugment: def __init__(self, n, m): self.n = n self.m = m # [0, 30] self.augment_list = augment_list() def __call__(self, img): ops = random.choices(self.augment_list, k=self.n) for op, minval, maxval in ops: val = (float(self.m) / 30) * float(maxval - minval) + minval img = op(img, val) return img
6,990
24.892593
134
py
Mr.Right
Mr.Right-main/models/METER/meter/transforms/__init__.py
from .transform import ( pixelbert_transform, pixelbert_transform_randaug, vit_transform, vit_transform_randaug, imagenet_transform, imagenet_transform_randaug, clip_transform, clip_transform_randaug, ) _transforms = { "pixelbert": pixelbert_transform, "pixelbert_randaug": pixelbert_transform_randaug, "vit": vit_transform, "vit_randaug": vit_transform_randaug, "imagenet": imagenet_transform, "imagenet_randaug": imagenet_transform_randaug, "clip": clip_transform, "clip_randaug": clip_transform_randaug, } def keys_to_transforms(keys: list, size=224): return [_transforms[key](size=size) for key in keys]
678
26.16
56
py
Mr.Right
Mr.Right-main/models/ALBEF/models/xbert.py
# coding=utf-8 # Copyright 2018 The Google AI Language Team Authors and The HuggingFace Inc. team. # Copyright (c) 2018, NVIDIA CORPORATION. All rights reserved. # # Licensed under the Apache License, Version 2.0 (the "License"); # you may not use this file except in compliance with the License. # You may obtain a copy of the License at # # http://www.apache.org/licenses/LICENSE-2.0 # # Unless required by applicable law or agreed to in writing, software # distributed under the License is distributed on an "AS IS" BASIS, # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. # See the License for the specific language governing permissions and # limitations under the License. """PyTorch BERT model. """ import math import os import warnings from dataclasses import dataclass from typing import Optional, Tuple import torch from torch import Tensor, device, dtype, nn import torch.utils.checkpoint from torch import nn from torch.nn import CrossEntropyLoss, MSELoss import torch.nn.functional as F from transformers.activations import ACT2FN from transformers.file_utils import ( ModelOutput, add_code_sample_docstrings, add_start_docstrings, add_start_docstrings_to_model_forward, replace_return_docstrings, ) from transformers.modeling_outputs import ( BaseModelOutputWithPastAndCrossAttentions, BaseModelOutputWithPoolingAndCrossAttentions, CausalLMOutputWithCrossAttentions, MaskedLMOutput, MultipleChoiceModelOutput, NextSentencePredictorOutput, QuestionAnsweringModelOutput, SequenceClassifierOutput, TokenClassifierOutput, ) from transformers.modeling_utils import ( PreTrainedModel, apply_chunking_to_forward, find_pruneable_heads_and_indices, prune_linear_layer, ) from transformers.utils import logging from transformers.models.bert.configuration_bert import BertConfig import transformers transformers.logging.set_verbosity_error() logger = logging.get_logger(__name__) _CONFIG_FOR_DOC = "BertConfig" _TOKENIZER_FOR_DOC = "BertTokenizer" BERT_PRETRAINED_MODEL_ARCHIVE_LIST = [ "bert-base-uncased", "bert-large-uncased", "bert-base-cased", "bert-large-cased", "bert-base-multilingual-uncased", "bert-base-multilingual-cased", "bert-base-chinese", "bert-base-german-cased", "bert-large-uncased-whole-word-masking", "bert-large-cased-whole-word-masking", "bert-large-uncased-whole-word-masking-finetuned-squad", "bert-large-cased-whole-word-masking-finetuned-squad", "bert-base-cased-finetuned-mrpc", "bert-base-german-dbmdz-cased", "bert-base-german-dbmdz-uncased", "cl-tohoku/bert-base-japanese", "cl-tohoku/bert-base-japanese-whole-word-masking", "cl-tohoku/bert-base-japanese-char", "cl-tohoku/bert-base-japanese-char-whole-word-masking", "TurkuNLP/bert-base-finnish-cased-v1", "TurkuNLP/bert-base-finnish-uncased-v1", "wietsedv/bert-base-dutch-cased", # See all BERT models at https://huggingface.co/models?filter=bert ] def load_tf_weights_in_bert(model, config, tf_checkpoint_path): """Load tf checkpoints in a pytorch model.""" try: import re import numpy as np import tensorflow as tf except ImportError: logger.error( "Loading a TensorFlow model in PyTorch, requires TensorFlow to be installed. Please see " "https://www.tensorflow.org/install/ for installation instructions." ) raise tf_path = os.path.abspath(tf_checkpoint_path) logger.info("Converting TensorFlow checkpoint from {}".format(tf_path)) # Load weights from TF model init_vars = tf.train.list_variables(tf_path) names = [] arrays = [] for name, shape in init_vars: logger.info("Loading TF weight {} with shape {}".format(name, shape)) array = tf.train.load_variable(tf_path, name) names.append(name) arrays.append(array) for name, array in zip(names, arrays): name = name.split("/") # adam_v and adam_m are variables used in AdamWeightDecayOptimizer to calculated m and v # which are not required for using pretrained model if any( n in ["adam_v", "adam_m", "AdamWeightDecayOptimizer", "AdamWeightDecayOptimizer_1", "global_step"] for n in name ): logger.info("Skipping {}".format("/".join(name))) continue pointer = model for m_name in name: if re.fullmatch(r"[A-Za-z]+_\d+", m_name): scope_names = re.split(r"_(\d+)", m_name) else: scope_names = [m_name] if scope_names[0] == "kernel" or scope_names[0] == "gamma": pointer = getattr(pointer, "weight") elif scope_names[0] == "output_bias" or scope_names[0] == "beta": pointer = getattr(pointer, "bias") elif scope_names[0] == "output_weights": pointer = getattr(pointer, "weight") elif scope_names[0] == "squad": pointer = getattr(pointer, "classifier") else: try: pointer = getattr(pointer, scope_names[0]) except AttributeError: logger.info("Skipping {}".format("/".join(name))) continue if len(scope_names) >= 2: num = int(scope_names[1]) pointer = pointer[num] if m_name[-11:] == "_embeddings": pointer = getattr(pointer, "weight") elif m_name == "kernel": array = np.transpose(array) try: assert ( pointer.shape == array.shape ), f"Pointer shape {pointer.shape} and array shape {array.shape} mismatched" except AssertionError as e: e.args += (pointer.shape, array.shape) raise logger.info("Initialize PyTorch weight {}".format(name)) pointer.data = torch.from_numpy(array) return model class BertEmbeddings(nn.Module): """Construct the embeddings from word, position and token_type embeddings.""" def __init__(self, config): super().__init__() self.word_embeddings = nn.Embedding(config.vocab_size, config.hidden_size, padding_idx=config.pad_token_id) self.position_embeddings = nn.Embedding(config.max_position_embeddings, config.hidden_size) self.token_type_embeddings = nn.Embedding(config.type_vocab_size, config.hidden_size) # self.LayerNorm is not snake-cased to stick with TensorFlow model variable name and be able to load # any TensorFlow checkpoint file self.LayerNorm = nn.LayerNorm(config.hidden_size, eps=config.layer_norm_eps) self.dropout = nn.Dropout(config.hidden_dropout_prob) # position_ids (1, len position emb) is contiguous in memory and exported when serialized self.register_buffer("position_ids", torch.arange(config.max_position_embeddings).expand((1, -1))) self.position_embedding_type = getattr(config, "position_embedding_type", "absolute") self.config = config def forward( self, input_ids=None, token_type_ids=None, position_ids=None, inputs_embeds=None, past_key_values_length=0 ): if input_ids is not None: input_shape = input_ids.size() else: input_shape = inputs_embeds.size()[:-1] seq_length = input_shape[1] if position_ids is None: position_ids = self.position_ids[:, past_key_values_length : seq_length + past_key_values_length] if token_type_ids is None: token_type_ids = torch.zeros(input_shape, dtype=torch.long, device=self.position_ids.device) if inputs_embeds is None: inputs_embeds = self.word_embeddings(input_ids) token_type_embeddings = self.token_type_embeddings(token_type_ids) embeddings = inputs_embeds + token_type_embeddings if self.position_embedding_type == "absolute": position_embeddings = self.position_embeddings(position_ids) embeddings += position_embeddings embeddings = self.LayerNorm(embeddings) embeddings = self.dropout(embeddings) return embeddings class BertSelfAttention(nn.Module): def __init__(self, config, is_cross_attention): super().__init__() self.config = config if config.hidden_size % config.num_attention_heads != 0 and not hasattr(config, "embedding_size"): raise ValueError( "The hidden size (%d) is not a multiple of the number of attention " "heads (%d)" % (config.hidden_size, config.num_attention_heads) ) self.num_attention_heads = config.num_attention_heads self.attention_head_size = int(config.hidden_size / config.num_attention_heads) self.all_head_size = self.num_attention_heads * self.attention_head_size self.query = nn.Linear(config.hidden_size, self.all_head_size) if is_cross_attention: self.key = nn.Linear(config.encoder_width, self.all_head_size) self.value = nn.Linear(config.encoder_width, self.all_head_size) else: self.key = nn.Linear(config.hidden_size, self.all_head_size) self.value = nn.Linear(config.hidden_size, self.all_head_size) self.dropout = nn.Dropout(config.attention_probs_dropout_prob) self.position_embedding_type = getattr(config, "position_embedding_type", "absolute") if self.position_embedding_type == "relative_key" or self.position_embedding_type == "relative_key_query": self.max_position_embeddings = config.max_position_embeddings self.distance_embedding = nn.Embedding(2 * config.max_position_embeddings - 1, self.attention_head_size) self.save_attention = False def save_attn_gradients(self, attn_gradients): self.attn_gradients = attn_gradients def get_attn_gradients(self): return self.attn_gradients def save_attention_map(self, attention_map): self.attention_map = attention_map def get_attention_map(self): return self.attention_map def transpose_for_scores(self, x): new_x_shape = x.size()[:-1] + (self.num_attention_heads, self.attention_head_size) x = x.view(*new_x_shape) return x.permute(0, 2, 1, 3) def forward( self, hidden_states, attention_mask=None, head_mask=None, encoder_hidden_states=None, encoder_attention_mask=None, past_key_value=None, output_attentions=False, ): mixed_query_layer = self.query(hidden_states) # If this is instantiated as a cross-attention module, the keys # and values come from an encoder; the attention mask needs to be # such that the encoder's padding tokens are not attended to. is_cross_attention = encoder_hidden_states is not None if is_cross_attention: key_layer = self.transpose_for_scores(self.key(encoder_hidden_states)) value_layer = self.transpose_for_scores(self.value(encoder_hidden_states)) attention_mask = encoder_attention_mask elif past_key_value is not None: key_layer = self.transpose_for_scores(self.key(hidden_states)) value_layer = self.transpose_for_scores(self.value(hidden_states)) key_layer = torch.cat([past_key_value[0], key_layer], dim=2) value_layer = torch.cat([past_key_value[1], value_layer], dim=2) else: key_layer = self.transpose_for_scores(self.key(hidden_states)) value_layer = self.transpose_for_scores(self.value(hidden_states)) query_layer = self.transpose_for_scores(mixed_query_layer) past_key_value = (key_layer, value_layer) # Take the dot product between "query" and "key" to get the raw attention scores. attention_scores = torch.matmul(query_layer, key_layer.transpose(-1, -2)) if self.position_embedding_type == "relative_key" or self.position_embedding_type == "relative_key_query": seq_length = hidden_states.size()[1] position_ids_l = torch.arange(seq_length, dtype=torch.long, device=hidden_states.device).view(-1, 1) position_ids_r = torch.arange(seq_length, dtype=torch.long, device=hidden_states.device).view(1, -1) distance = position_ids_l - position_ids_r positional_embedding = self.distance_embedding(distance + self.max_position_embeddings - 1) positional_embedding = positional_embedding.to(dtype=query_layer.dtype) # fp16 compatibility if self.position_embedding_type == "relative_key": relative_position_scores = torch.einsum("bhld,lrd->bhlr", query_layer, positional_embedding) attention_scores = attention_scores + relative_position_scores elif self.position_embedding_type == "relative_key_query": relative_position_scores_query = torch.einsum("bhld,lrd->bhlr", query_layer, positional_embedding) relative_position_scores_key = torch.einsum("bhrd,lrd->bhlr", key_layer, positional_embedding) attention_scores = attention_scores + relative_position_scores_query + relative_position_scores_key attention_scores = attention_scores / math.sqrt(self.attention_head_size) if attention_mask is not None: # Apply the attention mask is (precomputed for all layers in BertModel forward() function) attention_scores = attention_scores + attention_mask # Normalize the attention scores to probabilities. attention_probs = nn.Softmax(dim=-1)(attention_scores) if self.save_attention: self.save_attention_map(attention_probs) attention_probs.register_hook(self.save_attn_gradients) # This is actually dropping out entire tokens to attend to, which might # seem a bit unusual, but is taken from the original Transformer paper. attention_probs_dropped = self.dropout(attention_probs) # Mask heads if we want to if head_mask is not None: attention_probs_dropped = attention_probs_dropped * head_mask context_layer = torch.matmul(attention_probs_dropped, value_layer) context_layer = context_layer.permute(0, 2, 1, 3).contiguous() new_context_layer_shape = context_layer.size()[:-2] + (self.all_head_size,) context_layer = context_layer.view(*new_context_layer_shape) outputs = (context_layer, attention_probs) if output_attentions else (context_layer,) outputs = outputs + (past_key_value,) return outputs class BertSelfOutput(nn.Module): def __init__(self, config): super().__init__() self.dense = nn.Linear(config.hidden_size, config.hidden_size) self.LayerNorm = nn.LayerNorm(config.hidden_size, eps=config.layer_norm_eps) self.dropout = nn.Dropout(config.hidden_dropout_prob) def forward(self, hidden_states, input_tensor): hidden_states = self.dense(hidden_states) hidden_states = self.dropout(hidden_states) hidden_states = self.LayerNorm(hidden_states + input_tensor) return hidden_states class BertAttention(nn.Module): def __init__(self, config, is_cross_attention=False): super().__init__() self.self = BertSelfAttention(config, is_cross_attention) self.output = BertSelfOutput(config) self.pruned_heads = set() def prune_heads(self, heads): if len(heads) == 0: return heads, index = find_pruneable_heads_and_indices( heads, self.self.num_attention_heads, self.self.attention_head_size, self.pruned_heads ) # Prune linear layers self.self.query = prune_linear_layer(self.self.query, index) self.self.key = prune_linear_layer(self.self.key, index) self.self.value = prune_linear_layer(self.self.value, index) self.output.dense = prune_linear_layer(self.output.dense, index, dim=1) # Update hyper params and store pruned heads self.self.num_attention_heads = self.self.num_attention_heads - len(heads) self.self.all_head_size = self.self.attention_head_size * self.self.num_attention_heads self.pruned_heads = self.pruned_heads.union(heads) def forward( self, hidden_states, attention_mask=None, head_mask=None, encoder_hidden_states=None, encoder_attention_mask=None, past_key_value=None, output_attentions=False, ): self_outputs = self.self( hidden_states, attention_mask, head_mask, encoder_hidden_states, encoder_attention_mask, past_key_value, output_attentions, ) attention_output = self.output(self_outputs[0], hidden_states) outputs = (attention_output,) + self_outputs[1:] # add attentions if we output them return outputs class BertIntermediate(nn.Module): def __init__(self, config): super().__init__() self.dense = nn.Linear(config.hidden_size, config.intermediate_size) if isinstance(config.hidden_act, str): self.intermediate_act_fn = ACT2FN[config.hidden_act] else: self.intermediate_act_fn = config.hidden_act def forward(self, hidden_states): hidden_states = self.dense(hidden_states) hidden_states = self.intermediate_act_fn(hidden_states) return hidden_states class BertOutput(nn.Module): def __init__(self, config): super().__init__() self.dense = nn.Linear(config.intermediate_size, config.hidden_size) self.LayerNorm = nn.LayerNorm(config.hidden_size, eps=config.layer_norm_eps) self.dropout = nn.Dropout(config.hidden_dropout_prob) def forward(self, hidden_states, input_tensor): hidden_states = self.dense(hidden_states) hidden_states = self.dropout(hidden_states) hidden_states = self.LayerNorm(hidden_states + input_tensor) return hidden_states class BertLayer(nn.Module): def __init__(self, config, layer_num): super().__init__() self.config = config self.chunk_size_feed_forward = config.chunk_size_feed_forward self.seq_len_dim = 1 self.attention = BertAttention(config) self.has_cross_attention = (layer_num >= config.fusion_layer) if self.has_cross_attention: self.layer_num = layer_num self.crossattention = BertAttention(config, is_cross_attention=True) self.intermediate = BertIntermediate(config) self.output = BertOutput(config) def forward( self, hidden_states, attention_mask=None, head_mask=None, encoder_hidden_states=None, encoder_attention_mask=None, past_key_value=None, output_attentions=False, ): # decoder uni-directional self-attention cached key/values tuple is at positions 1,2 self_attn_past_key_value = past_key_value[:2] if past_key_value is not None else None self_attention_outputs = self.attention( hidden_states, attention_mask, head_mask, output_attentions=output_attentions, past_key_value=self_attn_past_key_value, ) attention_output = self_attention_outputs[0] outputs = self_attention_outputs[1:-1] present_key_value = self_attention_outputs[-1] if self.has_cross_attention: assert encoder_hidden_states is not None, "encoder_hidden_states must be given for cross-attention layers" if type(encoder_hidden_states) == list: cross_attention_outputs = self.crossattention( attention_output, attention_mask, head_mask, encoder_hidden_states[(self.layer_num-self.config.fusion_layer)%len(encoder_hidden_states)], encoder_attention_mask[(self.layer_num-self.config.fusion_layer)%len(encoder_hidden_states)], output_attentions=output_attentions, ) attention_output = cross_attention_outputs[0] outputs = outputs + cross_attention_outputs[1:-1] else: cross_attention_outputs = self.crossattention( attention_output, attention_mask, head_mask, encoder_hidden_states, encoder_attention_mask, output_attentions=output_attentions, ) attention_output = cross_attention_outputs[0] outputs = outputs + cross_attention_outputs[1:-1] # add cross attentions if we output attention weights layer_output = apply_chunking_to_forward( self.feed_forward_chunk, self.chunk_size_feed_forward, self.seq_len_dim, attention_output ) outputs = (layer_output,) + outputs outputs = outputs + (present_key_value,) return outputs def feed_forward_chunk(self, attention_output): intermediate_output = self.intermediate(attention_output) layer_output = self.output(intermediate_output, attention_output) return layer_output class BertEncoder(nn.Module): def __init__(self, config): super().__init__() self.config = config self.layer = nn.ModuleList([BertLayer(config,i) for i in range(config.num_hidden_layers)]) def forward( self, hidden_states, attention_mask=None, head_mask=None, encoder_hidden_states=None, encoder_attention_mask=None, past_key_values=None, use_cache=None, output_attentions=False, output_hidden_states=False, return_dict=True, mode='multi_modal', ): all_hidden_states = () if output_hidden_states else None all_self_attentions = () if output_attentions else None all_cross_attentions = () if output_attentions and self.config.add_cross_attention else None next_decoder_cache = () if use_cache else None if mode=='text': start_layer = 0 output_layer = self.config.fusion_layer elif mode=='fusion': start_layer = self.config.fusion_layer output_layer = self.config.num_hidden_layers elif mode=='multi_modal': start_layer = 0 output_layer = self.config.num_hidden_layers for i in range(start_layer, output_layer): layer_module = self.layer[i] if output_hidden_states: all_hidden_states = all_hidden_states + (hidden_states,) layer_head_mask = head_mask[i] if head_mask is not None else None past_key_value = past_key_values[i] if past_key_values is not None else None if getattr(self.config, "gradient_checkpointing", False) and self.training: if use_cache: logger.warn( "`use_cache=True` is incompatible with `config.gradient_checkpointing=True`. Setting " "`use_cache=False`..." ) use_cache = False def create_custom_forward(module): def custom_forward(*inputs): return module(*inputs, past_key_value, output_attentions) return custom_forward layer_outputs = torch.utils.checkpoint.checkpoint( create_custom_forward(layer_module), hidden_states, attention_mask, layer_head_mask, encoder_hidden_states, encoder_attention_mask, ) else: layer_outputs = layer_module( hidden_states, attention_mask, layer_head_mask, encoder_hidden_states, encoder_attention_mask, past_key_value, output_attentions, ) hidden_states = layer_outputs[0] if use_cache: next_decoder_cache += (layer_outputs[-1],) if output_attentions: all_self_attentions = all_self_attentions + (layer_outputs[1],) if output_hidden_states: all_hidden_states = all_hidden_states + (hidden_states,) if not return_dict: return tuple( v for v in [ hidden_states, next_decoder_cache, all_hidden_states, all_self_attentions, all_cross_attentions, ] if v is not None ) return BaseModelOutputWithPastAndCrossAttentions( last_hidden_state=hidden_states, past_key_values=next_decoder_cache, hidden_states=all_hidden_states, attentions=all_self_attentions, cross_attentions=all_cross_attentions, ) class BertPooler(nn.Module): def __init__(self, config): super().__init__() self.dense = nn.Linear(config.hidden_size, config.hidden_size) self.activation = nn.Tanh() def forward(self, hidden_states): # We "pool" the model by simply taking the hidden state corresponding # to the first token. first_token_tensor = hidden_states[:, 0] pooled_output = self.dense(first_token_tensor) pooled_output = self.activation(pooled_output) return pooled_output class BertPredictionHeadTransform(nn.Module): def __init__(self, config): super().__init__() self.dense = nn.Linear(config.hidden_size, config.hidden_size) if isinstance(config.hidden_act, str): self.transform_act_fn = ACT2FN[config.hidden_act] else: self.transform_act_fn = config.hidden_act self.LayerNorm = nn.LayerNorm(config.hidden_size, eps=config.layer_norm_eps) def forward(self, hidden_states): hidden_states = self.dense(hidden_states) hidden_states = self.transform_act_fn(hidden_states) hidden_states = self.LayerNorm(hidden_states) return hidden_states class BertLMPredictionHead(nn.Module): def __init__(self, config): super().__init__() self.transform = BertPredictionHeadTransform(config) # The output weights are the same as the input embeddings, but there is # an output-only bias for each token. self.decoder = nn.Linear(config.hidden_size, config.vocab_size, bias=False) self.bias = nn.Parameter(torch.zeros(config.vocab_size)) # Need a link between the two variables so that the bias is correctly resized with `resize_token_embeddings` self.decoder.bias = self.bias def forward(self, hidden_states): hidden_states = self.transform(hidden_states) hidden_states = self.decoder(hidden_states) return hidden_states class BertOnlyMLMHead(nn.Module): def __init__(self, config): super().__init__() self.predictions = BertLMPredictionHead(config) def forward(self, sequence_output): prediction_scores = self.predictions(sequence_output) return prediction_scores class BertOnlyNSPHead(nn.Module): def __init__(self, config): super().__init__() self.seq_relationship = nn.Linear(config.hidden_size, 2) def forward(self, pooled_output): seq_relationship_score = self.seq_relationship(pooled_output) return seq_relationship_score class BertPreTrainingHeads(nn.Module): def __init__(self, config): super().__init__() self.predictions = BertLMPredictionHead(config) self.seq_relationship = nn.Linear(config.hidden_size, 2) def forward(self, sequence_output, pooled_output): prediction_scores = self.predictions(sequence_output) seq_relationship_score = self.seq_relationship(pooled_output) return prediction_scores, seq_relationship_score class BertPreTrainedModel(PreTrainedModel): """ An abstract class to handle weights initialization and a simple interface for downloading and loading pretrained models. """ config_class = BertConfig load_tf_weights = load_tf_weights_in_bert base_model_prefix = "bert" _keys_to_ignore_on_load_missing = [r"position_ids"] def _init_weights(self, module): """ Initialize the weights """ if isinstance(module, (nn.Linear, nn.Embedding)): # Slightly different from the TF version which uses truncated_normal for initialization # cf https://github.com/pytorch/pytorch/pull/5617 module.weight.data.normal_(mean=0.0, std=self.config.initializer_range) elif isinstance(module, nn.LayerNorm): module.bias.data.zero_() module.weight.data.fill_(1.0) if isinstance(module, nn.Linear) and module.bias is not None: module.bias.data.zero_() @dataclass class BertForPreTrainingOutput(ModelOutput): """ Output type of :class:`~transformers.BertForPreTraining`. Args: loss (`optional`, returned when ``labels`` is provided, ``torch.FloatTensor`` of shape :obj:`(1,)`): Total loss as the sum of the masked language modeling loss and the next sequence prediction (classification) loss. prediction_logits (:obj:`torch.FloatTensor` of shape :obj:`(batch_size, sequence_length, config.vocab_size)`): Prediction scores of the language modeling head (scores for each vocabulary token before SoftMax). seq_relationship_logits (:obj:`torch.FloatTensor` of shape :obj:`(batch_size, 2)`): Prediction scores of the next sequence prediction (classification) head (scores of True/False continuation before SoftMax). hidden_states (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_hidden_states=True`` is passed or when ``config.output_hidden_states=True``): Tuple of :obj:`torch.FloatTensor` (one for the output of the embeddings + one for the output of each layer) of shape :obj:`(batch_size, sequence_length, hidden_size)`. Hidden-states of the model at the output of each layer plus the initial embedding outputs. attentions (:obj:`tuple(torch.FloatTensor)`, `optional`, returned when ``output_attentions=True`` is passed or when ``config.output_attentions=True``): Tuple of :obj:`torch.FloatTensor` (one for each layer) of shape :obj:`(batch_size, num_heads, sequence_length, sequence_length)`. Attentions weights after the attention softmax, used to compute the weighted average in the self-attention heads. """ loss: Optional[torch.FloatTensor] = None prediction_logits: torch.FloatTensor = None seq_relationship_logits: torch.FloatTensor = None hidden_states: Optional[Tuple[torch.FloatTensor]] = None attentions: Optional[Tuple[torch.FloatTensor]] = None BERT_START_DOCSTRING = r""" This model inherits from :class:`~transformers.PreTrainedModel`. Check the superclass documentation for the generic methods the library implements for all its model (such as downloading or saving, resizing the input embeddings, pruning heads etc.) This model is also a PyTorch `torch.nn.Module <https://pytorch.org/docs/stable/nn.html#torch.nn.Module>`__ subclass. Use it as a regular PyTorch Module and refer to the PyTorch documentation for all matter related to general usage and behavior. Parameters: config (:class:`~transformers.BertConfig`): Model configuration class with all the parameters of the model. Initializing with a config file does not load the weights associated with the model, only the configuration. Check out the :meth:`~transformers.PreTrainedModel.from_pretrained` method to load the model weights. """ BERT_INPUTS_DOCSTRING = r""" Args: input_ids (:obj:`torch.LongTensor` of shape :obj:`({0})`): Indices of input sequence tokens in the vocabulary. Indices can be obtained using :class:`~transformers.BertTokenizer`. See :meth:`transformers.PreTrainedTokenizer.encode` and :meth:`transformers.PreTrainedTokenizer.__call__` for details. `What are input IDs? <../glossary.html#input-ids>`__ attention_mask (:obj:`torch.FloatTensor` of shape :obj:`({0})`, `optional`): Mask to avoid performing attention on padding token indices. Mask values selected in ``[0, 1]``: - 1 for tokens that are **not masked**, - 0 for tokens that are **masked**. `What are attention masks? <../glossary.html#attention-mask>`__ token_type_ids (:obj:`torch.LongTensor` of shape :obj:`({0})`, `optional`): Segment token indices to indicate first and second portions of the inputs. Indices are selected in ``[0, 1]``: - 0 corresponds to a `sentence A` token, - 1 corresponds to a `sentence B` token. `What are token type IDs? <../glossary.html#token-type-ids>`_ position_ids (:obj:`torch.LongTensor` of shape :obj:`({0})`, `optional`): Indices of positions of each input sequence tokens in the position embeddings. Selected in the range ``[0, config.max_position_embeddings - 1]``. `What are position IDs? <../glossary.html#position-ids>`_ head_mask (:obj:`torch.FloatTensor` of shape :obj:`(num_heads,)` or :obj:`(num_layers, num_heads)`, `optional`): Mask to nullify selected heads of the self-attention modules. Mask values selected in ``[0, 1]``: - 1 indicates the head is **not masked**, - 0 indicates the head is **masked**. inputs_embeds (:obj:`torch.FloatTensor` of shape :obj:`({0}, hidden_size)`, `optional`): Optionally, instead of passing :obj:`input_ids` you can choose to directly pass an embedded representation. This is useful if you want more control over how to convert :obj:`input_ids` indices into associated vectors than the model's internal embedding lookup matrix. output_attentions (:obj:`bool`, `optional`): Whether or not to return the attentions tensors of all attention layers. See ``attentions`` under returned tensors for more detail. output_hidden_states (:obj:`bool`, `optional`): Whether or not to return the hidden states of all layers. See ``hidden_states`` under returned tensors for more detail. return_dict (:obj:`bool`, `optional`): Whether or not to return a :class:`~transformers.file_utils.ModelOutput` instead of a plain tuple. """ @add_start_docstrings( "The bare Bert Model transformer outputting raw hidden-states without any specific head on top.", BERT_START_DOCSTRING, ) class BertModel(BertPreTrainedModel): """ The model can behave as an encoder (with only self-attention) as well as a decoder, in which case a layer of cross-attention is added between the self-attention layers, following the architecture described in `Attention is all you need <https://arxiv.org/abs/1706.03762>`__ by Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N. Gomez, Lukasz Kaiser and Illia Polosukhin. argument and :obj:`add_cross_attention` set to :obj:`True`; an :obj:`encoder_hidden_states` is then expected as an input to the forward pass. """ def __init__(self, config, add_pooling_layer=True): super().__init__(config) self.config = config self.embeddings = BertEmbeddings(config) self.encoder = BertEncoder(config) self.pooler = BertPooler(config) if add_pooling_layer else None self.init_weights() def get_input_embeddings(self): return self.embeddings.word_embeddings def set_input_embeddings(self, value): self.embeddings.word_embeddings = value def _prune_heads(self, heads_to_prune): """ Prunes heads of the model. heads_to_prune: dict of {layer_num: list of heads to prune in this layer} See base class PreTrainedModel """ for layer, heads in heads_to_prune.items(): self.encoder.layer[layer].attention.prune_heads(heads) @add_start_docstrings_to_model_forward(BERT_INPUTS_DOCSTRING.format("batch_size, sequence_length")) @add_code_sample_docstrings( tokenizer_class=_TOKENIZER_FOR_DOC, checkpoint="bert-base-uncased", output_type=BaseModelOutputWithPoolingAndCrossAttentions, config_class=_CONFIG_FOR_DOC, ) def get_extended_attention_mask(self, attention_mask: Tensor, input_shape: Tuple[int], device: device, is_decoder: bool) -> Tensor: """ Makes broadcastable attention and causal masks so that future and masked tokens are ignored. Arguments: attention_mask (:obj:`torch.Tensor`): Mask with ones indicating tokens to attend to, zeros for tokens to ignore. input_shape (:obj:`Tuple[int]`): The shape of the input to the model. device: (:obj:`torch.device`): The device of the input to the model. Returns: :obj:`torch.Tensor` The extended attention mask, with a the same dtype as :obj:`attention_mask.dtype`. """ # We can provide a self-attention mask of dimensions [batch_size, from_seq_length, to_seq_length] # ourselves in which case we just need to make it broadcastable to all heads. if attention_mask.dim() == 3: extended_attention_mask = attention_mask[:, None, :, :] elif attention_mask.dim() == 2: # Provided a padding mask of dimensions [batch_size, seq_length] # - if the model is a decoder, apply a causal mask in addition to the padding mask # - if the model is an encoder, make the mask broadcastable to [batch_size, num_heads, seq_length, seq_length] if is_decoder: batch_size, seq_length = input_shape seq_ids = torch.arange(seq_length, device=device) causal_mask = seq_ids[None, None, :].repeat(batch_size, seq_length, 1) <= seq_ids[None, :, None] # in case past_key_values are used we need to add a prefix ones mask to the causal mask # causal and attention masks must have same type with pytorch version < 1.3 causal_mask = causal_mask.to(attention_mask.dtype) if causal_mask.shape[1] < attention_mask.shape[1]: prefix_seq_len = attention_mask.shape[1] - causal_mask.shape[1] causal_mask = torch.cat( [ torch.ones( (batch_size, seq_length, prefix_seq_len), device=device, dtype=causal_mask.dtype ), causal_mask, ], axis=-1, ) extended_attention_mask = causal_mask[:, None, :, :] * attention_mask[:, None, None, :] else: extended_attention_mask = attention_mask[:, None, None, :] else: raise ValueError( "Wrong shape for input_ids (shape {}) or attention_mask (shape {})".format( input_shape, attention_mask.shape ) ) # Since attention_mask is 1.0 for positions we want to attend and 0.0 for # masked positions, this operation will create a tensor which is 0.0 for # positions we want to attend and -10000.0 for masked positions. # Since we are adding it to the raw scores before the softmax, this is # effectively the same as removing these entirely. extended_attention_mask = extended_attention_mask.to(dtype=self.dtype) # fp16 compatibility extended_attention_mask = (1.0 - extended_attention_mask) * -10000.0 return extended_attention_mask def forward( self, input_ids=None, attention_mask=None, token_type_ids=None, position_ids=None, head_mask=None, inputs_embeds=None, encoder_embeds=None, encoder_hidden_states=None, encoder_attention_mask=None, past_key_values=None, use_cache=None, output_attentions=None, output_hidden_states=None, return_dict=None, is_decoder=False, mode='multi_modal', ): r""" encoder_hidden_states (:obj:`torch.FloatTensor` of shape :obj:`(batch_size, sequence_length, hidden_size)`, `optional`): Sequence of hidden-states at the output of the last layer of the encoder. Used in the cross-attention if the model is configured as a decoder. encoder_attention_mask (:obj:`torch.FloatTensor` of shape :obj:`(batch_size, sequence_length)`, `optional`): Mask to avoid performing attention on the padding token indices of the encoder input. This mask is used in the cross-attention if the model is configured as a decoder. Mask values selected in ``[0, 1]``: - 1 for tokens that are **not masked**, - 0 for tokens that are **masked**. past_key_values (:obj:`tuple(tuple(torch.FloatTensor))` of length :obj:`config.n_layers` with each tuple having 4 tensors of shape :obj:`(batch_size, num_heads, sequence_length - 1, embed_size_per_head)`): Contains precomputed key and value hidden states of the attention blocks. Can be used to speed up decoding. If :obj:`past_key_values` are used, the user can optionally input only the last :obj:`decoder_input_ids` (those that don't have their past key value states given to this model) of shape :obj:`(batch_size, 1)` instead of all :obj:`decoder_input_ids` of shape :obj:`(batch_size, sequence_length)`. use_cache (:obj:`bool`, `optional`): If set to :obj:`True`, :obj:`past_key_values` key value states are returned and can be used to speed up decoding (see :obj:`past_key_values`). """ output_attentions = output_attentions if output_attentions is not None else self.config.output_attentions output_hidden_states = ( output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states ) return_dict = return_dict if return_dict is not None else self.config.use_return_dict if is_decoder: use_cache = use_cache if use_cache is not None else self.config.use_cache else: use_cache = False if input_ids is not None and inputs_embeds is not None: raise ValueError("You cannot specify both input_ids and inputs_embeds at the same time") elif input_ids is not None: input_shape = input_ids.size() batch_size, seq_length = input_shape device = input_ids.device elif inputs_embeds is not None: input_shape = inputs_embeds.size()[:-1] batch_size, seq_length = input_shape device = inputs_embeds.device elif encoder_embeds is not None: input_shape = encoder_embeds.size()[:-1] batch_size, seq_length = input_shape device = encoder_embeds.device else: raise ValueError("You have to specify either input_ids or inputs_embeds or encoder_embeds") # past_key_values_length past_key_values_length = past_key_values[0][0].shape[2] if past_key_values is not None else 0 if attention_mask is None: attention_mask = torch.ones(((batch_size, seq_length + past_key_values_length)), device=device) if token_type_ids is None: token_type_ids = torch.zeros(input_shape, dtype=torch.long, device=device) # We can provide a self-attention mask of dimensions [batch_size, from_seq_length, to_seq_length] # ourselves in which case we just need to make it broadcastable to all heads. extended_attention_mask: torch.Tensor = self.get_extended_attention_mask(attention_mask, input_shape, device, is_decoder) # If a 2D or 3D attention mask is provided for the cross-attention # we need to make broadcastable to [batch_size, num_heads, seq_length, seq_length] if encoder_hidden_states is not None: if type(encoder_hidden_states) == list: encoder_batch_size, encoder_sequence_length, _ = encoder_hidden_states[0].size() else: encoder_batch_size, encoder_sequence_length, _ = encoder_hidden_states.size() encoder_hidden_shape = (encoder_batch_size, encoder_sequence_length) if type(encoder_attention_mask) == list: encoder_extended_attention_mask = [self.invert_attention_mask(mask) for mask in encoder_attention_mask] elif encoder_attention_mask is None: encoder_attention_mask = torch.ones(encoder_hidden_shape, device=device) encoder_extended_attention_mask = self.invert_attention_mask(encoder_attention_mask) else: encoder_extended_attention_mask = self.invert_attention_mask(encoder_attention_mask) else: encoder_extended_attention_mask = None # Prepare head mask if needed # 1.0 in head_mask indicate we keep the head # attention_probs has shape bsz x n_heads x N x N # input head_mask has shape [num_heads] or [num_hidden_layers x num_heads] # and head_mask is converted to shape [num_hidden_layers x batch x num_heads x seq_length x seq_length] head_mask = self.get_head_mask(head_mask, self.config.num_hidden_layers) if encoder_embeds is None: embedding_output = self.embeddings( input_ids=input_ids, position_ids=position_ids, token_type_ids=token_type_ids, inputs_embeds=inputs_embeds, past_key_values_length=past_key_values_length, ) else: embedding_output = encoder_embeds encoder_outputs = self.encoder( embedding_output, attention_mask=extended_attention_mask, head_mask=head_mask, encoder_hidden_states=encoder_hidden_states, encoder_attention_mask=encoder_extended_attention_mask, past_key_values=past_key_values, use_cache=use_cache, output_attentions=output_attentions, output_hidden_states=output_hidden_states, return_dict=return_dict, mode=mode, ) sequence_output = encoder_outputs[0] pooled_output = self.pooler(sequence_output) if self.pooler is not None else None if not return_dict: return (sequence_output, pooled_output) + encoder_outputs[1:] return BaseModelOutputWithPoolingAndCrossAttentions( last_hidden_state=sequence_output, pooler_output=pooled_output, past_key_values=encoder_outputs.past_key_values, hidden_states=encoder_outputs.hidden_states, attentions=encoder_outputs.attentions, cross_attentions=encoder_outputs.cross_attentions, ) @add_start_docstrings( """ Bert Model with two heads on top as done during the pretraining: a `masked language modeling` head and a `next sentence prediction (classification)` head. """, BERT_START_DOCSTRING, ) class BertForPreTraining(BertPreTrainedModel): def __init__(self, config): super().__init__(config) self.bert = BertModel(config) self.cls = BertPreTrainingHeads(config) self.init_weights() def get_output_embeddings(self): return self.cls.predictions.decoder def set_output_embeddings(self, new_embeddings): self.cls.predictions.decoder = new_embeddings @add_start_docstrings_to_model_forward(BERT_INPUTS_DOCSTRING.format("batch_size, sequence_length")) @replace_return_docstrings(output_type=BertForPreTrainingOutput, config_class=_CONFIG_FOR_DOC) def forward( self, input_ids=None, attention_mask=None, token_type_ids=None, position_ids=None, head_mask=None, inputs_embeds=None, labels=None, next_sentence_label=None, output_attentions=None, output_hidden_states=None, return_dict=None, ): r""" labels (:obj:`torch.LongTensor` of shape ``(batch_size, sequence_length)``, `optional`): Labels for computing the masked language modeling loss. Indices should be in ``[-100, 0, ..., config.vocab_size]`` (see ``input_ids`` docstring) Tokens with indices set to ``-100`` are ignored (masked), the loss is only computed for the tokens with labels in ``[0, ..., config.vocab_size]`` next_sentence_label (``torch.LongTensor`` of shape ``(batch_size,)``, `optional`): Labels for computing the next sequence prediction (classification) loss. Input should be a sequence pair (see :obj:`input_ids` docstring) Indices should be in ``[0, 1]``: - 0 indicates sequence B is a continuation of sequence A, - 1 indicates sequence B is a random sequence. kwargs (:obj:`Dict[str, any]`, optional, defaults to `{}`): Used to hide legacy arguments that have been deprecated. Returns: Example:: >>> from transformers import BertTokenizer, BertForPreTraining >>> import torch >>> tokenizer = BertTokenizer.from_pretrained('bert-base-uncased') >>> model = BertForPreTraining.from_pretrained('bert-base-uncased') >>> inputs = tokenizer("Hello, my dog is cute", return_tensors="pt") >>> outputs = model(**inputs) >>> prediction_logits = outputs.prediction_logits >>> seq_relationship_logits = outputs.seq_relationship_logits """ return_dict = return_dict if return_dict is not None else self.config.use_return_dict outputs = self.bert( input_ids, attention_mask=attention_mask, token_type_ids=token_type_ids, position_ids=position_ids, head_mask=head_mask, inputs_embeds=inputs_embeds, output_attentions=output_attentions, output_hidden_states=output_hidden_states, return_dict=return_dict, ) sequence_output, pooled_output = outputs[:2] prediction_scores, seq_relationship_score = self.cls(sequence_output, pooled_output) total_loss = None if labels is not None and next_sentence_label is not None: loss_fct = CrossEntropyLoss() masked_lm_loss = loss_fct(prediction_scores.view(-1, self.config.vocab_size), labels.view(-1)) next_sentence_loss = loss_fct(seq_relationship_score.view(-1, 2), next_sentence_label.view(-1)) total_loss = masked_lm_loss + next_sentence_loss if not return_dict: output = (prediction_scores, seq_relationship_score) + outputs[2:] return ((total_loss,) + output) if total_loss is not None else output return BertForPreTrainingOutput( loss=total_loss, prediction_logits=prediction_scores, seq_relationship_logits=seq_relationship_score, hidden_states=outputs.hidden_states, attentions=outputs.attentions, ) @add_start_docstrings( """Bert Model with a `language modeling` head on top for CLM fine-tuning. """, BERT_START_DOCSTRING ) class BertLMHeadModel(BertPreTrainedModel): _keys_to_ignore_on_load_unexpected = [r"pooler"] _keys_to_ignore_on_load_missing = [r"position_ids", r"predictions.decoder.bias"] def __init__(self, config): super().__init__(config) self.bert = BertModel(config, add_pooling_layer=False) self.cls = BertOnlyMLMHead(config) self.init_weights() def get_output_embeddings(self): return self.cls.predictions.decoder def set_output_embeddings(self, new_embeddings): self.cls.predictions.decoder = new_embeddings @add_start_docstrings_to_model_forward(BERT_INPUTS_DOCSTRING.format("batch_size, sequence_length")) @replace_return_docstrings(output_type=CausalLMOutputWithCrossAttentions, config_class=_CONFIG_FOR_DOC) def forward( self, input_ids=None, attention_mask=None, token_type_ids=None, position_ids=None, head_mask=None, inputs_embeds=None, encoder_hidden_states=None, encoder_attention_mask=None, labels=None, past_key_values=None, use_cache=None, output_attentions=None, output_hidden_states=None, return_dict=None, is_decoder=True, reduction='mean', mode='multi_modal', soft_labels=None, alpha=0, return_logits=False, ): r""" encoder_hidden_states (:obj:`torch.FloatTensor` of shape :obj:`(batch_size, sequence_length, hidden_size)`, `optional`): Sequence of hidden-states at the output of the last layer of the encoder. Used in the cross-attention if the model is configured as a decoder. encoder_attention_mask (:obj:`torch.FloatTensor` of shape :obj:`(batch_size, sequence_length)`, `optional`): Mask to avoid performing attention on the padding token indices of the encoder input. This mask is used in the cross-attention if the model is configured as a decoder. Mask values selected in ``[0, 1]``: - 1 for tokens that are **not masked**, - 0 for tokens that are **masked**. labels (:obj:`torch.LongTensor` of shape :obj:`(batch_size, sequence_length)`, `optional`): Labels for computing the left-to-right language modeling loss (next word prediction). Indices should be in ``[-100, 0, ..., config.vocab_size]`` (see ``input_ids`` docstring) Tokens with indices set to ``-100`` are ignored (masked), the loss is only computed for the tokens with labels n ``[0, ..., config.vocab_size]`` past_key_values (:obj:`tuple(tuple(torch.FloatTensor))` of length :obj:`config.n_layers` with each tuple having 4 tensors of shape :obj:`(batch_size, num_heads, sequence_length - 1, embed_size_per_head)`): Contains precomputed key and value hidden states of the attention blocks. Can be used to speed up decoding. If :obj:`past_key_values` are used, the user can optionally input only the last :obj:`decoder_input_ids` (those that don't have their past key value states given to this model) of shape :obj:`(batch_size, 1)` instead of all :obj:`decoder_input_ids` of shape :obj:`(batch_size, sequence_length)`. use_cache (:obj:`bool`, `optional`): If set to :obj:`True`, :obj:`past_key_values` key value states are returned and can be used to speed up decoding (see :obj:`past_key_values`). Returns: Example:: >>> from transformers import BertTokenizer, BertLMHeadModel, BertConfig >>> import torch >>> tokenizer = BertTokenizer.from_pretrained('bert-base-cased') >>> config = BertConfig.from_pretrained("bert-base-cased") >>> model = BertLMHeadModel.from_pretrained('bert-base-cased', config=config) >>> inputs = tokenizer("Hello, my dog is cute", return_tensors="pt") >>> outputs = model(**inputs) >>> prediction_logits = outputs.logits """ return_dict = return_dict if return_dict is not None else self.config.use_return_dict if labels is not None: use_cache = False outputs = self.bert( input_ids, attention_mask=attention_mask, token_type_ids=token_type_ids, position_ids=position_ids, head_mask=head_mask, inputs_embeds=inputs_embeds, encoder_hidden_states=encoder_hidden_states, encoder_attention_mask=encoder_attention_mask, past_key_values=past_key_values, use_cache=use_cache, output_attentions=output_attentions, output_hidden_states=output_hidden_states, return_dict=return_dict, is_decoder=is_decoder, mode=mode, ) sequence_output = outputs[0] prediction_scores = self.cls(sequence_output) if return_logits: return prediction_scores[:, :-1, :].contiguous() lm_loss = None if labels is not None: # we are doing next-token prediction; shift prediction scores and input ids by one shifted_prediction_scores = prediction_scores[:, :-1, :].contiguous() labels = labels[:, 1:].contiguous() loss_fct = CrossEntropyLoss(reduction=reduction) lm_loss = loss_fct(shifted_prediction_scores.view(-1, self.config.vocab_size), labels.view(-1)) lm_loss = lm_loss.view(prediction_scores.size(0),-1).sum(1) if soft_labels is not None: loss_distill = -torch.sum(F.log_softmax(shifted_prediction_scores, dim=-1)*soft_labels,dim=-1) loss_distill = (loss_distill * (labels!=-100)).sum(1) lm_loss = (1-alpha)*lm_loss + alpha*loss_distill if not return_dict: output = (prediction_scores,) + outputs[2:] return ((lm_loss,) + output) if lm_loss is not None else output return CausalLMOutputWithCrossAttentions( loss=lm_loss, logits=prediction_scores, past_key_values=outputs.past_key_values, hidden_states=outputs.hidden_states, attentions=outputs.attentions, cross_attentions=outputs.cross_attentions, ) def prepare_inputs_for_generation(self, input_ids, past=None, attention_mask=None, **model_kwargs): input_shape = input_ids.shape # if model is used as a decoder in encoder-decoder model, the decoder attention mask is created on the fly if attention_mask is None: attention_mask = input_ids.new_ones(input_shape) # cut decoder_input_ids if past is used if past is not None: input_ids = input_ids[:, -1:] return { "input_ids": input_ids, "attention_mask": attention_mask, "past_key_values": past, "encoder_hidden_states": model_kwargs.get("encoder_hidden_states", None), "encoder_attention_mask": model_kwargs.get("encoder_attention_mask", None), "is_decoder": True, } def _reorder_cache(self, past, beam_idx): reordered_past = () for layer_past in past: reordered_past += (tuple(past_state.index_select(0, beam_idx) for past_state in layer_past),) return reordered_past @add_start_docstrings("""Bert Model with a `language modeling` head on top. """, BERT_START_DOCSTRING) class BertForMaskedLM(BertPreTrainedModel): _keys_to_ignore_on_load_unexpected = [r"pooler"] _keys_to_ignore_on_load_missing = [r"position_ids", r"predictions.decoder.bias"] def __init__(self, config): super().__init__(config) self.bert = BertModel(config, add_pooling_layer=False) self.cls = BertOnlyMLMHead(config) self.init_weights() def get_output_embeddings(self): return self.cls.predictions.decoder def set_output_embeddings(self, new_embeddings): self.cls.predictions.decoder = new_embeddings @add_start_docstrings_to_model_forward(BERT_INPUTS_DOCSTRING.format("batch_size, sequence_length")) @add_code_sample_docstrings( tokenizer_class=_TOKENIZER_FOR_DOC, checkpoint="bert-base-uncased", output_type=MaskedLMOutput, config_class=_CONFIG_FOR_DOC, ) def forward( self, input_ids=None, attention_mask=None, token_type_ids=None, position_ids=None, head_mask=None, inputs_embeds=None, encoder_embeds=None, encoder_hidden_states=None, encoder_attention_mask=None, labels=None, output_attentions=None, output_hidden_states=None, return_dict=None, is_decoder=False, mode='multi_modal', soft_labels=None, alpha=0, return_logits=False, ): r""" labels (:obj:`torch.LongTensor` of shape :obj:`(batch_size, sequence_length)`, `optional`): Labels for computing the masked language modeling loss. Indices should be in ``[-100, 0, ..., config.vocab_size]`` (see ``input_ids`` docstring) Tokens with indices set to ``-100`` are ignored (masked), the loss is only computed for the tokens with labels in ``[0, ..., config.vocab_size]`` """ return_dict = return_dict if return_dict is not None else self.config.use_return_dict outputs = self.bert( input_ids, attention_mask=attention_mask, token_type_ids=token_type_ids, position_ids=position_ids, head_mask=head_mask, inputs_embeds=inputs_embeds, encoder_embeds=encoder_embeds, encoder_hidden_states=encoder_hidden_states, encoder_attention_mask=encoder_attention_mask, output_attentions=output_attentions, output_hidden_states=output_hidden_states, return_dict=return_dict, is_decoder=is_decoder, mode=mode, ) sequence_output = outputs[0] prediction_scores = self.cls(sequence_output) if return_logits: return prediction_scores masked_lm_loss = None if labels is not None: loss_fct = CrossEntropyLoss() # -100 index = padding token masked_lm_loss = loss_fct(prediction_scores.view(-1, self.config.vocab_size), labels.view(-1)) if soft_labels is not None: loss_distill = -torch.sum(F.log_softmax(prediction_scores, dim=-1)*soft_labels,dim=-1) loss_distill = loss_distill[labels!=-100].mean() masked_lm_loss = (1-alpha)*masked_lm_loss + alpha*loss_distill if not return_dict: output = (prediction_scores,) + outputs[2:] return ((masked_lm_loss,) + output) if masked_lm_loss is not None else output return MaskedLMOutput( loss=masked_lm_loss, logits=prediction_scores, hidden_states=outputs.hidden_states, attentions=outputs.attentions, ) def prepare_inputs_for_generation(self, input_ids, attention_mask=None, **model_kwargs): input_shape = input_ids.shape effective_batch_size = input_shape[0] # add a dummy token assert self.config.pad_token_id is not None, "The PAD token should be defined for generation" attention_mask = torch.cat([attention_mask, attention_mask.new_zeros((attention_mask.shape[0], 1))], dim=-1) dummy_token = torch.full( (effective_batch_size, 1), self.config.pad_token_id, dtype=torch.long, device=input_ids.device ) input_ids = torch.cat([input_ids, dummy_token], dim=1) return {"input_ids": input_ids, "attention_mask": attention_mask} @add_start_docstrings( """Bert Model with a `next sentence prediction (classification)` head on top. """, BERT_START_DOCSTRING, ) class BertForNextSentencePrediction(BertPreTrainedModel): def __init__(self, config): super().__init__(config) self.bert = BertModel(config) self.cls = BertOnlyNSPHead(config) self.init_weights() @add_start_docstrings_to_model_forward(BERT_INPUTS_DOCSTRING.format("batch_size, sequence_length")) @replace_return_docstrings(output_type=NextSentencePredictorOutput, config_class=_CONFIG_FOR_DOC) def forward( self, input_ids=None, attention_mask=None, token_type_ids=None, position_ids=None, head_mask=None, inputs_embeds=None, labels=None, output_attentions=None, output_hidden_states=None, return_dict=None, **kwargs ): r""" labels (:obj:`torch.LongTensor` of shape :obj:`(batch_size,)`, `optional`): Labels for computing the next sequence prediction (classification) loss. Input should be a sequence pair (see ``input_ids`` docstring). Indices should be in ``[0, 1]``: - 0 indicates sequence B is a continuation of sequence A, - 1 indicates sequence B is a random sequence. Returns: Example:: >>> from transformers import BertTokenizer, BertForNextSentencePrediction >>> import torch >>> tokenizer = BertTokenizer.from_pretrained('bert-base-uncased') >>> model = BertForNextSentencePrediction.from_pretrained('bert-base-uncased') >>> prompt = "In Italy, pizza served in formal settings, such as at a restaurant, is presented unsliced." >>> next_sentence = "The sky is blue due to the shorter wavelength of blue light." >>> encoding = tokenizer(prompt, next_sentence, return_tensors='pt') >>> outputs = model(**encoding, labels=torch.LongTensor([1])) >>> logits = outputs.logits >>> assert logits[0, 0] < logits[0, 1] # next sentence was random """ if "next_sentence_label" in kwargs: warnings.warn( "The `next_sentence_label` argument is deprecated and will be removed in a future version, use `labels` instead.", FutureWarning, ) labels = kwargs.pop("next_sentence_label") return_dict = return_dict if return_dict is not None else self.config.use_return_dict outputs = self.bert( input_ids, attention_mask=attention_mask, token_type_ids=token_type_ids, position_ids=position_ids, head_mask=head_mask, inputs_embeds=inputs_embeds, output_attentions=output_attentions, output_hidden_states=output_hidden_states, return_dict=return_dict, ) pooled_output = outputs[1] seq_relationship_scores = self.cls(pooled_output) next_sentence_loss = None if labels is not None: loss_fct = CrossEntropyLoss() next_sentence_loss = loss_fct(seq_relationship_scores.view(-1, 2), labels.view(-1)) if not return_dict: output = (seq_relationship_scores,) + outputs[2:] return ((next_sentence_loss,) + output) if next_sentence_loss is not None else output return NextSentencePredictorOutput( loss=next_sentence_loss, logits=seq_relationship_scores, hidden_states=outputs.hidden_states, attentions=outputs.attentions, ) @add_start_docstrings( """ Bert Model transformer with a sequence classification/regression head on top (a linear layer on top of the pooled output) e.g. for GLUE tasks. """, BERT_START_DOCSTRING, ) class BertForSequenceClassification(BertPreTrainedModel): def __init__(self, config): super().__init__(config) self.num_labels = config.num_labels self.bert = BertModel(config) self.dropout = nn.Dropout(config.hidden_dropout_prob) self.classifier = nn.Linear(config.hidden_size, config.num_labels) self.init_weights() @add_start_docstrings_to_model_forward(BERT_INPUTS_DOCSTRING.format("batch_size, sequence_length")) @add_code_sample_docstrings( tokenizer_class=_TOKENIZER_FOR_DOC, checkpoint="bert-base-uncased", output_type=SequenceClassifierOutput, config_class=_CONFIG_FOR_DOC, ) def forward( self, input_ids=None, attention_mask=None, token_type_ids=None, position_ids=None, head_mask=None, inputs_embeds=None, labels=None, output_attentions=None, output_hidden_states=None, return_dict=None, ): r""" labels (:obj:`torch.LongTensor` of shape :obj:`(batch_size,)`, `optional`): Labels for computing the sequence classification/regression loss. Indices should be in :obj:`[0, ..., config.num_labels - 1]`. If :obj:`config.num_labels == 1` a regression loss is computed (Mean-Square loss), If :obj:`config.num_labels > 1` a classification loss is computed (Cross-Entropy). """ return_dict = return_dict if return_dict is not None else self.config.use_return_dict outputs = self.bert( input_ids, attention_mask=attention_mask, token_type_ids=token_type_ids, position_ids=position_ids, head_mask=head_mask, inputs_embeds=inputs_embeds, output_attentions=output_attentions, output_hidden_states=output_hidden_states, return_dict=return_dict, ) pooled_output = outputs[1] pooled_output = self.dropout(pooled_output) logits = self.classifier(pooled_output) loss = None if labels is not None: if self.num_labels == 1: # We are doing regression loss_fct = MSELoss() loss = loss_fct(logits.view(-1), labels.view(-1)) else: loss_fct = CrossEntropyLoss() loss = loss_fct(logits.view(-1, self.num_labels), labels.view(-1)) if not return_dict: output = (logits,) + outputs[2:] return ((loss,) + output) if loss is not None else output return SequenceClassifierOutput( loss=loss, logits=logits, hidden_states=outputs.hidden_states, attentions=outputs.attentions, ) @add_start_docstrings( """ Bert Model with a multiple choice classification head on top (a linear layer on top of the pooled output and a softmax) e.g. for RocStories/SWAG tasks. """, BERT_START_DOCSTRING, ) class BertForMultipleChoice(BertPreTrainedModel): def __init__(self, config): super().__init__(config) self.bert = BertModel(config) self.dropout = nn.Dropout(config.hidden_dropout_prob) self.classifier = nn.Linear(config.hidden_size, 1) self.init_weights() @add_start_docstrings_to_model_forward(BERT_INPUTS_DOCSTRING.format("batch_size, num_choices, sequence_length")) @add_code_sample_docstrings( tokenizer_class=_TOKENIZER_FOR_DOC, checkpoint="bert-base-uncased", output_type=MultipleChoiceModelOutput, config_class=_CONFIG_FOR_DOC, ) def forward( self, input_ids=None, attention_mask=None, token_type_ids=None, position_ids=None, head_mask=None, inputs_embeds=None, labels=None, output_attentions=None, output_hidden_states=None, return_dict=None, ): r""" labels (:obj:`torch.LongTensor` of shape :obj:`(batch_size,)`, `optional`): Labels for computing the multiple choice classification loss. Indices should be in ``[0, ..., num_choices-1]`` where :obj:`num_choices` is the size of the second dimension of the input tensors. (See :obj:`input_ids` above) """ return_dict = return_dict if return_dict is not None else self.config.use_return_dict num_choices = input_ids.shape[1] if input_ids is not None else inputs_embeds.shape[1] input_ids = input_ids.view(-1, input_ids.size(-1)) if input_ids is not None else None attention_mask = attention_mask.view(-1, attention_mask.size(-1)) if attention_mask is not None else None token_type_ids = token_type_ids.view(-1, token_type_ids.size(-1)) if token_type_ids is not None else None position_ids = position_ids.view(-1, position_ids.size(-1)) if position_ids is not None else None inputs_embeds = ( inputs_embeds.view(-1, inputs_embeds.size(-2), inputs_embeds.size(-1)) if inputs_embeds is not None else None ) outputs = self.bert( input_ids, attention_mask=attention_mask, token_type_ids=token_type_ids, position_ids=position_ids, head_mask=head_mask, inputs_embeds=inputs_embeds, output_attentions=output_attentions, output_hidden_states=output_hidden_states, return_dict=return_dict, ) pooled_output = outputs[1] pooled_output = self.dropout(pooled_output) logits = self.classifier(pooled_output) reshaped_logits = logits.view(-1, num_choices) loss = None if labels is not None: loss_fct = CrossEntropyLoss() loss = loss_fct(reshaped_logits, labels) if not return_dict: output = (reshaped_logits,) + outputs[2:] return ((loss,) + output) if loss is not None else output return MultipleChoiceModelOutput( loss=loss, logits=reshaped_logits, hidden_states=outputs.hidden_states, attentions=outputs.attentions, ) @add_start_docstrings( """ Bert Model with a token classification head on top (a linear layer on top of the hidden-states output) e.g. for Named-Entity-Recognition (NER) tasks. """, BERT_START_DOCSTRING, ) class BertForTokenClassification(BertPreTrainedModel): _keys_to_ignore_on_load_unexpected = [r"pooler"] def __init__(self, config): super().__init__(config) self.num_labels = config.num_labels self.bert = BertModel(config, add_pooling_layer=False) self.dropout = nn.Dropout(config.hidden_dropout_prob) self.classifier = nn.Linear(config.hidden_size, config.num_labels) self.init_weights() @add_start_docstrings_to_model_forward(BERT_INPUTS_DOCSTRING.format("batch_size, sequence_length")) @add_code_sample_docstrings( tokenizer_class=_TOKENIZER_FOR_DOC, checkpoint="bert-base-uncased", output_type=TokenClassifierOutput, config_class=_CONFIG_FOR_DOC, ) def forward( self, input_ids=None, attention_mask=None, token_type_ids=None, position_ids=None, head_mask=None, inputs_embeds=None, labels=None, output_attentions=None, output_hidden_states=None, return_dict=None, ): r""" labels (:obj:`torch.LongTensor` of shape :obj:`(batch_size, sequence_length)`, `optional`): Labels for computing the token classification loss. Indices should be in ``[0, ..., config.num_labels - 1]``. """ return_dict = return_dict if return_dict is not None else self.config.use_return_dict outputs = self.bert( input_ids, attention_mask=attention_mask, token_type_ids=token_type_ids, position_ids=position_ids, head_mask=head_mask, inputs_embeds=inputs_embeds, output_attentions=output_attentions, output_hidden_states=output_hidden_states, return_dict=return_dict, ) sequence_output = outputs[0] sequence_output = self.dropout(sequence_output) logits = self.classifier(sequence_output) loss = None if labels is not None: loss_fct = CrossEntropyLoss() # Only keep active parts of the loss if attention_mask is not None: active_loss = attention_mask.view(-1) == 1 active_logits = logits.view(-1, self.num_labels) active_labels = torch.where( active_loss, labels.view(-1), torch.tensor(loss_fct.ignore_index).type_as(labels) ) loss = loss_fct(active_logits, active_labels) else: loss = loss_fct(logits.view(-1, self.num_labels), labels.view(-1)) if not return_dict: output = (logits,) + outputs[2:] return ((loss,) + output) if loss is not None else output return TokenClassifierOutput( loss=loss, logits=logits, hidden_states=outputs.hidden_states, attentions=outputs.attentions, ) @add_start_docstrings( """ Bert Model with a span classification head on top for extractive question-answering tasks like SQuAD (a linear layers on top of the hidden-states output to compute `span start logits` and `span end logits`). """, BERT_START_DOCSTRING, ) class BertForQuestionAnswering(BertPreTrainedModel): _keys_to_ignore_on_load_unexpected = [r"pooler"] def __init__(self, config): super().__init__(config) self.num_labels = config.num_labels self.bert = BertModel(config, add_pooling_layer=False) self.qa_outputs = nn.Linear(config.hidden_size, config.num_labels) self.init_weights() @add_start_docstrings_to_model_forward(BERT_INPUTS_DOCSTRING.format("batch_size, sequence_length")) @add_code_sample_docstrings( tokenizer_class=_TOKENIZER_FOR_DOC, checkpoint="bert-base-uncased", output_type=QuestionAnsweringModelOutput, config_class=_CONFIG_FOR_DOC, ) def forward( self, input_ids=None, attention_mask=None, token_type_ids=None, position_ids=None, head_mask=None, inputs_embeds=None, start_positions=None, end_positions=None, output_attentions=None, output_hidden_states=None, return_dict=None, ): r""" start_positions (:obj:`torch.LongTensor` of shape :obj:`(batch_size,)`, `optional`): Labels for position (index) of the start of the labelled span for computing the token classification loss. Positions are clamped to the length of the sequence (:obj:`sequence_length`). Position outside of the sequence are not taken into account for computing the loss. end_positions (:obj:`torch.LongTensor` of shape :obj:`(batch_size,)`, `optional`): Labels for position (index) of the end of the labelled span for computing the token classification loss. Positions are clamped to the length of the sequence (:obj:`sequence_length`). Position outside of the sequence are not taken into account for computing the loss. """ return_dict = return_dict if return_dict is not None else self.config.use_return_dict outputs = self.bert( input_ids, attention_mask=attention_mask, token_type_ids=token_type_ids, position_ids=position_ids, head_mask=head_mask, inputs_embeds=inputs_embeds, output_attentions=output_attentions, output_hidden_states=output_hidden_states, return_dict=return_dict, ) sequence_output = outputs[0] logits = self.qa_outputs(sequence_output) start_logits, end_logits = logits.split(1, dim=-1) start_logits = start_logits.squeeze(-1) end_logits = end_logits.squeeze(-1) total_loss = None if start_positions is not None and end_positions is not None: # If we are on multi-GPU, split add a dimension if len(start_positions.size()) > 1: start_positions = start_positions.squeeze(-1) if len(end_positions.size()) > 1: end_positions = end_positions.squeeze(-1) # sometimes the start/end positions are outside our model inputs, we ignore these terms ignored_index = start_logits.size(1) start_positions.clamp_(0, ignored_index) end_positions.clamp_(0, ignored_index) loss_fct = CrossEntropyLoss(ignore_index=ignored_index) start_loss = loss_fct(start_logits, start_positions) end_loss = loss_fct(end_logits, end_positions) total_loss = (start_loss + end_loss) / 2 if not return_dict: output = (start_logits, end_logits) + outputs[2:] return ((total_loss,) + output) if total_loss is not None else output return QuestionAnsweringModelOutput( loss=total_loss, start_logits=start_logits, end_logits=end_logits, hidden_states=outputs.hidden_states, attentions=outputs.attentions, )
82,187
41.873239
213
py
Mr.Right
Mr.Right-main/models/ALBEF/models/vit.py
import torch import torch.nn as nn import torch.nn.functional as F from functools import partial from timm.models.vision_transformer import _cfg, PatchEmbed from timm.models.registry import register_model from timm.models.layers import trunc_normal_, DropPath class Mlp(nn.Module): """ MLP as used in Vision Transformer, MLP-Mixer and related networks """ def __init__(self, in_features, hidden_features=None, out_features=None, act_layer=nn.GELU, drop=0.): super().__init__() out_features = out_features or in_features hidden_features = hidden_features or in_features self.fc1 = nn.Linear(in_features, hidden_features) self.act = act_layer() self.fc2 = nn.Linear(hidden_features, out_features) self.drop = nn.Dropout(drop) def forward(self, x): x = self.fc1(x) x = self.act(x) x = self.drop(x) x = self.fc2(x) x = self.drop(x) return x class Attention(nn.Module): def __init__(self, dim, num_heads=8, qkv_bias=False, qk_scale=None, attn_drop=0., proj_drop=0.): super().__init__() self.num_heads = num_heads head_dim = dim // num_heads # NOTE scale factor was wrong in my original version, can set manually to be compat with prev weights self.scale = qk_scale or head_dim ** -0.5 self.qkv = nn.Linear(dim, dim * 3, bias=qkv_bias) self.attn_drop = nn.Dropout(attn_drop) self.proj = nn.Linear(dim, dim) self.proj_drop = nn.Dropout(proj_drop) self.attn_gradients = None self.attention_map = None def save_attn_gradients(self, attn_gradients): self.attn_gradients = attn_gradients def get_attn_gradients(self): return self.attn_gradients def save_attention_map(self, attention_map): self.attention_map = attention_map def get_attention_map(self): return self.attention_map def forward(self, x, register_hook=False): B, N, C = x.shape qkv = self.qkv(x).reshape(B, N, 3, self.num_heads, C // self.num_heads).permute(2, 0, 3, 1, 4) q, k, v = qkv[0], qkv[1], qkv[2] # make torchscript happy (cannot use tensor as tuple) attn = (q @ k.transpose(-2, -1)) * self.scale attn = attn.softmax(dim=-1) attn = self.attn_drop(attn) if register_hook: self.save_attention_map(attn) attn.register_hook(self.save_attn_gradients) x = (attn @ v).transpose(1, 2).reshape(B, N, C) x = self.proj(x) x = self.proj_drop(x) return x class Block(nn.Module): def __init__(self, dim, num_heads, mlp_ratio=4., qkv_bias=False, qk_scale=None, drop=0., attn_drop=0., drop_path=0., act_layer=nn.GELU, norm_layer=nn.LayerNorm): super().__init__() self.norm1 = norm_layer(dim) self.attn = Attention( dim, num_heads=num_heads, qkv_bias=qkv_bias, qk_scale=qk_scale, attn_drop=attn_drop, proj_drop=drop) # NOTE: drop path for stochastic depth, we shall see if this is better than dropout here self.drop_path = DropPath(drop_path) if drop_path > 0. else nn.Identity() self.norm2 = norm_layer(dim) mlp_hidden_dim = int(dim * mlp_ratio) self.mlp = Mlp(in_features=dim, hidden_features=mlp_hidden_dim, act_layer=act_layer, drop=drop) def forward(self, x, register_hook=False): x = x + self.drop_path(self.attn(self.norm1(x), register_hook=register_hook)) x = x + self.drop_path(self.mlp(self.norm2(x))) return x class VisionTransformer(nn.Module): """ Vision Transformer A PyTorch impl of : `An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale` - https://arxiv.org/abs/2010.11929 """ def __init__(self, img_size=224, patch_size=16, in_chans=3, num_classes=1000, embed_dim=768, depth=12, num_heads=12, mlp_ratio=4., qkv_bias=True, qk_scale=None, representation_size=None, drop_rate=0., attn_drop_rate=0., drop_path_rate=0., norm_layer=None): """ Args: img_size (int, tuple): input image size patch_size (int, tuple): patch size in_chans (int): number of input channels num_classes (int): number of classes for classification head embed_dim (int): embedding dimension depth (int): depth of transformer num_heads (int): number of attention heads mlp_ratio (int): ratio of mlp hidden dim to embedding dim qkv_bias (bool): enable bias for qkv if True qk_scale (float): override default qk scale of head_dim ** -0.5 if set representation_size (Optional[int]): enable and set representation layer (pre-logits) to this value if set drop_rate (float): dropout rate attn_drop_rate (float): attention dropout rate drop_path_rate (float): stochastic depth rate norm_layer: (nn.Module): normalization layer """ super().__init__() self.num_features = self.embed_dim = embed_dim # num_features for consistency with other models norm_layer = norm_layer or partial(nn.LayerNorm, eps=1e-6) self.patch_embed = PatchEmbed( img_size=img_size, patch_size=patch_size, in_chans=in_chans, embed_dim=embed_dim) num_patches = self.patch_embed.num_patches self.cls_token = nn.Parameter(torch.zeros(1, 1, embed_dim)) self.pos_embed = nn.Parameter(torch.zeros(1, num_patches + 1, embed_dim)) self.pos_drop = nn.Dropout(p=drop_rate) dpr = [x.item() for x in torch.linspace(0, drop_path_rate, depth)] # stochastic depth decay rule self.blocks = nn.ModuleList([ Block( dim=embed_dim, num_heads=num_heads, mlp_ratio=mlp_ratio, qkv_bias=qkv_bias, qk_scale=qk_scale, drop=drop_rate, attn_drop=attn_drop_rate, drop_path=dpr[i], norm_layer=norm_layer) for i in range(depth)]) self.norm = norm_layer(embed_dim) trunc_normal_(self.pos_embed, std=.02) trunc_normal_(self.cls_token, std=.02) self.apply(self._init_weights) def _init_weights(self, m): if isinstance(m, nn.Linear): trunc_normal_(m.weight, std=.02) if isinstance(m, nn.Linear) and m.bias is not None: nn.init.constant_(m.bias, 0) elif isinstance(m, nn.LayerNorm): nn.init.constant_(m.bias, 0) nn.init.constant_(m.weight, 1.0) @torch.jit.ignore def no_weight_decay(self): return {'pos_embed', 'cls_token'} def forward(self, x, register_blk=-1): B = x.shape[0] x = self.patch_embed(x) cls_tokens = self.cls_token.expand(B, -1, -1) # stole cls_tokens impl from Phil Wang, thanks x = torch.cat((cls_tokens, x), dim=1) x = x + self.pos_embed[:,:x.size(1),:] x = self.pos_drop(x) for i,blk in enumerate(self.blocks): x = blk(x, register_blk==i) x = self.norm(x) return x def interpolate_pos_embed(pos_embed_checkpoint, visual_encoder): # interpolate position embedding embedding_size = pos_embed_checkpoint.shape[-1] num_patches = visual_encoder.patch_embed.num_patches num_extra_tokens = visual_encoder.pos_embed.shape[-2] - num_patches # height (== width) for the checkpoint position embedding orig_size = int((pos_embed_checkpoint.shape[-2] - num_extra_tokens) ** 0.5) # height (== width) for the new position embedding new_size = int(num_patches ** 0.5) if orig_size!=new_size: # class_token and dist_token are kept unchanged extra_tokens = pos_embed_checkpoint[:, :num_extra_tokens] # only the position tokens are interpolated pos_tokens = pos_embed_checkpoint[:, num_extra_tokens:] pos_tokens = pos_tokens.reshape(-1, orig_size, orig_size, embedding_size).permute(0, 3, 1, 2) pos_tokens = torch.nn.functional.interpolate( pos_tokens, size=(new_size, new_size), mode='bicubic', align_corners=False) pos_tokens = pos_tokens.permute(0, 2, 3, 1).flatten(1, 2) new_pos_embed = torch.cat((extra_tokens, pos_tokens), dim=1) print('reshape position embedding from %d to %d'%(orig_size ** 2,new_size ** 2)) return new_pos_embed else: return pos_embed_checkpoint
8,558
41.162562
118
py
Mr.Right
Mr.Right-main/models/ALBEF/models/model_retrieval.py
from functools import partial from models.ALBEF.models.vit import VisionTransformer from models.ALBEF.models.xbert import BertConfig, BertModel import torch from torch import nn import torch.nn.functional as F class ALBEF(nn.Module): def __init__(self, text_encoder = None, tokenizer = None, config = None, ): super().__init__() self.tokenizer = tokenizer embed_dim = config['embed_dim'] vision_width = config['vision_width'] self.visual_encoder = VisionTransformer( img_size=config['image_res'], patch_size=16, embed_dim=768, depth=12, num_heads=12, mlp_ratio=4, qkv_bias=True, norm_layer=partial(nn.LayerNorm, eps=1e-6)) self.bert_config = BertConfig.from_json_file(config['bert_config']) self.text_encoder = BertModel.from_pretrained(text_encoder, config=self.bert_config, add_pooling_layer=False) text_width = self.text_encoder.config.hidden_size self.vision_proj = nn.Linear(vision_width, embed_dim) self.text_proj = nn.Linear(text_width, embed_dim) def forward(self, doc_image, doc_text, query): output = dict() query_output = self.text_encoder(query["input_ids"], attention_mask=query["attention_mask"], return_dict=True, mode='text') query_embeds = query_output.last_hidden_state query_feat = F.normalize(self.text_proj(query_embeds[:,0,:]),dim=-1) image_embeds = self.visual_encoder(doc_image) image_atts = torch.ones(image_embeds.size()[:-1],dtype=torch.long).to(doc_image.device) image_feat = F.normalize(self.vision_proj(image_embeds[:,0,:]),dim=-1) text_output = self.text_encoder(doc_text["input_ids"], attention_mask = doc_text["attention_mask"], return_dict = True, mode = 'text') text_embeds = text_output.last_hidden_state text_feat = F.normalize(self.text_proj(text_embeds[:,0,:]),dim=-1) multi_output = self.text_encoder(encoder_embeds = text_embeds, attention_mask = doc_text["attention_mask"], encoder_hidden_states = image_embeds, encoder_attention_mask = image_atts, return_dict = True, mode = 'fusion', ) multi_embeds = multi_output.last_hidden_state multi_atts = torch.ones(multi_embeds.size()[:-1],dtype=torch.long).to(doc_image.device) multi_feat = F.normalize(self.text_proj(multi_embeds[:,0,:]),dim=-1) output['query_embeds'] = query_embeds output['query_cls'] = query_feat output['query_atts'] = query["attention_mask"] output['doctext_embeds'] = text_embeds output['doctext_cls'] = text_feat output['img_embeds'] = image_embeds output['img_cls'] = image_feat output['multi_embeds'] = multi_embeds output['multi_cls'] = multi_feat output['multi_atts'] = doc_text["attention_mask"] return output
3,499
45.666667
129
py
Mr.Right
Mr.Right-main/models/ViLT/vilt/modules/vilt_utils.py
import torch import random from transformers.optimization import AdamW from transformers import ( get_polynomial_decay_schedule_with_warmup, get_cosine_schedule_with_warmup, ) from models.ViLT.vilt.modules.dist_utils import all_gather from models.ViLT.vilt.modules.objectives import compute_irtr_recall from models.ViLT.vilt.gadgets.my_metrics import Accuracy, VQAScore, Scalar def set_metrics(pl_module): for split in ["train", "val"]: for k, v in pl_module.hparams.config["loss_names"].items(): if v < 1: continue if k == "vqa": setattr(pl_module, f"{split}_vqa_score", VQAScore()) setattr(pl_module, f"{split}_{k}_loss", Scalar()) elif k == "nlvr2": if split == "train": setattr(pl_module, f"train_{k}_accuracy", Accuracy()) setattr(pl_module, f"train_{k}_loss", Scalar()) else: setattr(pl_module, f"dev_{k}_accuracy", Accuracy()) setattr(pl_module, f"dev_{k}_loss", Scalar()) setattr(pl_module, f"test_{k}_accuracy", Accuracy()) setattr(pl_module, f"test_{k}_loss", Scalar()) elif k == "irtr": setattr(pl_module, f"{split}_irtr_loss", Scalar()) elif k == "mppd" or k == "mpfr": setattr(pl_module, f"{split}_{k}_loss", Scalar()) elif k == "itm": setattr(pl_module, f"{split}_{k}_accuracy", Accuracy()) setattr(pl_module, f"{split}_{k}_loss", Scalar()) setattr(pl_module, f"{split}_{k}_wpa_loss", Scalar()) else: setattr(pl_module, f"{split}_{k}_accuracy", Accuracy()) setattr(pl_module, f"{split}_{k}_loss", Scalar()) def epoch_wrapup(pl_module): phase = "train" if pl_module.training else "val" the_metric = 0 if pl_module.hparams.config["get_recall_metric"] and not pl_module.training: (ir_r1, ir_r5, ir_r10, tr_r1, tr_r5, tr_r10) = compute_irtr_recall(pl_module) print((ir_r1, ir_r5, ir_r10, tr_r1, tr_r5, tr_r10), pl_module.global_step) pl_module.logger.experiment.add_scalar( "recalls/ir_r1", ir_r1, pl_module.global_step ) pl_module.logger.experiment.add_scalar( "recalls/ir_r5", ir_r5, pl_module.global_step ) pl_module.logger.experiment.add_scalar( "recalls/ir_r10", ir_r10, pl_module.global_step ) pl_module.logger.experiment.add_scalar( "recalls/tr_r1", tr_r1, pl_module.global_step ) pl_module.logger.experiment.add_scalar( "recalls/tr_r5", tr_r5, pl_module.global_step ) pl_module.logger.experiment.add_scalar( "recalls/tr_r10", tr_r10, pl_module.global_step ) the_metric += ir_r1.item() + tr_r1.item() for loss_name, v in pl_module.hparams.config["loss_names"].items(): if v < 1: continue value = 0 if loss_name == "vqa": value = getattr(pl_module, f"{phase}_{loss_name}_score").compute() pl_module.log(f"{loss_name}/{phase}/score_epoch", value) getattr(pl_module, f"{phase}_{loss_name}_score").reset() pl_module.log( f"{loss_name}/{phase}/loss_epoch", getattr(pl_module, f"{phase}_{loss_name}_loss").compute(), ) getattr(pl_module, f"{phase}_{loss_name}_loss").reset() elif loss_name == "nlvr2": if phase == "train": value = getattr(pl_module, f"train_{loss_name}_accuracy").compute() pl_module.log(f"{loss_name}/train/accuracy_epoch", value) getattr(pl_module, f"train_{loss_name}_accuracy").reset() pl_module.log( f"{loss_name}/train/loss_epoch", getattr(pl_module, f"train_{loss_name}_loss").compute(), ) getattr(pl_module, f"train_{loss_name}_loss").reset() else: value = getattr(pl_module, f"dev_{loss_name}_accuracy").compute() pl_module.log(f"{loss_name}/dev/accuracy_epoch", value) getattr(pl_module, f"dev_{loss_name}_accuracy").reset() pl_module.log( f"{loss_name}/dev/loss_epoch", getattr(pl_module, f"dev_{loss_name}_loss").compute(), ) getattr(pl_module, f"dev_{loss_name}_loss").reset() value = getattr(pl_module, f"test_{loss_name}_accuracy").compute() pl_module.log(f"{loss_name}/test/accuracy_epoch", value) getattr(pl_module, f"test_{loss_name}_accuracy").reset() pl_module.log( f"{loss_name}/test/loss_epoch", getattr(pl_module, f"test_{loss_name}_loss").compute(), ) getattr(pl_module, f"test_{loss_name}_loss").reset() elif loss_name == "irtr": pl_module.log( f"{loss_name}/{phase}/irtr_loss_epoch", getattr(pl_module, f"{phase}_irtr_loss").compute(), ) getattr(pl_module, f"{phase}_irtr_loss").reset() elif loss_name == "mppd" or loss_name == "mpfr": pl_module.log( f"{loss_name}/{phase}/loss_epoch", getattr(pl_module, f"{phase}_{loss_name}_loss").compute(), ) getattr(pl_module, f"{phase}_{loss_name}_loss").reset() elif loss_name == "itm": value = getattr(pl_module, f"{phase}_{loss_name}_accuracy").compute() pl_module.log(f"{loss_name}/{phase}/accuracy_epoch", value) getattr(pl_module, f"{phase}_{loss_name}_accuracy").reset() pl_module.log( f"{loss_name}/{phase}/loss_epoch", getattr(pl_module, f"{phase}_{loss_name}_loss").compute(), ) getattr(pl_module, f"{phase}_{loss_name}_loss").reset() pl_module.log( f"{loss_name}/{phase}/wpa_loss_epoch", getattr(pl_module, f"{phase}_{loss_name}_wpa_loss").compute(), ) getattr(pl_module, f"{phase}_{loss_name}_wpa_loss").reset() else: value = getattr(pl_module, f"{phase}_{loss_name}_accuracy").compute() pl_module.log(f"{loss_name}/{phase}/accuracy_epoch", value) getattr(pl_module, f"{phase}_{loss_name}_accuracy").reset() pl_module.log( f"{loss_name}/{phase}/loss_epoch", getattr(pl_module, f"{phase}_{loss_name}_loss").compute(), ) getattr(pl_module, f"{phase}_{loss_name}_loss").reset() the_metric += value pl_module.log(f"{phase}/the_metric", the_metric) def check_non_acc_grad(pl_module): if pl_module.token_type_embeddings.weight.grad is None: return True else: grad = pl_module.token_type_embeddings.weight.grad return (grad.sum() == 0).item() def set_task(pl_module): pl_module.current_tasks = [ k for k, v in pl_module.hparams.config["loss_names"].items() if v >= 1 ] return def set_schedule(pl_module): lr = pl_module.hparams.config["learning_rate"] wd = pl_module.hparams.config["weight_decay"] no_decay = [ "bias", "LayerNorm.bias", "LayerNorm.weight", "norm.bias", "norm.weight", "norm1.bias", "norm1.weight", "norm2.bias", "norm2.weight", ] head_names = ["vqa_classifier", "nlvr2_classifier"] lr_mult = pl_module.hparams.config["lr_mult"] end_lr = pl_module.hparams.config["end_lr"] decay_power = pl_module.hparams.config["decay_power"] optim_type = pl_module.hparams.config["optim_type"] names = [n for n, p in pl_module.named_parameters()] optimizer_grouped_parameters = [ { "params": [ p for n, p in pl_module.named_parameters() if not any(nd in n for nd in no_decay) and not any(bb in n for bb in head_names) ], "weight_decay": wd, "lr": lr, }, { "params": [ p for n, p in pl_module.named_parameters() if any(nd in n for nd in no_decay) and not any(bb in n for bb in head_names) ], "weight_decay": 0.0, "lr": lr, }, { "params": [ p for n, p in pl_module.named_parameters() if not any(nd in n for nd in no_decay) and any(bb in n for bb in head_names) ], "weight_decay": wd, "lr": lr * lr_mult, }, { "params": [ p for n, p in pl_module.named_parameters() if any(nd in n for nd in no_decay) and any(bb in n for bb in head_names) ], "weight_decay": 0.0, "lr": lr * lr_mult, }, ] if optim_type == "adamw": optimizer = AdamW( optimizer_grouped_parameters, lr=lr, eps=1e-8, betas=(0.9, 0.98) ) elif optim_type == "adam": optimizer = torch.optim.Adam(optimizer_grouped_parameters, lr=lr) elif optim_type == "sgd": optimizer = torch.optim.SGD(optimizer_grouped_parameters, lr=lr, momentum=0.9) if pl_module.trainer.max_steps is None: max_steps = ( len(pl_module.trainer.datamodule.train_dataloader()) * pl_module.trainer.max_epochs // pl_module.trainer.accumulate_grad_batches ) else: max_steps = pl_module.trainer.max_steps warmup_steps = pl_module.hparams.config["warmup_steps"] if isinstance(pl_module.hparams.config["warmup_steps"], float): warmup_steps = int(max_steps * warmup_steps) if decay_power == "cosine": scheduler = get_cosine_schedule_with_warmup( optimizer, num_warmup_steps=warmup_steps, num_training_steps=max_steps, ) else: scheduler = get_polynomial_decay_schedule_with_warmup( optimizer, num_warmup_steps=warmup_steps, num_training_steps=max_steps, lr_end=end_lr, power=decay_power, ) sched = {"scheduler": scheduler, "interval": "step"} return ( [optimizer], [sched], )
10,650
37.451264
88
py
Mr.Right
Mr.Right-main/models/ViLT/vilt/modules/dist_utils.py
# Copyright (c) Facebook, Inc. and its affiliates. All Rights Reserved """ This file contains primitives for multi-gpu communication. This is useful when doing distributed training. """ import functools import logging import numpy as np import pickle import torch import torch.distributed as dist import torch _LOCAL_PROCESS_GROUP = None """ A torch process group which only includes processes that on the same machine as the current process. This variable is set when processes are spawned by `launch()` in "engine/launch.py". """ def get_world_size() -> int: if not dist.is_available(): return 1 if not dist.is_initialized(): return 1 return dist.get_world_size() def get_rank() -> int: if not dist.is_available(): return 0 if not dist.is_initialized(): return 0 return dist.get_rank() def get_local_rank() -> int: """ Returns: The rank of the current process within the local (per-machine) process group. """ if not dist.is_available(): return 0 if not dist.is_initialized(): return 0 assert _LOCAL_PROCESS_GROUP is not None return dist.get_rank(group=_LOCAL_PROCESS_GROUP) def get_local_size() -> int: """ Returns: The size of the per-machine process group, i.e. the number of processes per machine. """ if not dist.is_available(): return 1 if not dist.is_initialized(): return 1 return dist.get_world_size(group=_LOCAL_PROCESS_GROUP) def is_main_process() -> bool: return get_rank() == 0 def synchronize(): """ Helper function to synchronize (barrier) among all processes when using distributed training """ if not dist.is_available(): return if not dist.is_initialized(): return world_size = dist.get_world_size() if world_size == 1: return dist.barrier() @functools.lru_cache() def _get_global_gloo_group(): """ Return a process group based on gloo backend, containing all the ranks The result is cached. """ if dist.get_backend() == "nccl": return dist.new_group(backend="gloo") else: return dist.group.WORLD def _serialize_to_tensor(data, group): backend = dist.get_backend(group) assert backend in ["gloo", "nccl"] device = torch.device("cpu" if backend == "gloo" else "cuda") buffer = pickle.dumps(data) if len(buffer) > 1024 ** 3: logger = logging.getLogger(__name__) logger.warning( "Rank {} trying to all-gather {:.2f} GB of data on device {}".format( get_rank(), len(buffer) / (1024 ** 3), device ) ) storage = torch.ByteStorage.from_buffer(buffer) tensor = torch.ByteTensor(storage).to(device=device) return tensor def _pad_to_largest_tensor(tensor, group): """ Returns: list[int]: size of the tensor, on each rank Tensor: padded tensor that has the max size """ world_size = dist.get_world_size(group=group) assert ( world_size >= 1 ), "comm.gather/all_gather must be called from ranks within the given group!" local_size = torch.tensor([tensor.numel()], dtype=torch.int64, device=tensor.device) size_list = [ torch.zeros([1], dtype=torch.int64, device=tensor.device) for _ in range(world_size) ] dist.all_gather(size_list, local_size, group=group) size_list = [int(size.item()) for size in size_list] max_size = max(size_list) # we pad the tensor because torch all_gather does not support # gathering tensors of different shapes if local_size != max_size: padding = torch.zeros( (max_size - local_size,), dtype=torch.uint8, device=tensor.device ) tensor = torch.cat((tensor, padding), dim=0) return size_list, tensor def all_gather(data, group=None): """ Run all_gather on arbitrary picklable data (not necessarily tensors). Args: data: any picklable object group: a torch process group. By default, will use a group which contains all ranks on gloo backend. Returns: list[data]: list of data gathered from each rank """ if get_world_size() == 1: return [data] if group is None: group = _get_global_gloo_group() if dist.get_world_size(group) == 1: return [data] tensor = _serialize_to_tensor(data, group) size_list, tensor = _pad_to_largest_tensor(tensor, group) max_size = max(size_list) # receiving Tensor from all ranks tensor_list = [ torch.empty((max_size,), dtype=torch.uint8, device=tensor.device) for _ in size_list ] dist.all_gather(tensor_list, tensor, group=group) data_list = [] for size, tensor in zip(size_list, tensor_list): buffer = tensor.cpu().numpy().tobytes()[:size] data_list.append(pickle.loads(buffer)) return data_list def gather(data, dst=0, group=None): """ Run gather on arbitrary picklable data (not necessarily tensors). Args: data: any picklable object dst (int): destination rank group: a torch process group. By default, will use a group which contains all ranks on gloo backend. Returns: list[data]: on dst, a list of data gathered from each rank. Otherwise, an empty list. """ if get_world_size() == 1: return [data] if group is None: group = _get_global_gloo_group() if dist.get_world_size(group=group) == 1: return [data] rank = dist.get_rank(group=group) tensor = _serialize_to_tensor(data, group) size_list, tensor = _pad_to_largest_tensor(tensor, group) # receiving Tensor from all ranks if rank == dst: max_size = max(size_list) tensor_list = [ torch.empty((max_size,), dtype=torch.uint8, device=tensor.device) for _ in size_list ] dist.gather(tensor, tensor_list, dst=dst, group=group) data_list = [] for size, tensor in zip(size_list, tensor_list): buffer = tensor.cpu().numpy().tobytes()[:size] data_list.append(pickle.loads(buffer)) return data_list else: dist.gather(tensor, [], dst=dst, group=group) return [] def shared_random_seed(): """ Returns: int: a random number that is the same across all workers. If workers need a shared RNG, they can use this shared seed to create one. All workers must call this function, otherwise it will deadlock. """ ints = np.random.randint(2 ** 31) all_ints = all_gather(ints) return all_ints[0] def reduce_dict(input_dict, average=True): """ Reduce the values in the dictionary from all processes so that process with rank 0 has the reduced results. Args: input_dict (dict): inputs to be reduced. All the values must be scalar CUDA Tensor. average (bool): whether to do average or sum Returns: a dict with the same keys as input_dict, after reduction. """ world_size = get_world_size() if world_size < 2: return input_dict with torch.no_grad(): names = [] values = [] # sort the keys so that they are consistent across processes for k in sorted(input_dict.keys()): names.append(k) values.append(input_dict[k]) values = torch.stack(values, dim=0) dist.reduce(values, dst=0) if dist.get_rank() == 0 and average: # only main process gets accumulated, so only divide by # world_size in this case values /= world_size reduced_dict = {k: v for k, v in zip(names, values)} return reduced_dict
7,814
27.837638
100
py
Mr.Right
Mr.Right-main/models/ViLT/vilt/modules/objectives.py
import torch import torch.nn as nn import torch.nn.functional as F import os import glob import json import tqdm import functools from torch.utils.data.distributed import DistributedSampler from einops import rearrange from models.ViLT.vilt.modules.dist_utils import all_gather def cost_matrix_cosine(x, y, eps=1e-5): """Compute cosine distnace across every pairs of x, y (batched) [B, L_x, D] [B, L_y, D] -> [B, Lx, Ly]""" assert x.dim() == y.dim() assert x.size(0) == y.size(0) assert x.size(2) == y.size(2) x_norm = F.normalize(x, p=2, dim=-1, eps=eps) y_norm = F.normalize(y, p=2, dim=-1, eps=eps) cosine_sim = x_norm.matmul(y_norm.transpose(1, 2)) cosine_dist = 1 - cosine_sim return cosine_dist def trace(x): """ compute trace of input tensor (batched) """ b, m, n = x.size() assert m == n mask = torch.eye(n, dtype=torch.bool, device=x.device).unsqueeze(0).expand_as(x) trace = x.masked_select(mask).contiguous().view(b, n).sum(dim=-1, keepdim=False) return trace @torch.no_grad() def ipot(C, x_len, x_pad, y_len, y_pad, joint_pad, beta, iteration, k): """ [B, M, N], [B], [B, M], [B], [B, N], [B, M, N]""" b, m, n = C.size() sigma = torch.ones(b, m, dtype=C.dtype, device=C.device) / x_len.unsqueeze(1) T = torch.ones(b, n, m, dtype=C.dtype, device=C.device) A = torch.exp(-C.transpose(1, 2) / beta) # mask padded positions sigma.masked_fill_(x_pad, 0) joint_pad = joint_pad.transpose(1, 2) T.masked_fill_(joint_pad, 0) A.masked_fill_(joint_pad, 0) # broadcastable lengths x_len = x_len.unsqueeze(1).unsqueeze(2) y_len = y_len.unsqueeze(1).unsqueeze(2) # mask to zero out padding in delta and sigma x_mask = (x_pad.to(C.dtype) * 1e4).unsqueeze(1) y_mask = (y_pad.to(C.dtype) * 1e4).unsqueeze(1) for _ in range(iteration): Q = A * T # bs * n * m sigma = sigma.view(b, m, 1) for _ in range(k): delta = 1 / (y_len * Q.matmul(sigma).view(b, 1, n) + y_mask) sigma = 1 / (x_len * delta.matmul(Q) + x_mask) T = delta.view(b, n, 1) * Q * sigma T.masked_fill_(joint_pad, 0) return T def optimal_transport_dist( txt_emb, img_emb, txt_pad, img_pad, beta=0.5, iteration=50, k=1 ): """ [B, M, D], [B, N, D], [B, M], [B, N]""" cost = cost_matrix_cosine(txt_emb, img_emb) # mask the padded inputs joint_pad = txt_pad.unsqueeze(-1) | img_pad.unsqueeze(-2) cost.masked_fill_(joint_pad, 0) txt_len = (txt_pad.size(1) - txt_pad.sum(dim=1, keepdim=False)).to(dtype=cost.dtype) img_len = (img_pad.size(1) - img_pad.sum(dim=1, keepdim=False)).to(dtype=cost.dtype) T = ipot( cost.detach(), txt_len, txt_pad, img_len, img_pad, joint_pad, beta, iteration, k ) distance = trace(cost.matmul(T.detach())) return distance def compute_mlm(pl_module, batch): infer = pl_module.infer(batch, mask_text=True, mask_image=False) mlm_logits = pl_module.mlm_score(infer["text_feats"]) mlm_labels = infer["text_labels"] mlm_loss = F.cross_entropy( mlm_logits.view(-1, pl_module.hparams.config["vocab_size"]), mlm_labels.view(-1), ignore_index=-100, ) ret = { "mlm_loss": mlm_loss, "mlm_logits": mlm_logits, "mlm_labels": mlm_labels, "mlm_ids": infer["text_ids"], } phase = "train" if pl_module.training else "val" loss = getattr(pl_module, f"{phase}_mlm_loss")(ret["mlm_loss"]) acc = getattr(pl_module, f"{phase}_mlm_accuracy")( ret["mlm_logits"], ret["mlm_labels"] ) pl_module.log(f"mlm/{phase}/loss", loss) pl_module.log(f"mlm/{phase}/accuracy", acc) return ret def compute_mpp(pl_module, batch): infer = pl_module.infer(batch, mask_text=False, mask_image=True) mpp_logits = pl_module.mpp_score(infer["image_feats"]) mpp_logits = torch.stack( [ mpp_logits[:, :, 0:256], mpp_logits[:, :, 256:512], mpp_logits[:, :, 512:768], ], dim=2, ) mpp_labels = infer["image_labels"] mpp_loss = F.cross_entropy( mpp_logits.view(-1, 256), mpp_labels.view(-1), ignore_index=-100, ) ret = { "mpp_loss": mpp_loss, "mpp_logits": mpp_logits, "mpp_labels": mpp_labels, } phase = "train" if pl_module.training else "val" loss = getattr(pl_module, f"{phase}_mpp_loss")(ret["mpp_loss"]) acc = getattr(pl_module, f"{phase}_mpp_accuracy")( ret["mpp_logits"], ret["mpp_labels"] ) pl_module.log(f"mpp/{phase}/loss", loss) pl_module.log(f"mpp/{phase}/accuracy", acc) return ret def compute_mppd(pl_module, batch): infer = pl_module.infer(batch, mask_text=False, mask_image=True) mppd_logits = pl_module.mppd_score(infer["image_feats"]) mppd_labels = infer["image_labels_mppd"] filter_to_train = infer["image_labels"].float().mean(dim=-1) != -100 labels = mppd_labels[filter_to_train] logits = mppd_logits[filter_to_train] mppd_loss = F.mse_loss(logits, labels) ret = { "mppd_loss": mppd_loss, "mppd_logits": mppd_logits, "mppd_labels": mppd_labels, } phase = "train" if pl_module.training else "val" loss = getattr(pl_module, f"{phase}_mppd_loss")(ret["mppd_loss"]) pl_module.log(f"mppd/{phase}/loss", loss) return ret def compute_mpfr(pl_module, batch): infer = pl_module.infer(batch, mask_text=False, mask_image=True) mpfr_logits = pl_module.mpfr_score(infer["image_feats"]) mpfr_labels = infer["image_labels_mpfr"] filter_to_train = infer["image_labels"].float().mean(dim=-1) != -100 labels = mpfr_labels[filter_to_train] logits = mpfr_logits[filter_to_train] mpfr_loss = F.mse_loss(logits, labels) ret = { "mpfr_loss": mpfr_loss, "mpfr_logits": mpfr_logits, "mpfr_labels": mpfr_labels, } phase = "train" if pl_module.training else "val" loss = getattr(pl_module, f"{phase}_mpfr_loss")(ret["mpfr_loss"]) pl_module.log(f"mpfr/{phase}/loss", loss) return ret def compute_itm_wpa(pl_module, batch): pos_len = len(batch["text"]) // 2 neg_len = len(batch["text"]) - pos_len itm_labels = torch.cat([torch.ones(pos_len), torch.zeros(neg_len)]).to( pl_module.device ) itm_labels = itm_labels[torch.randperm(itm_labels.size(0))] itm_images = [ torch.stack( [ ti if itm_labels[i] == 1 else fi for i, (ti, fi) in enumerate(zip(bti, bfi)) ] ) for bti, bfi in zip(batch["image"], batch["false_image_0"]) ] batch = {k: v for k, v in batch.items()} batch["image"] = itm_images infer = pl_module.infer(batch, mask_text=False, mask_image=False) with torch.cuda.amp.autocast(enabled=False): txt_emb, img_emb = infer["text_feats"], infer["image_feats"] txt_mask, img_mask = infer["text_masks"].bool(), infer["image_masks"].bool() for i, _len in enumerate(txt_mask.sum(dim=1)): txt_mask[i, _len - 1] = False txt_mask[:, 0] = False img_mask[:, 0] = False if "deit" in pl_module.hparams.config["vit"]: img_mask[:, 1] = False txt_pad, img_pad = ~txt_mask, ~img_mask cost = cost_matrix_cosine(txt_emb.float(), img_emb.float()) joint_pad = txt_pad.unsqueeze(-1) | img_pad.unsqueeze(-2) cost.masked_fill_(joint_pad, 0) txt_len = (txt_pad.size(1) - txt_pad.sum(dim=1, keepdim=False)).to( dtype=cost.dtype ) img_len = (img_pad.size(1) - img_pad.sum(dim=1, keepdim=False)).to( dtype=cost.dtype ) T = ipot( cost.detach(), txt_len, txt_pad, img_len, img_pad, joint_pad, 0.5, 50, 1 ) distance = trace(cost.matmul(T.detach())) dist_pos = distance.masked_select(itm_labels == 1) dist_neg = distance.masked_select(itm_labels == 0) ot_loss = (dist_pos.sum() - dist_neg.sum()) / (dist_pos.size(0) + dist_neg.size(0)) itm_logits = pl_module.itm_score(infer["cls_feats"]) itm_loss = F.cross_entropy(itm_logits, itm_labels.long()) ret = { "itm_loss": itm_loss, "itm_wpa_loss": 0.1 * ot_loss, "itm_logits": itm_logits, "itm_labels": itm_labels, } phase = "train" if pl_module.training else "val" loss = getattr(pl_module, f"{phase}_itm_loss")(ret["itm_loss"]) wpa_loss = getattr(pl_module, f"{phase}_itm_wpa_loss")(ret["itm_wpa_loss"]) acc = getattr(pl_module, f"{phase}_itm_accuracy")( ret["itm_logits"], ret["itm_labels"] ) pl_module.log(f"itm/{phase}/loss", loss) pl_module.log(f"itm/{phase}/wpa_loss", wpa_loss) pl_module.log(f"itm/{phase}/accuracy", acc) return ret def compute_imgcls(pl_module, batch): infer = pl_module.infer(batch, mask_text=False, mask_image=False) imgcls_logits = pl_module.img_classifier(infer["cls_feats"]) imgcls_labels = batch["label"] imgcls_labels = torch.tensor(imgcls_labels).to(pl_module.device).long() imgcls_loss = F.cross_entropy(imgcls_logits, imgcls_labels) ret = { "imgcls_loss": imgcls_loss, "imgcls_logits": imgcls_logits, "imgcls_labels": imgcls_labels, } phase = "train" if pl_module.training else "val" loss = getattr(pl_module, f"{phase}_imgcls_loss")(ret["imgcls_loss"]) acc = getattr(pl_module, f"{phase}_imgcls_accuracy")( ret["imgcls_logits"], ret["imgcls_labels"] ) pl_module.log(f"imgcls/{phase}/loss", loss) pl_module.log(f"imgcls/{phase}/accuracy", acc) return ret def compute_vqa(pl_module, batch): infer = pl_module.infer(batch, mask_text=False, mask_image=False) vqa_logits = pl_module.vqa_classifier(infer["cls_feats"]) vqa_targets = torch.zeros( len(vqa_logits), pl_module.hparams.config["vqav2_label_size"] ).to(pl_module.device) vqa_labels = batch["vqa_labels"] vqa_scores = batch["vqa_scores"] for i, (_label, _score) in enumerate(zip(vqa_labels, vqa_scores)): for l, s in zip(_label, _score): vqa_targets[i, l] = s vqa_loss = ( F.binary_cross_entropy_with_logits(vqa_logits, vqa_targets) * vqa_targets.shape[1] ) # https://github.com/jnhwkim/ban-vqa/blob/master/train.py#L19 ret = { "vqa_loss": vqa_loss, "vqa_logits": vqa_logits, "vqa_targets": vqa_targets, "vqa_labels": vqa_labels, "vqa_scores": vqa_scores, } phase = "train" if pl_module.training else "val" loss = getattr(pl_module, f"{phase}_vqa_loss")(ret["vqa_loss"]) score = getattr(pl_module, f"{phase}_vqa_score")( ret["vqa_logits"], ret["vqa_targets"] ) pl_module.log(f"vqa/{phase}/loss", loss) pl_module.log(f"vqa/{phase}/score", score) return ret def compute_nlvr2(pl_module, batch): infer1 = pl_module.infer( batch, mask_text=False, mask_image=False, image_token_type_idx=1 ) infer2 = pl_module.infer( batch, mask_text=False, mask_image=False, image_token_type_idx=2 ) cls_feats = torch.cat([infer1["cls_feats"], infer2["cls_feats"]], dim=-1) nlvr2_logits = pl_module.nlvr2_classifier(cls_feats) nlvr2_labels = batch["answers"] nlvr2_labels = torch.tensor(nlvr2_labels).to(pl_module.device).long() nlvr2_loss = F.cross_entropy(nlvr2_logits, nlvr2_labels) ret = { "nlvr2_loss": nlvr2_loss, "nlvr2_logits": nlvr2_logits, "nlvr2_labels": nlvr2_labels, } phase = "train" if pl_module.training else "val" if phase == "train": loss = getattr(pl_module, f"{phase}_nlvr2_loss")(ret["nlvr2_loss"]) acc = getattr(pl_module, f"{phase}_nlvr2_accuracy")( ret["nlvr2_logits"], ret["nlvr2_labels"] ) pl_module.log(f"nlvr2/{phase}/loss", loss) pl_module.log(f"nlvr2/{phase}/accuracy", acc) else: dev_batches = [i for i, n in enumerate(batch["table_name"]) if "dev" in n] test_batches = [i for i, n in enumerate(batch["table_name"]) if "test" in n] if dev_batches: dev_loss = getattr(pl_module, f"dev_nlvr2_loss")( F.cross_entropy( ret["nlvr2_logits"][dev_batches], ret["nlvr2_labels"][dev_batches] ) ) dev_acc = getattr(pl_module, f"dev_nlvr2_accuracy")( ret["nlvr2_logits"][dev_batches], ret["nlvr2_labels"][dev_batches] ) pl_module.log(f"nlvr2/dev/loss", dev_loss) pl_module.log(f"nlvr2/dev/accuracy", dev_acc) if test_batches: test_loss = getattr(pl_module, f"test_nlvr2_loss")( F.cross_entropy( ret["nlvr2_logits"][test_batches], ret["nlvr2_labels"][test_batches] ) ) test_acc = getattr(pl_module, f"test_nlvr2_accuracy")( ret["nlvr2_logits"][test_batches], ret["nlvr2_labels"][test_batches] ) pl_module.log(f"nlvr2/test/loss", test_loss) pl_module.log(f"nlvr2/test/accuracy", test_acc) return ret def compute_irtr(pl_module, batch): is_training_phase = pl_module.training _bs, _c, _h, _w = batch["image"][0].shape false_len = pl_module.hparams.config["draw_false_text"] text_ids = torch.stack( [batch[f"false_text_{i}_ids"] for i in range(false_len)], dim=1 ) text_masks = torch.stack( [batch[f"false_text_{i}_masks"] for i in range(false_len)], dim=1 ) text_labels = torch.stack( [batch[f"false_text_{i}_labels"] for i in range(false_len)], dim=1 ) text_ids = torch.cat([batch["text_ids"].unsqueeze(1), text_ids], dim=1) text_masks = torch.cat([batch["text_masks"].unsqueeze(1), text_masks], dim=1) text_labels = torch.cat([batch["text_labels"].unsqueeze(1), text_labels], dim=1) images = batch["image"][0].unsqueeze(1).expand(_bs, false_len + 1, _c, _h, _w) infer = pl_module.infer( { "image": [rearrange(images, "bs fs c h w -> (bs fs) c h w")], "text_ids": rearrange(text_ids, "bs fs tl -> (bs fs) tl"), "text_masks": rearrange(text_masks, "bs fs tl -> (bs fs) tl"), "text_labels": rearrange(text_labels, "bs fs tl -> (bs fs) tl"), } ) score = pl_module.rank_output(infer["cls_feats"])[:, 0] score = rearrange(score, "(bs fs) -> bs fs", bs=_bs, fs=false_len + 1) answer = torch.zeros(_bs).to(score).long() irtr_loss = F.cross_entropy(score, answer) ret = { "irtr_loss": irtr_loss, } phase = "train" if pl_module.training else "val" irtr_loss = getattr(pl_module, f"{phase}_irtr_loss")(ret["irtr_loss"]) pl_module.log(f"irtr/{phase}/irtr_loss", irtr_loss) return ret @torch.no_grad() def compute_irtr_recall(pl_module): text_dset = pl_module.trainer.datamodule.dms[0].make_no_false_val_dset() text_dset.tokenizer = pl_module.trainer.datamodule.dms[0].tokenizer text_loader = torch.utils.data.DataLoader( text_dset, batch_size=64, num_workers=pl_module.hparams.config["num_workers"], pin_memory=True, collate_fn=functools.partial( text_dset.collate, mlm_collator=pl_module.trainer.datamodule.dms[0].mlm_collator, ), ) image_dset = pl_module.trainer.datamodule.dms[0].make_no_false_val_dset( image_only=True ) image_dset.tokenizer = pl_module.trainer.datamodule.dms[0].tokenizer dist_sampler = DistributedSampler(image_dset, shuffle=False) image_loader = torch.utils.data.DataLoader( image_dset, batch_size=1, num_workers=pl_module.hparams.config["num_workers"], sampler=dist_sampler, pin_memory=True, collate_fn=functools.partial( image_dset.collate, mlm_collator=pl_module.trainer.datamodule.dms[0].mlm_collator, ), ) text_preload = list() for _b in tqdm.tqdm(text_loader, desc="text prefetch loop"): text_preload.append( { "text_ids": _b["text_ids"].to(pl_module.device), "text_masks": _b["text_masks"].to(pl_module.device), "text_labels": _b["text_labels"].to(pl_module.device), "img_index": _b["img_index"], } ) tiids = list() for pre in text_preload: tiids += pre["img_index"] tiids = torch.tensor(tiids) image_preload = list() for _b in tqdm.tqdm(image_loader, desc="image prefetch loop"): (ie, im, _, _) = pl_module.transformer.visual_embed( _b["image"][0].to(pl_module.device), max_image_len=pl_module.hparams.config["max_image_len"], mask_it=False, ) image_preload.append((ie, im, _b["img_index"][0])) rank_scores = list() rank_iids = list() for img_batch in tqdm.tqdm(image_preload, desc="rank loop"): _ie, _im, _iid = img_batch _, l, c = _ie.shape img_batch_score = list() for txt_batch in text_preload: fblen = len(txt_batch["text_ids"]) ie = _ie.expand(fblen, l, c) im = _im.expand(fblen, l) with torch.cuda.amp.autocast(): score = pl_module.rank_output( pl_module.infer( { "text_ids": txt_batch["text_ids"], "text_masks": txt_batch["text_masks"], "text_labels": txt_batch["text_labels"], }, image_embeds=ie, image_masks=im, )["cls_feats"] )[:, 0] img_batch_score.append(score) img_batch_score = torch.cat(img_batch_score) rank_scores.append(img_batch_score.cpu().tolist()) rank_iids.append(_iid) torch.distributed.barrier() gather_rank_scores = all_gather(rank_scores) gather_rank_iids = all_gather(rank_iids) iids = torch.tensor(gather_rank_iids) iids = iids.view(-1) scores = torch.tensor(gather_rank_scores) scores = scores.view(len(iids), -1) topk10 = scores.topk(10, dim=1) topk5 = scores.topk(5, dim=1) topk1 = scores.topk(1, dim=1) topk10_iids = tiids[topk10.indices] topk5_iids = tiids[topk5.indices] topk1_iids = tiids[topk1.indices] tr_r10 = (iids.unsqueeze(1) == topk10_iids).float().max(dim=1)[0].mean() tr_r5 = (iids.unsqueeze(1) == topk5_iids).float().max(dim=1)[0].mean() tr_r1 = (iids.unsqueeze(1) == topk1_iids).float().max(dim=1)[0].mean() topk10 = scores.topk(10, dim=0) topk5 = scores.topk(5, dim=0) topk1 = scores.topk(1, dim=0) topk10_iids = iids[topk10.indices] topk5_iids = iids[topk5.indices] topk1_iids = iids[topk1.indices] ir_r10 = (tiids.unsqueeze(0) == topk10_iids).float().max(dim=0)[0].mean() ir_r5 = (tiids.unsqueeze(0) == topk5_iids).float().max(dim=0)[0].mean() ir_r1 = (tiids.unsqueeze(0) == topk1_iids).float().max(dim=0)[0].mean() return (ir_r1, ir_r5, ir_r10, tr_r1, tr_r5, tr_r10) def init_weights(module): if isinstance(module, (nn.Linear, nn.Embedding)): module.weight.data.normal_(mean=0.0, std=0.02) elif isinstance(module, nn.LayerNorm): module.bias.data.zero_() module.weight.data.fill_(1.0) if isinstance(module, nn.Linear) and module.bias is not None: module.bias.data.zero_() def vqa_test_step(pl_module, batch, output): id2answer = ( pl_module.trainer.datamodule.dm_dicts["vqa_trainval"].id2answer if "vqa_trainval" in pl_module.trainer.datamodule.dm_dicts else pl_module.trainer.datamodule.dm_dicts["vqa"].id2answer ) vqa_logits = output["vqa_logits"] vqa_preds = vqa_logits.argmax(dim=-1) vqa_preds = [id2answer[pred.item()] for pred in vqa_preds] questions = batch["text"] qids = batch["qid"] return {"qids": qids, "preds": vqa_preds} def arc_test_step(pl_module, batch, output): return output def vqa_test_wrapup(outs, model_name): rank = torch.distributed.get_rank() qids, preds = list(), list() for out in outs: qids += out["qids"] preds += out["preds"] rets = list() for qid, pred in zip(qids, preds): rets.append({"question_id": qid, "answer": pred}) with open(f"vqa_submit_{rank}.json", "w") as fp: json.dump(rets, fp, indent=4) torch.distributed.barrier() if rank == 0: jsons = list() paths = list(glob.glob("vqa_submit_*.json")) for path in paths: with open(path, "r") as fp: jsons += json.load(fp) os.makedirs("result", exist_ok=True) with open(f"result/vqa_submit_{model_name}.json", "w") as fp: json.dump(jsons, fp, indent=4) torch.distributed.barrier() os.remove(f"vqa_submit_{rank}.json") def arc_test_wrapup(outs, caplen, model_name): rank = torch.distributed.get_rank() iids, captions = list(), list() for out in outs: iids += out["iid"] captions += out["captions"] rets = list() for iid, caption in zip(iids, captions): rets.append({"image_id": iid, "caption": caption}) with open(f"coco_cap_len{caplen}_{rank}.json", "w") as fp: json.dump(rets, fp, indent=4) torch.distributed.barrier() if rank == 0: jsons = list() paths = list(glob.glob(f"coco_cap_len{caplen}_*.json")) for path in paths: with open(path, "r") as fp: jsons += json.load(fp) os.makedirs("result/arc", exist_ok=True) jsons = sorted(jsons, key=lambda x: x["image_id"]) with open(f"result/arc/coco_cap_{model_name}_len{caplen}.json", "w") as fp: json.dump(jsons, fp, indent=4) torch.distributed.barrier() os.remove(f"coco_cap_len{caplen}_{rank}.json")
22,098
32.842266
88
py
Mr.Right
Mr.Right-main/models/ViLT/vilt/modules/vilt_module.py
import torch torch.autograd.set_detect_anomaly(True) import torch.nn as nn import pytorch_lightning as pl import models.ViLT.vilt.modules.vision_transformer as vit import pdb from transformers.models.bert.modeling_bert import BertConfig, BertEmbeddings from models.ViLT.vilt.modules import heads, objectives # from models.ViLT.vilt.modules import vilt_utils class ViLTransformerSS(pl.LightningModule): def __init__(self, config): super().__init__() self.save_hyperparameters() self.bert_config = BertConfig( vocab_size=config["vocab_size"], hidden_size=config["hidden_size"], num_hidden_layers=config["num_layers"], num_attention_heads=config["num_heads"], intermediate_size=config["hidden_size"] * config["mlp_ratio"], max_position_embeddings=config["max_text_len"], hidden_dropout_prob=config["drop_rate"], attention_probs_dropout_prob=config["drop_rate"], ) self.text_embeddings = BertEmbeddings(self.bert_config) self.text_embeddings.apply(objectives.init_weights) self.token_type_embeddings = nn.Embedding(2, config["hidden_size"]) self.token_type_embeddings.apply(objectives.init_weights) if self.hparams.config["checkpoint"] == "": self.transformer = getattr(vit, self.hparams.config["vit"])( pretrained=True, config=self.hparams.config ) else: self.transformer = getattr(vit, self.hparams.config["vit"])( pretrained=False, config=self.hparams.config ) self.pooler = heads.Pooler(config["hidden_size"]) self.pooler.apply(objectives.init_weights) if config["loss_names"]["mlm"] > 0: self.mlm_score = heads.MLMHead(self.bert_config) self.mlm_score.apply(objectives.init_weights) if config["loss_names"]["itm"] > 0: self.itm_score = heads.ITMHead(config["hidden_size"]) self.itm_score.apply(objectives.init_weights) if config["loss_names"]["mpp"] > 0: self.mpp_score = heads.MPPHead(self.bert_config) self.mpp_score.apply(objectives.init_weights) # ===================== Downstream ===================== # hs = self.hparams.config["hidden_size"] if self.hparams.config["loss_names"]["vqa"] > 0: vs = self.hparams.config["vqav2_label_size"] self.vqa_classifier = nn.Sequential( nn.Linear(hs, hs * 2), nn.LayerNorm(hs * 2), nn.GELU(), nn.Linear(hs * 2, vs), ) self.vqa_classifier.apply(objectives.init_weights) if self.hparams.config["loss_names"]["nlvr2"] > 0: self.nlvr2_classifier = nn.Sequential( nn.Linear(hs * 2, hs * 2), nn.LayerNorm(hs * 2), nn.GELU(), nn.Linear(hs * 2, 2), ) self.nlvr2_classifier.apply(objectives.init_weights) emb_data = self.token_type_embeddings.weight.data self.token_type_embeddings = nn.Embedding(3, hs) self.token_type_embeddings.apply(objectives.init_weights) self.token_type_embeddings.weight.data[0, :] = emb_data[0, :] self.token_type_embeddings.weight.data[1, :] = emb_data[1, :] self.token_type_embeddings.weight.data[2, :] = emb_data[1, :] if self.hparams.config["loss_names"]["irtr"] > 0: self.rank_output = nn.Linear(hs, 1) self.rank_output.weight.data = self.itm_score.fc.weight.data[1:, :] self.rank_output.bias.data = self.itm_score.fc.bias.data[1:] self.margin = 0.2 for p in self.itm_score.parameters(): p.requires_grad = False # vilt_utils.set_metrics(self) self.current_tasks = list() # ===================== load downstream (test_only) ====================== if self.hparams.config["checkpoint"] != "": ckpt = torch.load(self.hparams.config["checkpoint"], map_location="cpu") state_dict = ckpt['state_dict'] # interpolate position embedding org_embeds = state_dict['text_embeddings.position_embeddings.weight'] org_tokens_len = org_embeds.shape[0] new_tokens_len = self.text_embeddings.position_embeddings.weight.shape[0] if org_tokens_len > new_tokens_len: state_dict['text_embeddings.position_embeddings.weight'] = org_embeds[:new_tokens_len,:] else: extra_tokens = new_tokens_len - org_tokens_len extra_tokens_embeds = nn.Embedding(extra_tokens, self.hparams.config["hidden_size"]) extra_tokens_embeds_weight = extra_tokens_embeds.weight state_dict['text_embeddings.position_embeddings.weight'] = torch.cat([org_embeds,extra_tokens_embeds_weight],dim=0) state_dict['text_embeddings.position_ids'] = self.text_embeddings.position_ids msg = self.load_state_dict(state_dict, strict=False) print(msg) def infer( self, batch, mask_text=False, mask_image=False, image_token_type_idx=1, image_embeds=None, image_masks=None, ): imgkey = "image" do_mlm = "_mlm" if mask_text else "" text_ids = batch[f"text_ids{do_mlm}"] # text_labels = batch[f"text_labels{do_mlm}"] text_masks = batch[f"text_masks"] text_embeds = self.text_embeddings(text_ids) query_ids = batch[f"query_ids"] query_embeds = self.text_embeddings(query_ids) query_masks = batch[f"query_masks"] check = torch.isnan(query_embeds) # print(-1,torch.isnan(check).any()) # print((check == True).nonzero(as_tuple=True)) for i, blk in enumerate(self.transformer.blocks): query_embeds, _ = blk(query_embeds, mask=query_masks) query_embeds = self.transformer.norm(query_embeds) if image_embeds is None and image_masks is None: img = batch[imgkey] ( image_embeds, image_masks, patch_index, image_labels, ) = self.transformer.visual_embed( img, max_image_len=self.hparams.config["max_image_len"], mask_it=mask_image, ) else: patch_index, image_labels = ( None, None, ) text_embeds, image_embeds = ( text_embeds + self.token_type_embeddings(torch.zeros_like(text_masks)), image_embeds + self.token_type_embeddings(torch.full_like(image_masks, image_token_type_idx)), ) # co_embeds = text_embeds # co_masks = text_masks co_embeds = torch.cat([text_embeds, image_embeds], dim=1) co_masks = torch.cat([text_masks, image_masks], dim=1) x = co_embeds for i, blk in enumerate(self.transformer.blocks): x, _attn = blk(x, mask=co_masks) x = self.transformer.norm(x) text_feats, image_feats = ( x[:, : text_embeds.shape[1]], x[:, text_embeds.shape[1] :], ) # text_feats, image_feats = x, None cls_feats = self.pooler(x) ret = { "query_embeds": query_embeds, "query_atts": query_masks, "query_cls": query_embeds[:,0,:], "text_feats": text_feats, "image_feats": image_feats, "multi_cls": cls_feats, "raw_cls_feats": x[:, 0], "image_labels": image_labels, "image_atts": image_masks, "multi_embeds": x, "multi_atts": co_masks, # "text_labels": text_labels, # "text_ids": text_ids, "text_masks": text_masks, "patch_index": patch_index, } return ret @torch.no_grad() def output_multi(self,batch,image_token_type_idx=1): imgkey = "image" text_ids = batch["text_ids"] text_masks = batch["text_masks"] text_embeds = self.text_embeddings(text_ids) img = batch[imgkey] ( image_embeds, image_masks, patch_index, image_labels, ) = self.transformer.visual_embed( img, max_image_len=self.hparams.config["max_image_len"], mask_it=False, ) text_embeds, image_embeds = ( text_embeds + self.token_type_embeddings(torch.zeros_like(text_masks)), image_embeds + self.token_type_embeddings( torch.full_like(image_masks, image_token_type_idx) ), ) # co_embeds = text_embeds # co_masks = text_masks co_embeds = torch.cat([text_embeds, image_embeds], dim=1) co_masks = torch.cat([text_masks, image_masks], dim=1) x = co_embeds for i, blk in enumerate(self.transformer.blocks): x, _attn = blk(x, mask=co_masks) x = self.transformer.norm(x) text_feats, image_feats = ( x[:, : text_embeds.shape[1]], x[:, text_embeds.shape[1] :], ) # text_feats, image_feats = x, None cls_feats = self.pooler(x) ret = { "text_feats": text_feats, "image_feats": image_feats, "multi_cls": cls_feats, "raw_cls_feats": x[:, 0], "image_labels": image_labels, "image_atts": image_masks, "multi_embeds": x, "multi_atts": co_masks, # "text_labels": text_labels, # "text_ids": text_ids, "text_masks": text_masks, "patch_index": patch_index, } return ret def forward(self, batch): ret = dict() if len(self.current_tasks) == 0: ret.update(self.infer(batch)) return ret # Masked Language Modeling if "mlm" in self.current_tasks: ret.update(objectives.compute_mlm(self, batch)) # Masked Patch Prediction if "mpp" in self.current_tasks: ret.update(objectives.compute_mpp(self, batch)) # Image Text Matching if "itm" in self.current_tasks: ret.update(objectives.compute_itm_wpa(self, batch)) # Visual Question Answering if "vqa" in self.current_tasks: ret.update(objectives.compute_vqa(self, batch)) # Natural Language for Visual Reasoning 2 if "nlvr2" in self.current_tasks: ret.update(objectives.compute_nlvr2(self, batch)) # Image Retrieval and Text Retrieval if "irtr" in self.current_tasks: ret.update(objectives.compute_irtr(self, batch)) return ret def training_step(self, batch, batch_idx): # vilt_utils.set_task(self) output = self(batch) total_loss = sum([v for k, v in output.items() if "loss" in k]) return total_loss def training_epoch_end(self, outs): pass # vilt_utils.epoch_wrapup(self) def validation_step(self, batch, batch_idx): # vilt_utils.set_task(self) output = self(batch) def validation_epoch_end(self, outs): pass # vilt_utils.epoch_wrapup(self) def test_step(self, batch, batch_idx): # vilt_utils.set_task(self) output = self(batch) ret = dict() if self.hparams.config["loss_names"]["vqa"] > 0: ret.update(objectives.vqa_test_step(self, batch, output)) return ret def test_epoch_end(self, outs): model_name = self.hparams.config["checkpoint"].split("/")[-1][:-5] if self.hparams.config["loss_names"]["vqa"] > 0: objectives.vqa_test_wrapup(outs, model_name) # vilt_utils.epoch_wrapup(self) def configure_optimizers(self): pass # return vilt_utils.set_schedule(self)
10,172
28.148997
119
py
Mr.Right
Mr.Right-main/models/ViLT/vilt/modules/vision_transformer.py
""" Vision Transformer (ViT) in PyTorch A PyTorch implement of Vision Transformers as described in 'An Image Is Worth 16 x 16 Words: Transformers for Image Recognition at Scale' - https://arxiv.org/abs/2010.11929 The official jax code is released and available at https://github.com/google-research/vision_transformer Acknowledgments: * The paper authors for releasing code and weights, thanks! * I fixed my class token impl based on Phil Wang's https://github.com/lucidrains/vit-pytorch ... check it out for some einops/einsum fun * Simple transformer style inspired by Andrej Karpathy's https://github.com/karpathy/minGPT * Bert reference code checks against Huggingface Transformers and Tensorflow Bert DeiT model defs and weights from https://github.com/facebookresearch/deit, paper `DeiT: Data-efficient Image Transformers` - https://arxiv.org/abs/2012.12877 Hacked together by / Copyright 2020 Ross Wightman """ import math import logging from functools import partial import torch import torch.nn as nn import torch.nn.functional as F import hashlib import os import urllib import warnings from functools import partial from tqdm import tqdm from timm.data import IMAGENET_DEFAULT_MEAN, IMAGENET_DEFAULT_STD from timm.models.helpers import load_pretrained from timm.models.layers import StdConv2dSame, DropPath, to_2tuple, trunc_normal_ from timm.models.resnet import resnet26d, resnet50d from timm.models.resnetv2 import ResNetV2 from timm.models.registry import register_model from torchvision import transforms _logger = logging.getLogger(__name__) def download_clip( url: str = "https://openaipublic.azureedge.net/clip/models/40d365715913c9da98579312b702a82c18be219cc2a73407c4526f58eba950af/ViT-B-32.pt", root: str = os.path.expanduser("~/.cache/clip"), ): os.makedirs(root, exist_ok=True) filename = os.path.basename(url) expected_sha256 = url.split("/")[-2] download_target = os.path.join(root, filename) if os.path.exists(download_target) and not os.path.isfile(download_target): raise RuntimeError(f"{download_target} exists and is not a regular file") if os.path.isfile(download_target): if ( hashlib.sha256(open(download_target, "rb").read()).hexdigest() == expected_sha256 ): return download_target else: warnings.warn( f"{download_target} exists, but the SHA256 checksum does not match; re-downloading the file" ) with urllib.request.urlopen(url) as source, open(download_target, "wb") as output: with tqdm(total=int(source.info().get("Content-Length")), ncols=80) as loop: while True: buffer = source.read(8192) if not buffer: break output.write(buffer) loop.update(len(buffer)) if ( hashlib.sha256(open(download_target, "rb").read()).hexdigest() != expected_sha256 ): raise RuntimeError( f"Model has been downloaded but the SHA256 checksum does not not match" ) return download_target class UnNormalize(object): def __init__(self, mean, std): self.mean = mean self.std = std def __call__(self, tensor): for t, m, s in zip(tensor, self.mean, self.std): t.mul_(s).add_(m) return tensor inception_unnormalize = transforms.Compose( [UnNormalize(mean=[0.5, 0.5, 0.5], std=[0.5, 0.5, 0.5])] ) def _cfg(url="", **kwargs): return { "url": url, "num_classes": 1000, "input_size": (3, 224, 224), "pool_size": None, "crop_pct": 0.9, "interpolation": "bicubic", "mean": IMAGENET_DEFAULT_MEAN, "std": IMAGENET_DEFAULT_STD, "first_conv": "patch_embed.proj", "classifier": "head", **kwargs, } default_cfgs = { # patch models (my experiments) "vit_small_patch16_224": _cfg( url="https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-weights/vit_small_p16_224-15ec54c9.pth", ), # patch models (weights ported from official Google JAX impl) "vit_base_patch16_224": _cfg( url="https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-vitjx/jx_vit_base_p16_224-80ecf9dd.pth", mean=(0.5, 0.5, 0.5), std=(0.5, 0.5, 0.5), ), "vit_base_patch32_224": _cfg( url="", # no official model weights for this combo, only for in21k mean=(0.5, 0.5, 0.5), std=(0.5, 0.5, 0.5), ), "vit_base_patch16_384": _cfg( url="https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-vitjx/jx_vit_base_p16_384-83fb41ba.pth", input_size=(3, 384, 384), mean=(0.5, 0.5, 0.5), std=(0.5, 0.5, 0.5), crop_pct=1.0, ), "vit_base_patch32_384": _cfg( url="https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-vitjx/jx_vit_base_p32_384-830016f5.pth", input_size=(3, 384, 384), mean=(0.5, 0.5, 0.5), std=(0.5, 0.5, 0.5), crop_pct=1.0, ), "vit_large_patch16_224": _cfg( url="https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-vitjx/jx_vit_large_p16_224-4ee7a4dc.pth", mean=(0.5, 0.5, 0.5), std=(0.5, 0.5, 0.5), ), "vit_large_patch32_224": _cfg( url="", # no official model weights for this combo, only for in21k mean=(0.5, 0.5, 0.5), std=(0.5, 0.5, 0.5), ), "vit_large_patch16_384": _cfg( url="https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-vitjx/jx_vit_large_p16_384-b3be5167.pth", input_size=(3, 384, 384), mean=(0.5, 0.5, 0.5), std=(0.5, 0.5, 0.5), crop_pct=1.0, ), "vit_large_patch32_384": _cfg( url="https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-vitjx/jx_vit_large_p32_384-9b920ba8.pth", input_size=(3, 384, 384), mean=(0.5, 0.5, 0.5), std=(0.5, 0.5, 0.5), crop_pct=1.0, ), # patch models, imagenet21k (weights ported from official Google JAX impl) "vit_base_patch16_224_in21k": _cfg( url="https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-vitjx/jx_vit_base_patch16_224_in21k-e5005f0a.pth", num_classes=21843, mean=(0.5, 0.5, 0.5), std=(0.5, 0.5, 0.5), ), "vit_base_patch32_224_in21k": _cfg( url="https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-vitjx/jx_vit_base_patch32_224_in21k-8db57226.pth", num_classes=21843, mean=(0.5, 0.5, 0.5), std=(0.5, 0.5, 0.5), ), "vit_large_patch16_224_in21k": _cfg( url="https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-vitjx/jx_vit_large_patch16_224_in21k-606da67d.pth", num_classes=21843, mean=(0.5, 0.5, 0.5), std=(0.5, 0.5, 0.5), ), "vit_large_patch32_224_in21k": _cfg( url="https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-vitjx/jx_vit_large_patch32_224_in21k-9046d2e7.pth", num_classes=21843, mean=(0.5, 0.5, 0.5), std=(0.5, 0.5, 0.5), ), "vit_huge_patch14_224_in21k": _cfg( url="", # FIXME I have weights for this but > 2GB limit for github release binaries num_classes=21843, mean=(0.5, 0.5, 0.5), std=(0.5, 0.5, 0.5), ), # hybrid models (weights ported from official Google JAX impl) "vit_base_resnet50_224_in21k": _cfg( url="https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-vitjx/jx_vit_base_resnet50_224_in21k-6f7c7740.pth", num_classes=21843, mean=(0.5, 0.5, 0.5), std=(0.5, 0.5, 0.5), crop_pct=0.9, first_conv="patch_embed.backbone.stem.conv", ), "vit_base_resnet50_384": _cfg( url="https://github.com/rwightman/pytorch-image-models/releases/download/v0.1-vitjx/jx_vit_base_resnet50_384-9fd3c705.pth", input_size=(3, 384, 384), mean=(0.5, 0.5, 0.5), std=(0.5, 0.5, 0.5), crop_pct=1.0, first_conv="patch_embed.backbone.stem.conv", ), # hybrid models (my experiments) "vit_small_resnet26d_224": _cfg(), "vit_small_resnet50d_s3_224": _cfg(), "vit_base_resnet26d_224": _cfg(), "vit_base_resnet50d_224": _cfg(), # deit models (FB weights) "vit_deit_tiny_patch16_224": _cfg( url="https://dl.fbaipublicfiles.com/deit/deit_tiny_patch16_224-a1311bcf.pth" ), "vit_deit_small_patch16_224": _cfg( url="https://dl.fbaipublicfiles.com/deit/deit_small_patch16_224-cd65a155.pth" ), "vit_deit_base_patch16_224": _cfg( url="https://dl.fbaipublicfiles.com/deit/deit_base_patch16_224-b5f2ef4d.pth", ), "vit_deit_base_patch16_384": _cfg( url="https://dl.fbaipublicfiles.com/deit/deit_base_patch16_384-8de9b5d1.pth", input_size=(3, 384, 384), crop_pct=1.0, ), "vit_deit_tiny_distilled_patch16_224": _cfg( url="https://dl.fbaipublicfiles.com/deit/deit_tiny_distilled_patch16_224-b40b3cf7.pth" ), "vit_deit_small_distilled_patch16_224": _cfg( url="https://dl.fbaipublicfiles.com/deit/deit_small_distilled_patch16_224-649709d9.pth" ), "vit_deit_base_distilled_patch16_224": _cfg( url="https://dl.fbaipublicfiles.com/deit/deit_base_distilled_patch16_224-df68dfff.pth", ), "vit_deit_base_distilled_patch16_384": _cfg( url="https://dl.fbaipublicfiles.com/deit/deit_base_distilled_patch16_384-d0272ac0.pth", input_size=(3, 384, 384), crop_pct=1.0, ), } class Mlp(nn.Module): def __init__( self, in_features, hidden_features=None, out_features=None, act_layer=nn.GELU, drop=0.0, ): super().__init__() out_features = out_features or in_features hidden_features = hidden_features or in_features self.fc1 = nn.Linear(in_features, hidden_features) self.act = act_layer() self.fc2 = nn.Linear(hidden_features, out_features) self.drop = nn.Dropout(drop) def forward(self, x): x = self.fc1(x) x = self.act(x) x = self.drop(x) x = self.fc2(x) x = self.drop(x) return x class Attention(nn.Module): def __init__( self, dim, num_heads=8, qkv_bias=False, qk_scale=None, attn_drop=0.0, proj_drop=0.0, ): super().__init__() self.num_heads = num_heads head_dim = dim // num_heads # NOTE scale factor was wrong in my original version, can set manually to be compat with prev weights self.scale = qk_scale or head_dim ** -0.5 self.qkv = nn.Linear(dim, dim * 3, bias=qkv_bias) self.attn_drop = nn.Dropout(attn_drop) self.proj = nn.Linear(dim, dim) self.proj_drop = nn.Dropout(proj_drop) def forward(self, x, mask=None): B, N, C = x.shape qkv = ( self.qkv(x) .reshape(B, N, 3, self.num_heads, C // self.num_heads) .permute(2, 0, 3, 1, 4) ) q, k, v = ( qkv[0], qkv[1], qkv[2], ) # make torchscript happy (cannot use tensor as tuple) attn = (q @ k.transpose(-2, -1)) * self.scale if mask is not None: mask = mask.bool() attn = attn.masked_fill(~mask[:, None, None, :], float("-inf")) attn = attn.softmax(dim=-1) attn = self.attn_drop(attn) x = (attn @ v).transpose(1, 2).reshape(B, N, C) x = self.proj(x) x = self.proj_drop(x) return x, attn class Block(nn.Module): def __init__( self, dim, num_heads, mlp_ratio=4.0, qkv_bias=False, qk_scale=None, drop=0.0, attn_drop=0.0, drop_path=0.0, act_layer=nn.GELU, norm_layer=nn.LayerNorm, ): super().__init__() self.norm1 = norm_layer(dim) self.attn = Attention( dim, num_heads=num_heads, qkv_bias=qkv_bias, qk_scale=qk_scale, attn_drop=attn_drop, proj_drop=drop, ) # NOTE: drop path for stochastic depth, we shall see if this is better than dropout here self.drop_path = DropPath(drop_path) if drop_path > 0.0 else nn.Identity() self.norm2 = norm_layer(dim) mlp_hidden_dim = int(dim * mlp_ratio) self.mlp = Mlp( in_features=dim, hidden_features=mlp_hidden_dim, act_layer=act_layer, drop=drop, ) def forward(self, x, mask=None): _x, attn = self.attn(self.norm1(x), mask=mask) x = x + self.drop_path(_x) x = x + self.drop_path(self.mlp(self.norm2(x))) return x, attn class PatchEmbed(nn.Module): """ Image to Patch Embedding""" def __init__( self, img_size=224, patch_size=16, in_chans=3, embed_dim=768, no_patch_embed_bias=False, ): super().__init__() img_size = to_2tuple(img_size) patch_size = to_2tuple(patch_size) num_patches = (img_size[1] // patch_size[1]) * (img_size[0] // patch_size[0]) self.img_size = img_size self.patch_size = patch_size self.num_patches = num_patches self.proj = nn.Conv2d( in_chans, embed_dim, kernel_size=patch_size, stride=patch_size, bias=False if no_patch_embed_bias else True, ) def forward(self, x): B, C, H, W = x.shape # FIXME look at relaxing size constraints x = self.proj(x) return x class VisionTransformer(nn.Module): """ Vision Transformer A PyTorch impl of : `An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale` - https://arxiv.org/abs/2010.11929 """ def __init__( self, img_size=224, patch_size=16, in_chans=3, num_classes=1000, embed_dim=768, depth=12, num_heads=12, mlp_ratio=4.0, qkv_bias=True, qk_scale=None, representation_size=None, drop_rate=0.0, attn_drop_rate=0.0, drop_path_rate=0.0, norm_layer=None, add_norm_before_transformer=False, no_patch_embed_bias=False, config=None, ): """ Args: img_size (int, tuple): input image size patch_size (int, tuple): patch size in_chans (int): number of input channels num_classes (int): number of classes for classification head embed_dim (int): embedding dimension depth (int): depth of transformer num_heads (int): number of attention heads mlp_ratio (int): ratio of mlp hidden dim to embedding dim qkv_bias (bool): enable bias for qkv if True qk_scale (float): override default qk scale of head_dim ** -0.5 if set representation_size (Optional[int]): enable and set representation layer (pre-logits) to this value if set drop_rate (float): dropout rate attn_drop_rate (float): attention dropout rate drop_path_rate (float): stochastic depth rate hybrid_backbone (nn.Module): CNN backbone to use in-place of PatchEmbed module norm_layer: (nn.Module): normalization layer """ super().__init__() drop_rate = drop_rate if config is None else config["drop_rate"] self.num_classes = num_classes self.num_features = ( self.embed_dim ) = embed_dim # num_features for consistency with other models norm_layer = norm_layer or partial(nn.LayerNorm, eps=1e-6) self.add_norm_before_transformer = add_norm_before_transformer self.patch_embed = PatchEmbed( img_size=img_size, patch_size=patch_size, in_chans=in_chans, embed_dim=embed_dim, ) num_patches = self.patch_embed.num_patches self.patch_size = patch_size self.patch_dim = img_size // patch_size self.cls_token = nn.Parameter(torch.zeros(1, 1, embed_dim)) self.pos_embed = nn.Parameter(torch.zeros(1, num_patches + 1, embed_dim)) self.pos_drop = nn.Dropout(p=drop_rate) if add_norm_before_transformer: self.pre_norm = norm_layer(embed_dim) dpr = [ x.item() for x in torch.linspace(0, drop_path_rate, depth) ] # stochastic depth decay rule self.blocks = nn.ModuleList( [ Block( dim=embed_dim, num_heads=num_heads, mlp_ratio=mlp_ratio, qkv_bias=qkv_bias, qk_scale=qk_scale, drop=drop_rate, attn_drop=attn_drop_rate, drop_path=dpr[i], norm_layer=norm_layer, ) for i in range(depth) ] ) self.norm = norm_layer(embed_dim) trunc_normal_(self.pos_embed, std=0.02) trunc_normal_(self.cls_token, std=0.02) self.apply(self._init_weights) def _init_weights(self, m): if isinstance(m, nn.Linear): trunc_normal_(m.weight, std=0.02) if isinstance(m, nn.Linear) and m.bias is not None: nn.init.constant_(m.bias, 0) elif isinstance(m, nn.LayerNorm): nn.init.constant_(m.bias, 0) nn.init.constant_(m.weight, 1.0) @torch.jit.ignore def no_weight_decay(self): return {"pos_embed", "cls_token"} def mask_tokens(self, orig_image, feats): """ Prepare masked tokens inputs/labels for masked patch prediction: 80% MASK, 10% random, 10% original. """ img_unnorm = orig_image * 0.5 + 0.5 _, _, ph, pw = self.patch_embed.proj.weight.shape with torch.no_grad(): img_unnorm_patch = F.conv2d( img_unnorm, weight=torch.ones(3, 1, ph, pw).to(img_unnorm) / (ph * pw), bias=None, stride=(ph, pw), padding=0, groups=3, ) labels = ( ((img_unnorm_patch * 255).long().flatten(start_dim=2, end_dim=3)) .permute(0, 2, 1) .contiguous() ) # We sample a few tokens in each sequence for MLM training (with probability `self.mlm_probability`) probability_matrix = torch.full(labels.shape[:-1], 0.15) masked_indices = torch.bernoulli(probability_matrix).bool() labels[~masked_indices] = -100 # We only compute loss on masked tokens # 80% of the time, we replace masked input tokens with tokenizer.mask_token ([MASK]) indices_replaced = ( torch.bernoulli(torch.full(labels.shape[:-1], 0.8)).bool() & masked_indices ) feats[indices_replaced] = self.mask_token.to(feats) return feats, labels def visual_embed(self, _x, max_image_len=200, mask_it=False): _, _, ph, pw = self.patch_embed.proj.weight.shape x = self.patch_embed(_x) x_mask = (_x.sum(dim=1) != 0).float()[:, None, :, :] x_mask = F.interpolate(x_mask, size=(x.shape[2], x.shape[3])).long() x_h = x_mask[:, 0].sum(dim=1)[:, 0] x_w = x_mask[:, 0].sum(dim=2)[:, 0] B, C, H, W = x.shape spatial_pos = ( self.pos_embed[:, 1:, :] .transpose(1, 2) .view(1, C, self.patch_dim, self.patch_dim) ) pos_embed = torch.cat( [ F.pad( F.interpolate( spatial_pos, size=(h, w), mode="bilinear", align_corners=True, ), (0, W - w, 0, H - h), ) for h, w in zip(x_h, x_w) ], dim=0, ) pos_embed = pos_embed.flatten(2).transpose(1, 2) x = x.flatten(2).transpose(1, 2) patch_index = ( torch.stack( torch.meshgrid( torch.arange(x_mask.shape[-2]), torch.arange(x_mask.shape[-1]) ), dim=-1, )[None, None, :, :, :] .expand(x_mask.shape[0], x_mask.shape[1], -1, -1, -1) .flatten(1, 3) ) x_mask = x_mask.flatten(1) if mask_it: x, label = self.mask_tokens(_x, x) if ( max_image_len < 0 or max_image_len is None or not isinstance(max_image_len, int) ): # suppose aug is 800 x 1333, then, maximum effective res is 800 x 1333 (if one side gets bigger, the other will be constrained and be shrinked) # (800 // self.patch_size) * (1333 // self.patch_size) is the maximum number of patches that single image can get. # if self.patch_size = 32, 25 * 41 = 1025 # if res is 384 x 640, 12 * 20 = 240 eff = x_h * x_w max_image_len = eff.max() else: eff = x_h * x_w max_image_len = min(eff.max(), max_image_len) valid_idx = x_mask.nonzero(as_tuple=False) non_valid_idx = (1 - x_mask).nonzero(as_tuple=False) unique_rows = valid_idx[:, 0].unique() valid_row_idx = [valid_idx[valid_idx[:, 0] == u] for u in unique_rows] non_valid_row_idx = [ non_valid_idx[non_valid_idx[:, 0] == u] for u in unique_rows ] valid_nums = [v.size(0) for v in valid_row_idx] non_valid_nums = [v.size(0) for v in non_valid_row_idx] pad_nums = [max_image_len - v for v in valid_nums] select = list() for i, (v, nv, p) in enumerate(zip(valid_nums, non_valid_nums, pad_nums)): if p <= 0: valid_choice = torch.multinomial(torch.ones(v).float(), max_image_len) select.append(valid_row_idx[i][valid_choice]) else: pad_choice = torch.multinomial( torch.ones(nv).float(), p, replacement=True ) select.append( torch.cat( [valid_row_idx[i], non_valid_row_idx[i][pad_choice]], dim=0, ) ) select = torch.cat(select, dim=0) x = x[select[:, 0], select[:, 1]].view(B, -1, C) x_mask = x_mask[select[:, 0], select[:, 1]].view(B, -1) patch_index = patch_index[select[:, 0], select[:, 1]].view(B, -1, 2) pos_embed = pos_embed[select[:, 0], select[:, 1]].view(B, -1, C) if mask_it: label = label[select[:, 0], select[:, 1]].view(B, -1, 3) label[x_mask == 0] = -100 label = torch.cat( [torch.full((label.shape[0], 1, 3), -100).to(label), label,], dim=1, ) cls_tokens = self.cls_token.expand(B, -1, -1) x = torch.cat((cls_tokens, x), dim=1) pos_embed = torch.cat( (self.pos_embed[:, 0, :][:, None, :].expand(B, -1, -1), pos_embed), dim=1 ) x = x + pos_embed x = self.pos_drop(x) if self.add_norm_before_transformer: x = self.pre_norm(x) x_mask = torch.cat([torch.ones(x_mask.shape[0], 1).to(x_mask), x_mask], dim=1) if mask_it: return x, x_mask, (patch_index, (H, W)), label else: return x, x_mask, (patch_index, (H, W)), None def forward_features(self, _x, max_image_len=144, mask_it=False): x, x_mask, patch_index, label = self.visual_embed( _x, max_image_len=max_image_len, mask_it=mask_it ) for blk in self.blocks: x, _ = blk(x, mask=x_mask) x = self.norm(x) return x, x_mask, label def forward(self, x, max_image_len=-1): x, _, _ = self.forward_features(x, max_image_len=max_image_len) x = x[:, 0] x = self.head(x) return x class DistilledVisionTransformer(VisionTransformer): """ Vision Transformer with distillation token. Paper: `Training data-efficient image transformers & distillation through attention` - https://arxiv.org/abs/2012.12877 This impl of distilled ViT is taken from https://github.com/facebookresearch/deit """ def __init__(self, *args, **kwargs): super().__init__(*args, **kwargs) self.dist_token = nn.Parameter(torch.zeros(1, 1, self.embed_dim)) num_patches = self.patch_embed.num_patches self.pos_embed = nn.Parameter(torch.zeros(1, num_patches + 2, self.embed_dim)) trunc_normal_(self.dist_token, std=0.02) trunc_normal_(self.pos_embed, std=0.02) def visual_embed(self, _x, max_image_len=200, mask_it=False): _, _, ph, pw = self.patch_embed.proj.weight.shape x = self.patch_embed(_x) x_mask = (_x.sum(dim=1) != 0).float()[:, None, :, :] x_mask = F.interpolate(x_mask, size=(x.shape[2], x.shape[3])).long() x_h = x_mask[:, 0].sum(dim=1)[:, 0] x_w = x_mask[:, 0].sum(dim=2)[:, 0] B, C, H, W = x.shape spatial_pos = ( self.pos_embed[:, 2:, :] .transpose(1, 2) .view(1, C, self.patch_dim, self.patch_dim) ) pos_embed = torch.cat( [ F.pad( F.interpolate( spatial_pos, size=(h, w), mode="bilinear", align_corners=True, ), (0, W - w, 0, H - h), ) for h, w in zip(x_h, x_w) ], dim=0, ) pos_embed = pos_embed.flatten(2).transpose(1, 2) x = x.flatten(2).transpose(1, 2) patch_index = ( torch.stack( torch.meshgrid( torch.arange(x_mask.shape[-2]), torch.arange(x_mask.shape[-1]) ), dim=-1, )[None, None, :, :, :] .expand(x_mask.shape[0], x_mask.shape[1], -1, -1, -1) .flatten(1, 3) ) x_mask = x_mask.flatten(1) if mask_it: x, label = self.mask_tokens(_x, x) if ( max_image_len < 0 or max_image_len is None or not isinstance(max_image_len, int) ): # suppose aug is 800 x 1333, then, maximum effective res is 800 x 1333 (if one side gets bigger, the other will be constrained and be shrinked) # (800 // self.patch_size) * (1333 // self.patch_size) is the maximum number of patches that single image can get. # if self.patch_size = 32, 25 * 41 = 1025 # if res is 384 x 640, 12 * 20 = 240 eff = x_h * x_w max_image_len = eff.max() else: eff = x_h * x_w max_image_len = min(eff.max(), max_image_len) valid_idx = x_mask.nonzero(as_tuple=False) non_valid_idx = (1 - x_mask).nonzero(as_tuple=False) unique_rows = valid_idx[:, 0].unique() valid_row_idx = [valid_idx[valid_idx[:, 0] == u] for u in unique_rows] non_valid_row_idx = [ non_valid_idx[non_valid_idx[:, 0] == u] for u in unique_rows ] valid_nums = [v.size(0) for v in valid_row_idx] non_valid_nums = [v.size(0) for v in non_valid_row_idx] pad_nums = [max_image_len - v for v in valid_nums] select = list() for i, (v, nv, p) in enumerate(zip(valid_nums, non_valid_nums, pad_nums)): if p <= 0: valid_choice = torch.multinomial(torch.ones(v).float(), max_image_len) select.append(valid_row_idx[i][valid_choice]) else: pad_choice = torch.multinomial( torch.ones(nv).float(), p, replacement=True ) select.append( torch.cat( [valid_row_idx[i], non_valid_row_idx[i][pad_choice]], dim=0, ) ) select = torch.cat(select, dim=0) x = x[select[:, 0], select[:, 1]].view(B, -1, C) x_mask = x_mask[select[:, 0], select[:, 1]].view(B, -1) patch_index = patch_index[select[:, 0], select[:, 1]].view(B, -1, 2) pos_embed = pos_embed[select[:, 0], select[:, 1]].view(B, -1, C) if mask_it: label = label[select[:, 0], select[:, 1]].view(B, -1, 3) label[x_mask == 0] = -100 label = torch.cat( [torch.full((label.shape[0], 1, 3), -100).to(label), label,], dim=1, ) cls_tokens = self.cls_token.expand(B, -1, -1) dist_token = self.dist_token.expand(B, -1, -1) x = torch.cat((cls_tokens, dist_token, x), dim=1) pos_embed = torch.cat( (self.pos_embed[:, :2, :].expand(B, -1, -1), pos_embed), dim=1 ) x = x + pos_embed x = self.pos_drop(x) if self.add_norm_before_transformer: x = self.pre_norm(x) x_mask = torch.cat([torch.ones(x_mask.shape[0], 2).to(x_mask), x_mask], dim=1) if mask_it: return x, x_mask, (patch_index, (H, W)), label else: return x, x_mask, (patch_index, (H, W)), None def forward_features(self, _x, max_image_len=144, mask_it=False): x, x_mask, patch_index, label = self.visual_embed( _x, max_image_len=max_image_len, mask_it=mask_it ) for blk in self.blocks: x, _ = blk(x, mask=x_mask) x = self.norm(x) return x, x_mask, label def forward(self, x, max_image_len=-1): x, _, _ = self.forward_features(x, max_image_len=max_image_len) x = x[:, 0] x = self.head(x) return x def resize_pos_embed(posemb, posemb_new): # Rescale the grid of position embeddings when loading from state_dict. Adapted from # https://github.com/google-research/vision_transformer/blob/00883dd691c63a6830751563748663526e811cee/vit_jax/checkpoint.py#L224 _logger.info("Resized position embedding: %s to %s", posemb.shape, posemb_new.shape) ntok_new = posemb_new.shape[1] if True: posemb_tok, posemb_grid = posemb[:, :1], posemb[0, 1:] ntok_new -= 1 else: posemb_tok, posemb_grid = posemb[:, :0], posemb[0] gs_old = int(math.sqrt(len(posemb_grid))) gs_new = int(math.sqrt(ntok_new)) _logger.info("Position embedding grid-size from %s to %s", gs_old, gs_new) posemb_grid = posemb_grid.reshape(1, gs_old, gs_old, -1).permute(0, 3, 1, 2) posemb_grid = F.interpolate(posemb_grid, size=(gs_new, gs_new), mode="bilinear") posemb_grid = posemb_grid.permute(0, 2, 3, 1).reshape(1, gs_new * gs_new, -1) posemb = torch.cat([posemb_tok, posemb_grid], dim=1) return posemb def checkpoint_filter_fn(state_dict, model): """ convert patch embedding weight from manual patchify + linear proj to conv""" out_dict = {} if "model" in state_dict: # For deit models state_dict = state_dict["model"] for k, v in state_dict.items(): if "patch_embed.proj.weight" in k and len(v.shape) < 4: # For old models that I trained prior to conv based patchification O, I, H, W = model.patch_embed.proj.weight.shape v = v.reshape(O, -1, H, W) elif k == "pos_embed" and v.shape != model.pos_embed.shape: # To resize pos embedding when using model at different size from pretrained weights v = resize_pos_embed(v, model.pos_embed) out_dict[k] = v return out_dict def _create_vision_transformer(variant, pretrained=False, distilled=False, **kwargs): default_cfg = default_cfgs[variant] default_num_classes = default_cfg["num_classes"] default_img_size = default_cfg["input_size"][-1] num_classes = kwargs.pop("num_classes", default_num_classes) img_size = kwargs.pop("img_size", default_img_size) repr_size = kwargs.pop("representation_size", None) if repr_size is not None and num_classes != default_num_classes: # Remove representation layer if fine-tuning. This may not always be the desired action, # but I feel better than doing nothing by default for fine-tuning. Perhaps a better interface? _logger.warning("Removing representation layer for fine-tuning.") repr_size = None model_cls = DistilledVisionTransformer if distilled else VisionTransformer model = model_cls( img_size=img_size, num_classes=num_classes, representation_size=repr_size, **kwargs, ) model.default_cfg = default_cfg if pretrained: load_pretrained( model, num_classes=num_classes, in_chans=kwargs.get("in_chans", 3), filter_fn=partial(checkpoint_filter_fn, model=model), strict=False, ) return model @register_model def vit_small_patch16_224(pretrained=False, **kwargs): """ My custom 'small' ViT model. Depth=8, heads=8= mlp_ratio=3.""" model_kwargs = dict( patch_size=16, embed_dim=768, depth=8, num_heads=8, mlp_ratio=3.0, qkv_bias=False, norm_layer=nn.LayerNorm, **kwargs, ) if pretrained: # NOTE my scale was wrong for original weights, leaving this here until I have better ones for this model model_kwargs.setdefault("qk_scale", 768 ** -0.5) model = _create_vision_transformer( "vit_small_patch16_224", pretrained=pretrained, **model_kwargs ) return model @register_model def vit_base_patch16_224(pretrained=False, **kwargs): """ ViT-Base (ViT-B/16) from original paper (https://arxiv.org/abs/2010.11929). ImageNet-1k weights fine-tuned from in21k @ 224x224, source https://github.com/google-research/vision_transformer. """ model_kwargs = dict(patch_size=16, embed_dim=768, depth=12, num_heads=12, **kwargs) model = _create_vision_transformer( "vit_base_patch16_224", pretrained=pretrained, **model_kwargs ) return model @register_model def vit_base_patch32_224(pretrained=False, **kwargs): """ ViT-Base (ViT-B/32) from original paper (https://arxiv.org/abs/2010.11929). No pretrained weights. """ model_kwargs = dict(patch_size=32, embed_dim=768, depth=12, num_heads=12, **kwargs) model = _create_vision_transformer( "vit_base_patch32_224", pretrained=pretrained, **model_kwargs ) return model @register_model def vit_base_patch16_384(pretrained=False, **kwargs): """ ViT-Base model (ViT-B/16) from original paper (https://arxiv.org/abs/2010.11929). ImageNet-1k weights fine-tuned from in21k @ 384x384, source https://github.com/google-research/vision_transformer. """ model_kwargs = dict(patch_size=16, embed_dim=768, depth=12, num_heads=12, **kwargs) model = _create_vision_transformer( "vit_base_patch16_384", pretrained=pretrained, **model_kwargs ) return model @register_model def vit_base_patch32_384(pretrained=False, **kwargs): """ ViT-Base model (ViT-B/32) from original paper (https://arxiv.org/abs/2010.11929). ImageNet-1k weights fine-tuned from in21k @ 384x384, source https://github.com/google-research/vision_transformer. """ model_kwargs = dict(patch_size=32, embed_dim=768, depth=12, num_heads=12, **kwargs) model = _create_vision_transformer( "vit_base_patch32_384", pretrained=pretrained, **model_kwargs ) return model @register_model def vit_large_patch16_224(pretrained=False, **kwargs): """ ViT-Large model (ViT-L/32) from original paper (https://arxiv.org/abs/2010.11929). ImageNet-1k weights fine-tuned from in21k @ 224x224, source https://github.com/google-research/vision_transformer. """ model_kwargs = dict(patch_size=16, embed_dim=1024, depth=24, num_heads=16, **kwargs) model = _create_vision_transformer( "vit_large_patch16_224", pretrained=pretrained, **model_kwargs ) return model @register_model def vit_large_patch32_224(pretrained=False, **kwargs): """ ViT-Large model (ViT-L/32) from original paper (https://arxiv.org/abs/2010.11929). No pretrained weights. """ model_kwargs = dict(patch_size=32, embed_dim=1024, depth=24, num_heads=16, **kwargs) model = _create_vision_transformer( "vit_large_patch32_224", pretrained=pretrained, **model_kwargs ) return model @register_model def vit_large_patch16_384(pretrained=False, **kwargs): """ ViT-Large model (ViT-L/16) from original paper (https://arxiv.org/abs/2010.11929). ImageNet-1k weights fine-tuned from in21k @ 384x384, source https://github.com/google-research/vision_transformer. """ model_kwargs = dict(patch_size=16, embed_dim=1024, depth=24, num_heads=16, **kwargs) model = _create_vision_transformer( "vit_large_patch16_384", pretrained=pretrained, **model_kwargs ) return model @register_model def vit_large_patch32_384(pretrained=False, **kwargs): """ ViT-Large model (ViT-L/32) from original paper (https://arxiv.org/abs/2010.11929). ImageNet-1k weights fine-tuned from in21k @ 384x384, source https://github.com/google-research/vision_transformer. """ model_kwargs = dict(patch_size=32, embed_dim=1024, depth=24, num_heads=16, **kwargs) model = _create_vision_transformer( "vit_large_patch32_384", pretrained=pretrained, **model_kwargs ) return model @register_model def vit_base_patch16_224_in21k(pretrained=False, **kwargs): """ ViT-Base model (ViT-B/16) from original paper (https://arxiv.org/abs/2010.11929). ImageNet-21k weights @ 224x224, source https://github.com/google-research/vision_transformer. """ model_kwargs = dict( patch_size=16, embed_dim=768, depth=12, num_heads=12, representation_size=768, **kwargs, ) model = _create_vision_transformer( "vit_base_patch16_224_in21k", pretrained=pretrained, **model_kwargs ) return model @register_model def vit_base_patch32_224_in21k(pretrained=False, **kwargs): """ ViT-Base model (ViT-B/32) from original paper (https://arxiv.org/abs/2010.11929). ImageNet-21k weights @ 224x224, source https://github.com/google-research/vision_transformer. """ model_kwargs = dict( patch_size=32, embed_dim=768, depth=12, num_heads=12, representation_size=768, **kwargs, ) model = _create_vision_transformer( "vit_base_patch32_224_in21k", pretrained=pretrained, **model_kwargs ) return model @register_model def vit_large_patch16_224_in21k(pretrained=False, **kwargs): """ ViT-Large model (ViT-L/16) from original paper (https://arxiv.org/abs/2010.11929). ImageNet-21k weights @ 224x224, source https://github.com/google-research/vision_transformer. """ model_kwargs = dict( patch_size=16, embed_dim=1024, depth=24, num_heads=16, representation_size=1024, **kwargs, ) model = _create_vision_transformer( "vit_large_patch16_224_in21k", pretrained=pretrained, **model_kwargs ) return model @register_model def vit_large_patch32_224_in21k(pretrained=False, **kwargs): """ ViT-Large model (ViT-L/32) from original paper (https://arxiv.org/abs/2010.11929). ImageNet-21k weights @ 224x224, source https://github.com/google-research/vision_transformer. """ model_kwargs = dict( patch_size=32, embed_dim=1024, depth=24, num_heads=16, representation_size=1024, **kwargs, ) model = _create_vision_transformer( "vit_large_patch32_224_in21k", pretrained=pretrained, **model_kwargs ) return model @register_model def vit_huge_patch14_224_in21k(pretrained=False, **kwargs): """ ViT-Huge model (ViT-H/14) from original paper (https://arxiv.org/abs/2010.11929). ImageNet-21k weights @ 224x224, source https://github.com/google-research/vision_transformer. NOTE: converted weights not currently available, too large for github release hosting. """ model_kwargs = dict( patch_size=14, embed_dim=1280, depth=32, num_heads=16, representation_size=1280, **kwargs, ) model = _create_vision_transformer( "vit_huge_patch14_224_in21k", pretrained=pretrained, **model_kwargs ) return model @register_model def vit_base_resnet50_224_in21k(pretrained=False, **kwargs): """ R50+ViT-B/16 hybrid model from original paper (https://arxiv.org/abs/2010.11929). ImageNet-21k weights @ 224x224, source https://github.com/google-research/vision_transformer. """ # create a ResNetV2 w/o pre-activation, that uses StdConv and GroupNorm and has 3 stages, no head backbone = ResNetV2( layers=(3, 4, 9), num_classes=0, global_pool="", in_chans=kwargs.get("in_chans", 3), preact=False, stem_type="same", conv_layer=StdConv2dSame, ) model_kwargs = dict( embed_dim=768, depth=12, num_heads=12, hybrid_backbone=backbone, representation_size=768, **kwargs, ) model = _create_vision_transformer( "vit_base_resnet50_224_in21k", pretrained=pretrained, **model_kwargs ) return model @register_model def vit_base_resnet50_384(pretrained=False, **kwargs): """ R50+ViT-B/16 hybrid from original paper (https://arxiv.org/abs/2010.11929). ImageNet-1k weights fine-tuned from in21k @ 384x384, source https://github.com/google-research/vision_transformer. """ # create a ResNetV2 w/o pre-activation, that uses StdConv and GroupNorm and has 3 stages, no head backbone = ResNetV2( layers=(3, 4, 9), num_classes=0, global_pool="", in_chans=kwargs.get("in_chans", 3), preact=False, stem_type="same", conv_layer=StdConv2dSame, ) model_kwargs = dict( embed_dim=768, depth=12, num_heads=12, hybrid_backbone=backbone, **kwargs ) model = _create_vision_transformer( "vit_base_resnet50_384", pretrained=pretrained, **model_kwargs ) return model @register_model def vit_small_resnet26d_224(pretrained=False, **kwargs): """ Custom ViT small hybrid w/ ResNet26D stride 32. No pretrained weights. """ backbone = resnet26d( pretrained=pretrained, in_chans=kwargs.get("in_chans", 3), features_only=True, out_indices=[4], ) model_kwargs = dict( embed_dim=768, depth=8, num_heads=8, mlp_ratio=3, hybrid_backbone=backbone, **kwargs, ) model = _create_vision_transformer( "vit_small_resnet26d_224", pretrained=pretrained, **model_kwargs ) return model @register_model def vit_small_resnet50d_s3_224(pretrained=False, **kwargs): """ Custom ViT small hybrid w/ ResNet50D 3-stages, stride 16. No pretrained weights. """ backbone = resnet50d( pretrained=pretrained, in_chans=kwargs.get("in_chans", 3), features_only=True, out_indices=[3], ) model_kwargs = dict( embed_dim=768, depth=8, num_heads=8, mlp_ratio=3, hybrid_backbone=backbone, **kwargs, ) model = _create_vision_transformer( "vit_small_resnet50d_s3_224", pretrained=pretrained, **model_kwargs ) return model @register_model def vit_base_resnet26d_224(pretrained=False, **kwargs): """ Custom ViT base hybrid w/ ResNet26D stride 32. No pretrained weights. """ backbone = resnet26d( pretrained=pretrained, in_chans=kwargs.get("in_chans", 3), features_only=True, out_indices=[4], ) model_kwargs = dict( embed_dim=768, depth=12, num_heads=12, hybrid_backbone=backbone, **kwargs ) model = _create_vision_transformer( "vit_base_resnet26d_224", pretrained=pretrained, **model_kwargs ) return model @register_model def vit_base_resnet50d_224(pretrained=False, **kwargs): """ Custom ViT base hybrid w/ ResNet50D stride 32. No pretrained weights. """ backbone = resnet50d( pretrained=pretrained, in_chans=kwargs.get("in_chans", 3), features_only=True, out_indices=[4], ) model_kwargs = dict( embed_dim=768, depth=12, num_heads=12, hybrid_backbone=backbone, **kwargs ) model = _create_vision_transformer( "vit_base_resnet50d_224", pretrained=pretrained, **model_kwargs ) return model @register_model def vit_deit_tiny_patch16_224(pretrained=False, **kwargs): """ DeiT-tiny model @ 224x224 from paper (https://arxiv.org/abs/2012.12877). ImageNet-1k weights from https://github.com/facebookresearch/deit. """ model_kwargs = dict(patch_size=16, embed_dim=192, depth=12, num_heads=3, **kwargs) model = _create_vision_transformer( "vit_deit_tiny_patch16_224", pretrained=pretrained, **model_kwargs ) return model @register_model def vit_deit_small_patch16_224(pretrained=False, **kwargs): """ DeiT-small model @ 224x224 from paper (https://arxiv.org/abs/2012.12877). ImageNet-1k weights from https://github.com/facebookresearch/deit. """ model_kwargs = dict(patch_size=16, embed_dim=384, depth=12, num_heads=6, **kwargs) model = _create_vision_transformer( "vit_deit_small_patch16_224", pretrained=pretrained, **model_kwargs ) return model @register_model def vit_deit_base_patch16_224(pretrained=False, **kwargs): """ DeiT base model @ 224x224 from paper (https://arxiv.org/abs/2012.12877). ImageNet-1k weights from https://github.com/facebookresearch/deit. """ model_kwargs = dict(patch_size=16, embed_dim=768, depth=12, num_heads=12, **kwargs) model = _create_vision_transformer( "vit_deit_base_patch16_224", pretrained=pretrained, **model_kwargs ) return model @register_model def vit_deit_base_patch16_384(pretrained=False, **kwargs): """ DeiT base model @ 384x384 from paper (https://arxiv.org/abs/2012.12877). ImageNet-1k weights from https://github.com/facebookresearch/deit. """ model_kwargs = dict(patch_size=16, embed_dim=768, depth=12, num_heads=12, **kwargs) model = _create_vision_transformer( "vit_deit_base_patch16_384", pretrained=pretrained, **model_kwargs ) return model @register_model def vit_deit_tiny_distilled_patch16_224(pretrained=False, **kwargs): """ DeiT-tiny distilled model @ 224x224 from paper (https://arxiv.org/abs/2012.12877). ImageNet-1k weights from https://github.com/facebookresearch/deit. """ model_kwargs = dict(patch_size=16, embed_dim=192, depth=12, num_heads=3, **kwargs) model = _create_vision_transformer( "vit_deit_tiny_distilled_patch16_224", pretrained=pretrained, distilled=True, **model_kwargs, ) return model @register_model def vit_deit_small_distilled_patch16_224(pretrained=False, **kwargs): """ DeiT-small distilled model @ 224x224 from paper (https://arxiv.org/abs/2012.12877). ImageNet-1k weights from https://github.com/facebookresearch/deit. """ model_kwargs = dict(patch_size=16, embed_dim=384, depth=12, num_heads=6, **kwargs) model = _create_vision_transformer( "vit_deit_small_distilled_patch16_224", pretrained=pretrained, distilled=True, **model_kwargs, ) return model @register_model def vit_deit_base_distilled_patch16_224(pretrained=False, **kwargs): """ DeiT-base distilled model @ 224x224 from paper (https://arxiv.org/abs/2012.12877). ImageNet-1k weights from https://github.com/facebookresearch/deit. """ model_kwargs = dict(patch_size=16, embed_dim=768, depth=12, num_heads=12, **kwargs) model = _create_vision_transformer( "vit_deit_base_distilled_patch16_224", pretrained=pretrained, distilled=True, **model_kwargs, ) return model @register_model def vit_deit_base_distilled_patch16_384(pretrained=False, **kwargs): """ DeiT-base distilled model @ 384x384 from paper (https://arxiv.org/abs/2012.12877). ImageNet-1k weights from https://github.com/facebookresearch/deit. """ model_kwargs = dict(patch_size=16, embed_dim=768, depth=12, num_heads=12, **kwargs) model = _create_vision_transformer( "vit_deit_base_distilled_patch16_384", pretrained=pretrained, distilled=True, **model_kwargs, ) return model
49,034
34.558376
155
py
Mr.Right
Mr.Right-main/models/ViLT/vilt/modules/heads.py
import torch import torch.nn as nn import torch.nn.functional as F from transformers.models.bert.modeling_bert import BertPredictionHeadTransform class Pooler(nn.Module): def __init__(self, hidden_size): super().__init__() self.dense = nn.Linear(hidden_size, hidden_size) self.activation = nn.Tanh() def forward(self, hidden_states): first_token_tensor = hidden_states[:, 0] pooled_output = self.dense(first_token_tensor) pooled_output = self.activation(pooled_output) return pooled_output class ITMHead(nn.Module): def __init__(self, hidden_size): super().__init__() self.fc = nn.Linear(hidden_size, 2) def forward(self, x): x = self.fc(x) return x class MLMHead(nn.Module): def __init__(self, config, weight=None): super().__init__() self.transform = BertPredictionHeadTransform(config) self.decoder = nn.Linear(config.hidden_size, config.vocab_size, bias=False) self.bias = nn.Parameter(torch.zeros(config.vocab_size)) if weight is not None: self.decoder.weight = weight def forward(self, x): x = self.transform(x) x = self.decoder(x) + self.bias return x class MPPHead(nn.Module): def __init__(self, config): super().__init__() self.transform = BertPredictionHeadTransform(config) self.decoder = nn.Linear(config.hidden_size, 256 * 3) def forward(self, x): x = self.transform(x) x = self.decoder(x) return x
1,569
27.035714
83
py
Mr.Right
Mr.Right-main/models/ViLT/vilt/modules/__init__.py
from .vilt_module import ViLTransformerSS
42
20.5
41
py
Mr.Right
Mr.Right-main/models/ViLT/vilt/transforms/utils.py
from torchvision import transforms from PIL import Image class MinMaxResize: def __init__(self, shorter=800, longer=1333): self.min = shorter self.max = longer def __call__(self, x): w, h = x.size scale = self.min / min(w, h) if h < w: newh, neww = self.min, scale * w else: newh, neww = scale * h, self.min if max(newh, neww) > self.max: scale = self.max / max(newh, neww) newh = newh * scale neww = neww * scale newh, neww = int(newh + 0.5), int(neww + 0.5) newh, neww = newh // 32 * 32, neww // 32 * 32 return x.resize((neww, newh), resample=Image.BICUBIC) class UnNormalize(object): def __init__(self, mean, std): self.mean = mean self.std = std def __call__(self, tensor): """ Args: tensor (Tensor): Tensor image of size (C, H, W) to be normalized. Returns: Tensor: Normalized image. """ for t, m, s in zip(tensor, self.mean, self.std): t.mul_(s).add_(m) # The normalize code -> t.sub_(m).div_(s) return tensor # This is simple maximum entropy normalization performed in Inception paper inception_normalize = transforms.Compose( [transforms.Normalize(mean=[0.5, 0.5, 0.5], std=[0.5, 0.5, 0.5])] ) # ViT uses simple non-biased inception normalization # https://github.com/google-research/vision_transformer/blob/master/vit_jax/input_pipeline.py#L132 inception_unnormalize = transforms.Compose( [UnNormalize(mean=[0.5, 0.5, 0.5], std=[0.5, 0.5, 0.5])] )
1,645
27.877193
98
py
Mr.Right
Mr.Right-main/models/ViLT/vilt/transforms/randaug.py
# code in this file is adpated from rpmcruz/autoaugment # https://github.com/rpmcruz/autoaugment/blob/master/transformations.py import random import PIL, PIL.ImageOps, PIL.ImageEnhance, PIL.ImageDraw import numpy as np import torch from PIL import Image def ShearX(img, v): # [-0.3, 0.3] assert -0.3 <= v <= 0.3 if random.random() > 0.5: v = -v return img.transform(img.size, PIL.Image.AFFINE, (1, v, 0, 0, 1, 0)) def ShearY(img, v): # [-0.3, 0.3] assert -0.3 <= v <= 0.3 if random.random() > 0.5: v = -v return img.transform(img.size, PIL.Image.AFFINE, (1, 0, 0, v, 1, 0)) def TranslateX(img, v): # [-150, 150] => percentage: [-0.45, 0.45] assert -0.45 <= v <= 0.45 if random.random() > 0.5: v = -v v = v * img.size[0] return img.transform(img.size, PIL.Image.AFFINE, (1, 0, v, 0, 1, 0)) def TranslateXabs(img, v): # [-150, 150] => percentage: [-0.45, 0.45] assert 0 <= v if random.random() > 0.5: v = -v return img.transform(img.size, PIL.Image.AFFINE, (1, 0, v, 0, 1, 0)) def TranslateY(img, v): # [-150, 150] => percentage: [-0.45, 0.45] assert -0.45 <= v <= 0.45 if random.random() > 0.5: v = -v v = v * img.size[1] return img.transform(img.size, PIL.Image.AFFINE, (1, 0, 0, 0, 1, v)) def TranslateYabs(img, v): # [-150, 150] => percentage: [-0.45, 0.45] assert 0 <= v if random.random() > 0.5: v = -v return img.transform(img.size, PIL.Image.AFFINE, (1, 0, 0, 0, 1, v)) def Rotate(img, v): # [-30, 30] assert -30 <= v <= 30 if random.random() > 0.5: v = -v return img.rotate(v) def AutoContrast(img, _): return PIL.ImageOps.autocontrast(img) def Invert(img, _): return PIL.ImageOps.invert(img) def Equalize(img, _): return PIL.ImageOps.equalize(img) def Flip(img, _): # not from the paper return PIL.ImageOps.mirror(img) def Solarize(img, v): # [0, 256] assert 0 <= v <= 256 return PIL.ImageOps.solarize(img, v) def SolarizeAdd(img, addition=0, threshold=128): img_np = np.array(img).astype(np.int) img_np = img_np + addition img_np = np.clip(img_np, 0, 255) img_np = img_np.astype(np.uint8) img = Image.fromarray(img_np) return PIL.ImageOps.solarize(img, threshold) def Posterize(img, v): # [4, 8] v = int(v) v = max(1, v) return PIL.ImageOps.posterize(img, v) def Contrast(img, v): # [0.1,1.9] assert 0.1 <= v <= 1.9 return PIL.ImageEnhance.Contrast(img).enhance(v) def Color(img, v): # [0.1,1.9] assert 0.1 <= v <= 1.9 return PIL.ImageEnhance.Color(img).enhance(v) def Brightness(img, v): # [0.1,1.9] assert 0.1 <= v <= 1.9 return PIL.ImageEnhance.Brightness(img).enhance(v) def Sharpness(img, v): # [0.1,1.9] assert 0.1 <= v <= 1.9 return PIL.ImageEnhance.Sharpness(img).enhance(v) def Cutout(img, v): # [0, 60] => percentage: [0, 0.2] assert 0.0 <= v <= 0.2 if v <= 0.0: return img v = v * img.size[0] return CutoutAbs(img, v) def CutoutAbs(img, v): # [0, 60] => percentage: [0, 0.2] # assert 0 <= v <= 20 if v < 0: return img w, h = img.size x0 = np.random.uniform(w) y0 = np.random.uniform(h) x0 = int(max(0, x0 - v / 2.0)) y0 = int(max(0, y0 - v / 2.0)) x1 = min(w, x0 + v) y1 = min(h, y0 + v) xy = (x0, y0, x1, y1) color = (125, 123, 114) # color = (0, 0, 0) img = img.copy() PIL.ImageDraw.Draw(img).rectangle(xy, color) return img def SamplePairing(imgs): # [0, 0.4] def f(img1, v): i = np.random.choice(len(imgs)) img2 = PIL.Image.fromarray(imgs[i]) return PIL.Image.blend(img1, img2, v) return f def Identity(img, v): return img def augment_list(): # 16 oeprations and their ranges # https://github.com/google-research/uda/blob/master/image/randaugment/policies.py#L57 # l = [ # (Identity, 0., 1.0), # (ShearX, 0., 0.3), # 0 # (ShearY, 0., 0.3), # 1 # (TranslateX, 0., 0.33), # 2 # (TranslateY, 0., 0.33), # 3 # (Rotate, 0, 30), # 4 # (AutoContrast, 0, 1), # 5 # (Invert, 0, 1), # 6 # (Equalize, 0, 1), # 7 # (Solarize, 0, 110), # 8 # (Posterize, 4, 8), # 9 # # (Contrast, 0.1, 1.9), # 10 # (Color, 0.1, 1.9), # 11 # (Brightness, 0.1, 1.9), # 12 # (Sharpness, 0.1, 1.9), # 13 # # (Cutout, 0, 0.2), # 14 # # (SamplePairing(imgs), 0, 0.4), # 15 # ] # https://github.com/tensorflow/tpu/blob/8462d083dd89489a79e3200bcc8d4063bf362186/models/official/efficientnet/autoaugment.py#L505 l = [ (AutoContrast, 0, 1), (Equalize, 0, 1), # (Invert, 0, 1), (Rotate, 0, 30), (Posterize, 0, 4), (Solarize, 0, 256), (SolarizeAdd, 0, 110), (Color, 0.1, 1.9), (Contrast, 0.1, 1.9), (Brightness, 0.1, 1.9), (Sharpness, 0.1, 1.9), (ShearX, 0.0, 0.3), (ShearY, 0.0, 0.3), # (CutoutAbs, 0, 40), (TranslateXabs, 0.0, 100), (TranslateYabs, 0.0, 100), ] return l class Lighting(object): """Lighting noise(AlexNet - style PCA - based noise)""" def __init__(self, alphastd, eigval, eigvec): self.alphastd = alphastd self.eigval = torch.Tensor(eigval) self.eigvec = torch.Tensor(eigvec) def __call__(self, img): if self.alphastd == 0: return img alpha = img.new().resize_(3).normal_(0, self.alphastd) rgb = ( self.eigvec.type_as(img) .clone() .mul(alpha.view(1, 3).expand(3, 3)) .mul(self.eigval.view(1, 3).expand(3, 3)) .sum(1) .squeeze() ) return img.add(rgb.view(3, 1, 1).expand_as(img)) class CutoutDefault(object): """ Reference : https://github.com/quark0/darts/blob/master/cnn/utils.py """ def __init__(self, length): self.length = length def __call__(self, img): h, w = img.size(1), img.size(2) mask = np.ones((h, w), np.float32) y = np.random.randint(h) x = np.random.randint(w) y1 = np.clip(y - self.length // 2, 0, h) y2 = np.clip(y + self.length // 2, 0, h) x1 = np.clip(x - self.length // 2, 0, w) x2 = np.clip(x + self.length // 2, 0, w) mask[y1:y2, x1:x2] = 0.0 mask = torch.from_numpy(mask) mask = mask.expand_as(img) img *= mask return img class RandAugment: def __init__(self, n, m): self.n = n self.m = m # [0, 30] self.augment_list = augment_list() def __call__(self, img): ops = random.choices(self.augment_list, k=self.n) for op, minval, maxval in ops: val = (float(self.m) / 30) * float(maxval - minval) + minval img = op(img, val) return img
6,990
24.892593
134
py
Mr.Right
Mr.Right-main/models/ViLT/vilt/transforms/pixelbert.py
from .utils import ( inception_normalize, MinMaxResize, ) from torchvision import transforms from .randaug import RandAugment def pixelbert_transform(size=800): longer = int((1333 / 800) * size) return transforms.Compose( [ MinMaxResize(shorter=size, longer=longer), transforms.ToTensor(), inception_normalize, ] ) def pixelbert_transform_randaug(size=800): longer = int((1333 / 800) * size) trs = transforms.Compose( [ MinMaxResize(shorter=size, longer=longer), transforms.ToTensor(), inception_normalize, ] ) trs.transforms.insert(0, RandAugment(2, 9)) return trs
714
22.064516
54
py
Mr.Right
Mr.Right-main/models/ViLT/vilt/transforms/__init__.py
from .pixelbert import ( pixelbert_transform, pixelbert_transform_randaug, ) _transforms = { "pixelbert": pixelbert_transform, "pixelbert_randaug": pixelbert_transform_randaug, } def keys_to_transforms(keys: list, size=224): return [_transforms[key](size=size) for key in keys]
301
20.571429
56
py
Mr.Right
Mr.Right-main/data/utils.py
import re import cv2 import numpy as np # ref: https://github.com/salesforce/ALBEF def pre_caption(caption,max_words): caption = re.sub( r"([,.'!?\"()*#:;~])", '', caption.lower(), ).replace('-', ' ').replace('/', ' ').replace('<person>', 'person') caption = re.sub( r"\s{2,}", ' ', caption, ) caption = caption.rstrip('\n') caption = caption.strip(' ') #truncate caption caption_words = caption.split(' ') if len(caption_words)>max_words: caption = ' '.join(caption_words[:max_words]) return caption ## aug functions def identity_func(img): return img def autocontrast_func(img, cutoff=0): ''' same output as PIL.ImageOps.autocontrast ''' n_bins = 256 def tune_channel(ch): n = ch.size cut = cutoff * n // 100 if cut == 0: high, low = ch.max(), ch.min() else: hist = cv2.calcHist([ch], [0], None, [n_bins], [0, n_bins]) low = np.argwhere(np.cumsum(hist) > cut) low = 0 if low.shape[0] == 0 else low[0] high = np.argwhere(np.cumsum(hist[::-1]) > cut) high = n_bins - 1 if high.shape[0] == 0 else n_bins - 1 - high[0] if high <= low: table = np.arange(n_bins) else: scale = (n_bins - 1) / (high - low) offset = -low * scale table = np.arange(n_bins) * scale + offset table[table < 0] = 0 table[table > n_bins - 1] = n_bins - 1 table = table.clip(0, 255).astype(np.uint8) return table[ch] channels = [tune_channel(ch) for ch in cv2.split(img)] out = cv2.merge(channels) return out def equalize_func(img): ''' same output as PIL.ImageOps.equalize PIL's implementation is different from cv2.equalize ''' n_bins = 256 def tune_channel(ch): hist = cv2.calcHist([ch], [0], None, [n_bins], [0, n_bins]) non_zero_hist = hist[hist != 0].reshape(-1) step = np.sum(non_zero_hist[:-1]) // (n_bins - 1) if step == 0: return ch n = np.empty_like(hist) n[0] = step // 2 n[1:] = hist[:-1] table = (np.cumsum(n) // step).clip(0, 255).astype(np.uint8) return table[ch] channels = [tune_channel(ch) for ch in cv2.split(img)] out = cv2.merge(channels) return out def rotate_func(img, degree, fill=(0, 0, 0)): ''' like PIL, rotate by degree, not radians ''' H, W = img.shape[0], img.shape[1] center = W / 2, H / 2 M = cv2.getRotationMatrix2D(center, degree, 1) out = cv2.warpAffine(img, M, (W, H), borderValue=fill) return out def solarize_func(img, thresh=128): ''' same output as PIL.ImageOps.posterize ''' table = np.array([el if el < thresh else 255 - el for el in range(256)]) table = table.clip(0, 255).astype(np.uint8) out = table[img] return out def color_func(img, factor): ''' same output as PIL.ImageEnhance.Color ''' ## implementation according to PIL definition, quite slow # degenerate = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)[:, :, np.newaxis] # out = blend(degenerate, img, factor) # M = ( # np.eye(3) * factor # + np.float32([0.114, 0.587, 0.299]).reshape(3, 1) * (1. - factor) # )[np.newaxis, np.newaxis, :] M = ( np.float32([ [0.886, -0.114, -0.114], [-0.587, 0.413, -0.587], [-0.299, -0.299, 0.701]]) * factor + np.float32([[0.114], [0.587], [0.299]]) ) out = np.matmul(img, M).clip(0, 255).astype(np.uint8) return out def contrast_func(img, factor): """ same output as PIL.ImageEnhance.Contrast """ mean = np.sum(np.mean(img, axis=(0, 1)) * np.array([0.114, 0.587, 0.299])) table = np.array([( el - mean) * factor + mean for el in range(256) ]).clip(0, 255).astype(np.uint8) out = table[img] return out def brightness_func(img, factor): ''' same output as PIL.ImageEnhance.Contrast ''' table = (np.arange(256, dtype=np.float32) * factor).clip(0, 255).astype(np.uint8) out = table[img] return out def sharpness_func(img, factor): ''' The differences the this result and PIL are all on the 4 boundaries, the center areas are same ''' kernel = np.ones((3, 3), dtype=np.float32) kernel[1][1] = 5 kernel /= 13 degenerate = cv2.filter2D(img, -1, kernel) if factor == 0.0: out = degenerate elif factor == 1.0: out = img else: out = img.astype(np.float32) degenerate = degenerate.astype(np.float32)[1:-1, 1:-1, :] out[1:-1, 1:-1, :] = degenerate + factor * (out[1:-1, 1:-1, :] - degenerate) out = out.astype(np.uint8) return out def shear_x_func(img, factor, fill=(0, 0, 0)): H, W = img.shape[0], img.shape[1] M = np.float32([[1, factor, 0], [0, 1, 0]]) out = cv2.warpAffine(img, M, (W, H), borderValue=fill, flags=cv2.INTER_LINEAR).astype(np.uint8) return out def translate_x_func(img, offset, fill=(0, 0, 0)): ''' same output as PIL.Image.transform ''' H, W = img.shape[0], img.shape[1] M = np.float32([[1, 0, -offset], [0, 1, 0]]) out = cv2.warpAffine(img, M, (W, H), borderValue=fill, flags=cv2.INTER_LINEAR).astype(np.uint8) return out def translate_y_func(img, offset, fill=(0, 0, 0)): ''' same output as PIL.Image.transform ''' H, W = img.shape[0], img.shape[1] M = np.float32([[1, 0, 0], [0, 1, -offset]]) out = cv2.warpAffine(img, M, (W, H), borderValue=fill, flags=cv2.INTER_LINEAR).astype(np.uint8) return out def posterize_func(img, bits): ''' same output as PIL.ImageOps.posterize ''' out = np.bitwise_and(img, np.uint8(255 << (8 - bits))) return out def shear_y_func(img, factor, fill=(0, 0, 0)): H, W = img.shape[0], img.shape[1] M = np.float32([[1, 0, 0], [factor, 1, 0]]) out = cv2.warpAffine(img, M, (W, H), borderValue=fill, flags=cv2.INTER_LINEAR).astype(np.uint8) return out def cutout_func(img, pad_size, replace=(0, 0, 0)): replace = np.array(replace, dtype=np.uint8) H, W = img.shape[0], img.shape[1] rh, rw = np.random.random(2) pad_size = pad_size // 2 ch, cw = int(rh * H), int(rw * W) x1, x2 = max(ch - pad_size, 0), min(ch + pad_size, H) y1, y2 = max(cw - pad_size, 0), min(cw + pad_size, W) out = img.copy() out[x1:x2, y1:y2, :] = replace return out ### level to args def enhance_level_to_args(MAX_LEVEL): def level_to_args(level): return ((level / MAX_LEVEL) * 1.8 + 0.1,) return level_to_args def shear_level_to_args(MAX_LEVEL, replace_value): def level_to_args(level): level = (level / MAX_LEVEL) * 0.3 if np.random.random() > 0.5: level = -level return (level, replace_value) return level_to_args def translate_level_to_args(translate_const, MAX_LEVEL, replace_value): def level_to_args(level): level = (level / MAX_LEVEL) * float(translate_const) if np.random.random() > 0.5: level = -level return (level, replace_value) return level_to_args def cutout_level_to_args(cutout_const, MAX_LEVEL, replace_value): def level_to_args(level): level = int((level / MAX_LEVEL) * cutout_const) return (level, replace_value) return level_to_args def solarize_level_to_args(MAX_LEVEL): def level_to_args(level): level = int((level / MAX_LEVEL) * 256) return (level, ) return level_to_args def none_level_to_args(level): return () def posterize_level_to_args(MAX_LEVEL): def level_to_args(level): level = int((level / MAX_LEVEL) * 4) return (level, ) return level_to_args def rotate_level_to_args(MAX_LEVEL, replace_value): def level_to_args(level): level = (level / MAX_LEVEL) * 30 if np.random.random() < 0.5: level = -level return (level, replace_value) return level_to_args func_dict = { 'Identity': identity_func, 'AutoContrast': autocontrast_func, 'Equalize': equalize_func, 'Rotate': rotate_func, 'Solarize': solarize_func, 'Color': color_func, 'Contrast': contrast_func, 'Brightness': brightness_func, 'Sharpness': sharpness_func, 'ShearX': shear_x_func, 'TranslateX': translate_x_func, 'TranslateY': translate_y_func, 'Posterize': posterize_func, 'ShearY': shear_y_func, } translate_const = 10 MAX_LEVEL = 10 replace_value = (128, 128, 128) arg_dict = { 'Identity': none_level_to_args, 'AutoContrast': none_level_to_args, 'Equalize': none_level_to_args, 'Rotate': rotate_level_to_args(MAX_LEVEL, replace_value), 'Solarize': solarize_level_to_args(MAX_LEVEL), 'Color': enhance_level_to_args(MAX_LEVEL), 'Contrast': enhance_level_to_args(MAX_LEVEL), 'Brightness': enhance_level_to_args(MAX_LEVEL), 'Sharpness': enhance_level_to_args(MAX_LEVEL), 'ShearX': shear_level_to_args(MAX_LEVEL, replace_value), 'TranslateX': translate_level_to_args( translate_const, MAX_LEVEL, replace_value ), 'TranslateY': translate_level_to_args( translate_const, MAX_LEVEL, replace_value ), 'Posterize': posterize_level_to_args(MAX_LEVEL), 'ShearY': shear_level_to_args(MAX_LEVEL, replace_value), } class RandomAugment(object): def __init__(self, N=2, M=10, isPIL=False, augs=[]): self.N = N self.M = M self.isPIL = isPIL if augs: self.augs = augs else: self.augs = list(arg_dict.keys()) def get_random_ops(self): sampled_ops = np.random.choice(self.augs, self.N) return [(op, 0.5, self.M) for op in sampled_ops] def __call__(self, img): if self.isPIL: img = np.array(img) ops = self.get_random_ops() for name, prob, level in ops: if np.random.random() > prob: continue args = arg_dict[name](level) img = func_dict[name](img, *args) return img
10,271
27.773109
99
py
Mr.Right
Mr.Right-main/data/extract_multimodal_val.py
import os import json import pdb import random from argparse import ArgumentParser random.seed(42) def main(args): document = json.load(open(args.mul_doc,'r')) val_query = json.load(open(args.mul_val,'r')) document_dict = dict() for doc in document: document_dict[doc['id']] = doc val_document = [] for idx, query in enumerate(val_query): new_doc = document_dict[query['id']].copy() del document_dict[query['id']] query["id"] = idx new_doc["id"] = idx val_document.append(new_doc) remaining_doc = list(document_dict.values()) add_doc = random.sample(remaining_doc,args.val_amount - len(val_document)) idx = len(val_document) for doc in add_doc: doc['id'] = idx idx += 1 val_document += add_doc with open(args.mul_val, "w") as outfile: json.dump(val_query, outfile, indent = 4) with open(args.output, "w") as outfile: json.dump(val_document, outfile, indent = 4) if __name__ == '__main__': parser = ArgumentParser() parser.add_argument('--mul_doc', default='multimodal_documents.json') parser.add_argument('--mul_val', default='multimodal_val_queries.json') parser.add_argument('--val_amount', default=10000,type=int) parser.add_argument('--output', default="multimodal_val_documents.json") args = parser.parse_args() print(args) main(args)
1,298
24.470588
75
py
Mr.Right
Mr.Right-main/data/data_module.py
import random import torch import os import json import pickle from torch.utils.data import Dataset, DataLoader from torchvision.transforms import Compose, ToTensor, Normalize, Resize, RandomResizedCrop, RandomHorizontalFlip from pytorch_lightning import LightningDataModule from data.utils import pre_caption, RandomAugment from PIL import Image,ImageFile ImageFile.LOAD_TRUNCATED_IMAGES = True class TextToMultiDataset(Dataset): def __init__(self, args, configs, data, task, tokenizer): self.data = data self.image_size = configs.image_res self.image_root = configs["image_root"] self.task = task self.tokenizer = tokenizer self.configs = configs self.args = args self.q_max_len = configs.query_length self.d_max_len = configs.text_length normalize = Normalize((0.48145466, 0.4578275, 0.40821073), (0.26862954, 0.26130258, 0.27577711)) if task == "train": self.transform = Compose([ RandomResizedCrop(self.image_size, scale=(0.5, 1.), interpolation=Image.BICUBIC), RandomHorizontalFlip(), RandomAugment(2,7,isPIL=True,augs=['Identity','AutoContrast','Equalize','Brightness','Sharpness', 'ShearX', 'ShearY', 'TranslateX', 'TranslateY', 'Rotate']), ToTensor(), normalize, ]) elif task == "val_queries": self.data = self.data elif task == 'test_queries': self.data = self.data elif task == "docs": self.data = self.data self.transform = Compose([ Resize((self.image_size,self.image_size), interpolation=Image.BICUBIC), ToTensor(), normalize, ]) def __len__(self): return len(self.data) def __getitem__(self, index): result = {} data = self.data[index] if self.task == "train": # query str result['query_text_str'] = data["txt_query_str"] result['query_image_str'] = data["img_query_str"] r = random.random() if r < 0.33: query_str = data["txt_query_str"] elif 0.33 < r < 0.66: query_str = data["img_query_str"] elif 0.66< r < 0.83: query_str = data["txt_query_str"] + data["img_query_str"] else: query_str = data["img_query_str"] + data["txt_query_str"] image = data["doc_image"] image = image.replace("./","") image_path = os.path.join(self.image_root,image) image = Image.open(image_path).convert('RGB') image = self.transform(image) result['query_str'] = query_str result['doc_id'] = data["doc_id"] result['doc_str'] = data["doc_text_str"] result['doc_image'] = image result['image_path'] = image_path elif self.task == 'val_queries': result['img_query_str'] = data['img_query_str'] result['txt_query_str'] = data['txt_query_str'] result['multi_query_str'] = data['multi_query_str'] result['doc_id'] = data["doc_id"] elif self.task == 'test_queries': result['img_query_str'] = data['img_query_str'] result['txt_query_str'] = data['txt_query_str'] result['multi_query_str'] = data['multi_query_str'] result['doc_id'] = data["doc_id"] elif self.task == 'docs': image = data["doc_image"] image = image.replace("./","") image_path = os.path.join(self.image_root,image) image = Image.open(image_path).convert('RGB') image = self.transform(image) result['doc_str'] = data["doc_text_str"] result['doc_image'] = image result['image_path'] = image_path result['doc_id'] = data["doc_id"] return result def collate_fn(self, batch): if self.task == 'test_queries' or self.task == 'val_queries': batch_dict = { "doc_id": torch.tensor([b["doc_id"] for b in batch]).long(), "img_query_str": [b["img_query_str"] for b in batch], "txt_query_str": [b["txt_query_str"] for b in batch], "multi_query_str": [b["multi_query_str"] for b in batch], } img_query_str_tensor = self.tokenizer( text=[b["img_query_str"] for b in batch], max_length=self.q_max_len, padding="max_length", truncation=True, return_tensors="pt") txt_query_str_tensor = self.tokenizer( text=[b["txt_query_str"] for b in batch], max_length=self.q_max_len, padding="max_length", truncation=True, return_tensors="pt") multi_query_str_tensor = self.tokenizer( text=[b["multi_query_str"] for b in batch], max_length=self.q_max_len, padding="max_length", truncation=True, return_tensors="pt") batch_dict["img_query_str_tensor"] = img_query_str_tensor batch_dict["txt_query_str_tensor"] = txt_query_str_tensor batch_dict["multi_query_str_tensor"] = multi_query_str_tensor return batch_dict if self.task == 'docs': batch_dict = { "doc_id": torch.tensor([b["doc_id"] for b in batch]).long(), "image_path": [b["image_path"] for b in batch], "doc_str" : [b["doc_str"] for b in batch], "doc_image_tensor": torch.stack([b["doc_image"] for b in batch]), } doc_str_tensor = self.tokenizer( text=[b["doc_str"] for b in batch], max_length=self.d_max_len, padding="max_length", truncation=True, return_tensors="pt") batch_dict["doc_str_tensor"] = doc_str_tensor return batch_dict batch_dict = { "doc_id": torch.tensor([b["doc_id"] for b in batch]).long(), "query_str": [b["query_str"] for b in batch], "doc_str" : [b["doc_str"] for b in batch], "doc_image_tensor": torch.stack([b["doc_image"] for b in batch]), } query_str_tensor = self.tokenizer( text=[b["query_str"] for b in batch], max_length=self.q_max_len, padding="longest" if self.task == "train" else "max_length", # padding="max_length", truncation=True, return_tensors="pt") batch_dict["query_str_tensor"] = query_str_tensor doc_str_tensor = self.tokenizer( text=[b["doc_str"] for b in batch], max_length=self.d_max_len, padding="longest" if self.task == "train" else "max_length", # padding="max_length", truncation=True, return_tensors="pt") batch_dict["doc_str_tensor"] = doc_str_tensor query_str_tensor_total = query_str_tensor if self.task == "train": query_str_tensor_total = self.tokenizer( text=[b["query_text_str"] + ' ' + b["query_image_str"] for b in batch], max_length=self.q_max_len, padding="longest", # padding="max_length", truncation=True, return_tensors="pt") if self.args.ctx_prediction: context_labels = torch.zeros(len(batch), self.tokenizer.vocab_size) pad_id = self.tokenizer.pad_token_id sep_id = self.tokenizer.sep_token_id cls_id = self.tokenizer.cls_token_id context_labels[torch.arange(len(batch)).unsqueeze(1), query_str_tensor_total['input_ids']] = 1 context_labels[torch.arange(len(batch)).unsqueeze(1), doc_str_tensor['input_ids']] = 1 context_labels[:, pad_id] = 0 context_labels[:, sep_id] = 0 context_labels[:, cls_id] = 0 batch_dict["context_labels"] = context_labels return batch_dict class TextToMultiDataModule(LightningDataModule): def __init__(self,args,configs,tokenizer): super().__init__() self.args = args self.shuffle = args.shuffle self.batch_size = args.batch_size self.num_workers = args.num_workers self.configs = configs self.max_len = configs.text_length self.tokenizer = tokenizer def prepare_data(self,train=None,val=None,test=None,document=None): # called only on 1 GPU def load_data(config, mode): datas = self.prepare_text2multi_data(config,mode) return datas if train is not None: self.train_datas = load_data(train, 'train') if val is not None: self.val_datas = load_data(val, 'val') if test is not None: self.test_datas = load_data(test, 'test') if document is not None: self.documents = load_data(document,'doc') def setup(self): if hasattr(self, 'train_datas'): self.train_dataset = TextToMultiDataset(self.args, self.configs, self.train_datas,"train", self.tokenizer) if hasattr(self, 'val_datas'): self.val_queries_dataset = TextToMultiDataset(self.args, self.configs, self.val_datas, "val_queries", self.tokenizer) self.val_docs_dataset = TextToMultiDataset(self.args, self.configs, self.documents, "docs", self.tokenizer) if hasattr(self, 'test_datas'): self.test_queries_dataset = TextToMultiDataset(self.args, self.configs, self.test_datas, "test_queries", self.tokenizer) self.test_docs_dataset = TextToMultiDataset(self.args, self.configs, self.documents, "docs", self.tokenizer) def train_dataloader(self): return DataLoader(self.train_dataset, batch_size=self.batch_size, num_workers=self.num_workers,pin_memory=True,drop_last=True, collate_fn=self.train_dataset.collate_fn) def val_dataloader(self): return [ DataLoader(self.val_queries_dataset, batch_size=self.batch_size, num_workers=self.num_workers, pin_memory=True, collate_fn=self.val_queries_dataset.collate_fn), DataLoader(self.val_docs_dataset, batch_size=self.batch_size, num_workers=self.num_workers, pin_memory=True, collate_fn=self.val_docs_dataset.collate_fn), ] def test_dataloader(self): return [ DataLoader(self.test_queries_dataset, batch_size=self.batch_size, num_workers=self.num_workers, pin_memory=True, collate_fn=self.test_queries_dataset.collate_fn), DataLoader(self.test_docs_dataset, batch_size=self.batch_size, num_workers=self.num_workers, pin_memory=True, collate_fn=self.test_docs_dataset.collate_fn), ] def prepare_text2multi_data(self,files,task="train"): print("\nReading json files") data = [] for f in files: print(f"File: {f}",end="\r") data += json.load(open(f,'r')) result = [] for idx,pairs in enumerate(data): if task == "train": result.append({ "img_query_str": pre_caption(pairs["query_img"],self.max_len), "txt_query_str": pre_caption(pairs["query_text"],self.max_len), "doc_text_str": pre_caption(pairs["doc_text"], self.max_len), "doc_image": pairs["doc_image"], "doc_id": idx, }) if task == "val": result.append({ "multi_query_str": pre_caption(pairs["query_multi"],self.max_len), "img_query_str": pre_caption(pairs["query_img"],self.max_len), "txt_query_str": pre_caption(pairs["query_text"],self.max_len), "doc_id": pairs["id"], }) if task == "test": result.append({ "img_query_str": pre_caption(pairs["query_img"],self.max_len), "txt_query_str": pre_caption(pairs["query_text"],self.max_len), "multi_query_str": pre_caption(pairs["query_multi"],self.max_len), "doc_id": pairs["id"], }) if task == "doc": result.append({ "doc_text_str": pre_caption(pairs["doc_text"], self.max_len), "doc_image": pairs["doc_image"], "doc_id": pairs["id"], }) return result
13,856
44.136808
176
py
NORPPA
NORPPA-main/sql.py
import sqlite3 import numpy as np from datetime import datetime def create_connection(path="/app/mount/tasks.db"): """ create a database connection to the SQLite database """ conn = None try: conn = sqlite3.connect(path, check_same_thread=False) except sqlite3.Error as e: print(e) return conn def select_class_path_by_ids(conn, ids): c = conn.cursor() bindings = ",".join(["?"] * len(ids)) c.execute(f"SELECT image_path, seal_id FROM database WHERE image_id IN ({bindings})", [int(x) for x in ids]) return c.fetchall() def create_database_table(conn): c = conn.cursor() c.execute("PRAGMA foreign_keys=ON") c.execute(''' CREATE TABLE IF NOT EXISTS database ([image_id] INTEGER PRIMARY KEY AUTOINCREMENT, [seal_id] TEXT, [image_path] TEXT, [seal_type] TEXT,[encoding] TEXT, [date] TEXT) ''') c.close() conn.commit() def get_patch_features(conn, image_id): c = conn.cursor() c.execute("SELECT encoding FROM patches WHERE image_id = ?", (int(image_id),)) result = c.fetchall() db_features = np.array([np.fromstring(res[0], dtype=np.float32, sep=' ') for res in result]) c.close() return db_features def get_patch_features_multiple_ids(conn, ids): c = conn.cursor() bindings = ",".join(["?"] * len(ids)) c.execute(f"SELECT image_id, encoding FROM patches WHERE image_id IN ({bindings})", [int(x) for x in ids]) result = c.fetchall() db_features = np.array([np.fromstring(res[1], dtype=np.float32, sep=' ') for res in result]) db_ids = np.array([int(res[0]) for res in result]) c.close() return db_ids, db_features def get_fisher_vectors(conn, ids): c = conn.cursor() bindings = ",".join(["?"] * len(ids)) c.execute(f"SELECT encoding FROM database WHERE image_id IN ({bindings})", [int(x) for x in ids]) result = c.fetchall() db_features = np.array([np.fromstring(res[0], dtype=float, sep=' ') for res in result]) return db_features def get_label(conn, image_id): c = conn.cursor() c.execute("SELECT seal_id FROM database WHERE image_id = ?", (image_id,)) result = c.fetchone() return result def get_db_ids(conn, seal_type="norppa"): c = conn.cursor() c.execute("SELECT image_id FROM database WHERE seal_type = ? ORDER BY image_id", (seal_type,)) result = c.fetchall() result = [res[0] for res in result] return result def get_img_paths_by_id(conn, seal_id): c = conn.cursor() c.execute("SELECT image_id, image_path FROM database WHERE seal_id = ?", (seal_id,)) result = c.fetchall() image_ids = [res[0] for res in result] image_paths = [res[1] for res in result] return image_ids, image_paths def get_patches(conn, image_id): c = conn.cursor() c.execute("SELECT coordinates, encoding FROM patches WHERE image_id = ?", (int(image_id),)) result = c.fetchall() coordinates = np.array([np.fromstring(res[0], dtype=np.float32, sep=' ') for res in result]) db_features = np.array([np.fromstring(res[1], dtype=np.float32, sep=' ') for res in result]) return db_features, coordinates def get_patches_multiple(conn, ids): c = conn.cursor() bindings = ",".join(["?"] * len(ids)) c.execute(f"SELECT coordinates, encoding FROM patches WHERE image_id IN ({bindings})", [int(x) for x in ids]) result = c.fetchall() coordinates = np.array([np.fromstring(res[0], dtype=np.float32, sep=' ') for res in result]) db_features = np.array([np.fromstring(res[1], dtype=np.float32, sep=' ') for res in result]) return db_features, coordinates def get_patch_features(conn, image_id): c = conn.cursor() c.execute("SELECT encoding FROM patches WHERE image_id = ?", (int(image_id),)) result = c.fetchall() db_features = np.array([np.fromstring(res[0], dtype=np.float32, sep=' ') for res in result]) return db_features def get_patch_features_multiple_ids_with_labels(conn, ids): c = conn.cursor() bindings = ",".join(["?"] * len(ids)) c.execute(f"SELECT patches.image_id, patches.encoding, database.seal_id FROM patches INNER JOIN database ON patches.image_id = database.image_id WHERE patches.image_id IN ({bindings})", [int(x) for x in ids]) result = c.fetchall() db_features = np.array([np.fromstring(res[1], dtype=np.float64, sep=' ') for res in result]) db_ids = np.array([int(res[0]) for res in result]) db_labels = np.array([res[2] for res in result]) return db_ids, db_features, db_labels def create_database_table(conn): c = conn.cursor() c.execute("PRAGMA foreign_keys=ON") c.execute(''' CREATE TABLE IF NOT EXISTS database ([image_id] INTEGER PRIMARY KEY AUTOINCREMENT, [seal_id] TEXT, [image_path] TEXT, [seal_type] TEXT,[encoding] TEXT, [date] TEXT) ''') c.close() conn.commit() def create_patches_table(conn): c = conn.cursor() c.execute(''' CREATE TABLE IF NOT EXISTS patches ([patch_id] INTEGER PRIMARY KEY AUTOINCREMENT, [image_id] INTEGER, [coordinates] TEXT, [encoding] TEXT, FOREIGN KEY (image_id) REFERENCES database(image_id) ON DELETE CASCADE) ''') c.close() conn.commit() def insert_patches(conn, image_id, coordinates, encoding): c = conn.cursor() c.execute("INSERT INTO patches (image_id, coordinates, encoding) VALUES (?, ?, ?)", (image_id, np.array2string(coordinates)[1:-1], np.array2string(encoding)[1:-1])) conn.commit() c.close() def update_encoding(conn, image_id, encoding, date): c = conn.cursor() sql = """ UPDATE database SET encoding = ?, date = ? WHERE image_id = ? """ conn.execute(sql, (np.array2string(encoding)[1:-1], date, image_id)) conn.commit() c.close() def insert_database(conn, image_path, seal_id, seal_type, encoding, date): c = conn.cursor() c.execute("INSERT INTO database (image_path, seal_id, seal_type, encoding, date) VALUES (?, ?, ?, ?, ?)", (image_path, seal_id, seal_type, np.array2string(encoding)[1:-1], date)) row_id = c.lastrowid conn.commit() c.close() return row_id
6,354
29.552885
212
py
NORPPA
NORPPA-main/tools.py
from datetime import datetime import random import string import os import sys import shutil from zipfile import ZipFile from skimage import color import numpy as np from datetime import datetime from six.moves import urllib from pattern_extraction.utils import thickness_resize from pattern_extraction.extract_pattern import smart_resize from PIL import Image, ImageFile ImageFile.LOAD_TRUNCATED_IMAGES = True from PIL import Image import json import gc from tqdm import tqdm import pickle def save_pickle(x, file): with open(file, 'wb') as f_file: pickle.dump(x, f_file, protocol=4) return x def load_pickle(file): with open(file, 'rb') as f_file: result = pickle.load(f_file) return result def print_step(text): def identity_print(x): print(text) return x return identity_print def flatten(t): return [item for sublist in t for item in sublist] def apply_step(input, step, rest): result = step(input) if rest: result = flatten((apply_step(x, rest[0], rest[1:]) for x in result)) return result def apply_pipeline(image, pipeline, verbose=False): return apply_pipeline_dataset([image], pipeline, verbose=verbose) def apply_pipeline_dataset(dataset, pipeline, verbose=False): result = dataset for (i,step) in enumerate(pipeline): result = step(result) if verbose: print(f"Completed {i+1}/{len(pipeline)} steps") return result def process(SOURCE_DIR, pipeline): result = [] num_files = sum([len(files) for r, d, files in os.walk(SOURCE_DIR)]) for root, _, files in os.walk(SOURCE_DIR): for f in files: num_files -= 1 result.extend(apply_pipeline(os.path.join(root, f), SOURCE_DIR, pipeline)) gc.collect(0) return result def curry(func, *params, **kw): return lambda x: func(x, *params, **kw) def cat(array): return [y for x in array for y in x] def apply_sequential(func): return lambda dataset: cat([func(x) for x in tqdm(dataset)]) def curry_sequential(func, *params, **kw): return apply_sequential(curry(func, *params, **kw)) def compose(*funcs): return curry(apply_step, step=funcs[0], rest=funcs[1:]) def compose_sequential(*funcs): return apply_sequential(compose(*funcs)) def save_id_result(result, source_dir, dest_path): res_json_path = os.path.join(dest_path, 'result.json') data = [] for _, res in result: res["query"] = os.path.join(source_dir, res["query"]) data.append(res) with open(res_json_path, mode='w') as outfile: json.dump(data, outfile) def save_upload_result(result, dest_path): res_json_path = os.path.join(dest_path, 'result.json') data = [] for image, label in result: if image is None: res = "Fail" else: res = "Success" data.append((label,res)) with open(res_json_path, mode='w') as outfile: json.dump(data, outfile) def randomString(stringLength=8): letters = string.ascii_lowercase return ''.join(random.choice(letters) for i in range(stringLength)) def generate_task_id(task_type): now = datetime.now() date_time = now.strftime("%d%m%y_%H%M%S") return f"{task_type[:4]}_{date_time}" def clean_dir(folder): for filename in os.listdir(folder): file_path = os.path.join(folder, filename) try: if os.path.isfile(file_path) or os.path.islink(file_path): os.unlink(file_path) elif os.path.isdir(file_path): shutil.rmtree(file_path) except Exception as e: print('Failed to delete %s. Reason: %s' % (file_path, e)) def to_zip_file(file, zip_file): if os.path.isfile(file): zip_file.write(file) else: addFolderToZip(zip_file, file, file) def toZip(file, filename): zip_file = ZipFile(filename, 'w') to_zip_file(file, zip_file) zip_file.close() def addFolderToZip(zip_file, folder, base_path): for file in os.listdir(folder): full_path = os.path.join(folder, file) if os.path.isfile(full_path): # print('File added: ' + str(full_path)) zip_file.write(full_path, os.path.relpath(full_path, start=base_path)) # print(full_path, base_path) elif os.path.isdir(full_path): # print('Entering folder: ' + str(full_path)) addFolderToZip(zip_file, full_path, base_path) def is_image(filename): return filename.lower().endswith(('.png', '.jpg', '.jpeg', 'cr2', 'pef', 'nef')) def is_raw_image(filename): return filename.lower().endswith(('cr2', 'pef', 'nef')) def crop_to_bb(img, return_bb=False): labels = color.rgb2gray(np.asarray(img)) labels = (labels > 0).astype(np.uint8) where = np.where(labels) y1, x1 = np.amin(where, axis=1) y2, x2 = np.amax(where, axis=1) img_cropped = img.crop((x1, y1, x2, y2)) if return_bb: return (img_cropped, (x1,y1,x2-x1,y2-y1)) else: return img_cropped def crop_imgs_in_dir(src): for dirpath, _, filenames in os.walk(src): for filename in filenames: source = os.path.join(dirpath, filename) img = Image.open(source) img = img.convert('RGB') img_cropped = crop_to_bb(img) img_cropped.save(source) def crop_step(input): image, label = input if image is not None: image, bb = crop_to_bb(image, True) if type(label) is dict: label["bb"] = bb return [(image, label)] def crop_step_sequential(input): return apply_sequential(crop_step)(input) def thickness_resize_step(input): image, label = input if image is not None: image, ratio = thickness_resize(image, return_ratio=True) if type(label) is dict: label["resize_ratio"] = ratio return [(image, label)] def change_dir(path, new_dir): name = os.path.basename(path) return os.path.join(new_dir, name) def get_save_step(dest_dir): return lambda x: save_step(x, dest_dir) def get_save_step(dest_dir, new_path_name=None): return curry_sequential(save_step, dest_dir, new_path_name=new_path_name) def test_save_step(dest_dir): return lambda x: test_step(x, dest_dir) def test_step(input, dest_dir): image, label = input if image is not None: if type(label) is dict and 'dataset_dir' in label: new_path = os.path.join(dest_dir, os.path.relpath(label['file'], label['dataset_dir'])) else: new_path = change_dir(label['file'], dest_dir) if os.path.exists(new_path): return [] return [(image, label)] def save_step(input, dest_dir, new_path_name=None): image, label = input if image is not None: os.makedirs(dest_dir, exist_ok=True) if type(label) is dict and 'dataset_dir' in label: new_path = os.path.join(dest_dir, os.path.relpath(label['file'], label['dataset_dir'])) else: new_path = change_dir(label['file'], dest_dir) os.makedirs(os.path.dirname(new_path), exist_ok = True) image.save(new_path, format="png") if type(label) is dict and new_path_name is not None: label[new_path_name] = new_path return [(image, label)] def download_url(url, dst): print('* url="{}"'.format(url)) print('* destination="{}"'.format(dst)) def _reporthook(count, block_size, total_size): global start_time if count == 0: start_time = time.time() return duration = time.time() - start_time progress_size = int(count * block_size) speed = int(progress_size / (1024 * duration)) percent = int(count * block_size * 100 / total_size) sys.stdout.write( '\r...%d%%, %d MB, %d KB/s, %d seconds passed' % (percent, progress_size / (1024 * 1024), speed, duration) ) sys.stdout.flush() urllib.request.urlretrieve(url, dst, _reporthook) sys.stdout.write('\n') def read_image(path): if not os.path.exists(path): raise IOError(f'"{path}" does not exist') try: img = Image.open(path).convert('RGB') except IOError: print(f'IOError when reading "{path}"') return img def get_topk_matrix(identification_result): result = [] for (db_labels, q_labels) in (identification_result): q_class = q_labels['class_id'] q_ln = len(q_labels['labels']) result.extend([[db_label['db_label']['class_id']==q_class for db_label in db_labels]]*q_ln) topk = len(identification_result[0][0]) return (np.asarray(result), topk) def calculate_accuracy(result, max_topk=None): topk_matrix, topk = get_topk_matrix(result) topk = max_topk if max_topk is not None else topk print(f'TOP-k={topk}') hits = topk_matrix # hits = (db_labels.T == q_labels).T print([sum((np.sum(hits[:, :j+1], axis=1) > 0)) / len(topk_matrix) for j in range(topk)]) def get_topk_matrix(identification_result): result = [] for (db_labels, q_labels) in identification_result: q_class = q_labels['class_id'] q_ln = len(q_labels['labels']) result.append([db_label['db_label']['class_id']==q_class for db_label in db_labels]*q_ln) return np.asarray(result) def get_topk_accuracy(identification_result): result = [] for (db_labels, q_labels) in identification_result: q_class = q_labels['class_id'] q_ln = len(q_labels['labels']) hits = [db_label['db_label']['class_id']==q_class for db_label in db_labels] for _ in range(q_ln): result.append(hits) result = np.asarray(result) return [sum((np.sum(result[:, :j+1], axis=1) > 0)) / len(result) for j in range(result.shape[1])] def print_topk_accuracy(identification_result, label=""): topk_acc = get_topk_accuracy(identification_result) print(label) for (i, acc) in enumerate(topk_acc): print(f"Top-{i+1} accuracy: {acc*100}%") return identification_result def resize_dataset(input, size): image, img_label = input if image is None: return [input] result, ratio = smart_resize(image, size, return_ratio=True) img_label["resize_ratio"] = ratio return [(result, img_label)] def update_codebooks(input, cfg): codebooks, encoded = input cfg["codebooks"] = codebooks return encoded
10,545
29.391931
101
py
NORPPA
NORPPA-main/config.py
import os import sys from pathlib import Path import cv2 import numpy as np file_folder = Path(__file__).resolve().parent sys.path.append(str(file_folder / "reidentification/hesaff_pytorch")) from HessianAffinePatches import init_affnet, init_orinet, init_hardnet from segmentation.detectron_segment import create_predictor from pattern_extraction.extract_pattern import create_unet from torchvision.datasets.utils import download_url from sql import create_connection def init_file(path, url, allow_download=True): if Path(path).exists(): return path elif allow_download: download_url(url, Path(path).parent, Path(path).name) return path else: raise Exception("The file {path} is not found!") def config(use_cuda=True, allow_download=True): config = {} base_dir = Path(__file__).resolve().parent mount_path = "/ekaterina/work/data/" path_db = mount_path + "DB.db" config["conn"] = create_connection(path_db) config["detectron_predictor"] = create_predictor(init_file(base_dir/"models/R-101-FPN_150ims.pth", "https://github.com/kwadraterry/NORPPA/raw/models/models/R-101-FPN_150ims.pth", allow_download=allow_download), not use_cuda ) config["unet"] = create_unet(init_file(base_dir/"models/unet_seals_512.hdf5", "https://github.com/kwadraterry/NORPPA/raw/models/models/unet_seals_512.hdf5", allow_download=allow_download)) config["net"] = init_hardnet(init_file(base_dir/"models/HardNet++.pth", "https://github.com/kwadraterry/NORPPA/raw/models/models/HardNet++.pth", allow_download=allow_download), use_cuda=use_cuda) affnet = init_affnet(init_file(base_dir/"models/AffNet.pth", "https://github.com/kwadraterry/NORPPA/raw/models/models/AffNet.pth", allow_download=allow_download), use_cuda=use_cuda) orinet = init_orinet(init_file(base_dir/"models/OriNet.pth", "https://github.com/kwadraterry/NORPPA/raw/models/models/OriNet.pth", allow_download=allow_download), use_cuda=use_cuda) codebooks_path = init_file(base_dir/'codebooks/codebooks.pickle', "https://github.com/kwadraterry/NORPPA/raw/models/codebooks/codebooks.pickle", allow_download=allow_download) config["codebooks_path"] = Path(base_dir/"codebooks/norppa.pickle") config["codebooks"] = None config["hesaff_args"] = {'init_sigma': 1.3213713243956968, 'mrSize': 9.348280997446642, 'nlevels': 10, 'num_features': 480, 'unsharp_amount': 6.80631647207343, 'unsharp_radius': None, 'use_cuda' :use_cuda} config["hesaff_args"]["AffNet"] = affnet config["hesaff_args"]["OriNet"] = orinet config["hesaff_args"]["patch_size"] = 32 config["use_hesaff"] = True config["pca"] = 64 config["use_kpca"] = False config["n_clusters"] = 1400 config["features_shape"] = 64 config["topk"] = 10 config["kernel"] = "rbf" config["use_cuda"] = use_cuda config["dataset_dir"] = base_dir/'data' config["sequence_dataset_dir"] = '/ekaterina/work/data/many_dataset/original_small' config["batch_size"] = 256 config["geometric"] = { "method": cv2.RANSAC, "max_iters": 5000, "max_reproj_err": 0.2, "estimator": lambda d, mask: d ** np.sum(mask) } return config
4,039
41.526316
123
py