# Copyright 2020-2026 The HuggingFace Team. 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. import asyncio import atexit import copy import importlib.resources as pkg_resources import inspect import math import os import sys import textwrap import time import warnings from collections import defaultdict, deque from collections.abc import Callable from contextlib import nullcontext from pathlib import Path from typing import Any, Protocol import numpy as np import pandas as pd import torch import torch.utils.data import transformers from accelerate.logging import get_logger from accelerate.utils import gather, gather_object, is_peft_model, set_seed from datasets import Dataset, IterableDataset from huggingface_hub import CommitScheduler, DatasetCard, DatasetCardData, create_repo from packaging.version import Version from torch import nn from torch.distributed.fsdp import FullyShardedDataParallel as FSDP from torch.utils.data import Sampler from transformers import ( AutoModelForSequenceClassification, AutoProcessor, AutoTokenizer, GenerationConfig, PreTrainedModel, PreTrainedTokenizerBase, ProcessorMixin, TrainerCallback, is_trackio_available, is_wandb_available, ) from transformers.utils import is_peft_available, is_rich_available from ..chat_template_utils import add_response_schema, get_training_chat_template, parse_response from ..data_utils import apply_chat_template, is_conversational, prepare_multimodal_messages from ..extras.profiling import profiling_context, profiling_decorator from ..generation.vllm_generation import VLLMGeneration from ..import_utils import is_jmespath_available, is_liger_kernel_available from ..models import prepare_deepspeed, prepare_fsdp, unwrap_model_for_generation from ..models.utils import _ForwardRedirection, disable_gradient_checkpointing from .base_trainer import _BaseTrainer from .callbacks import SyncRefModelCallback from .grpo_config import GRPOConfig from .utils import ( RepeatSampler, create_model_from_path, disable_dropout_in_model, entropy_from_logits, get_config_model_id, identity, nanmax, nanmin, nanstd, pad, print_prompt_completions_sample, selective_log_softmax, shuffle_sequence_dict, shutdown_event_loop_in_daemon, split_pixel_values_by_grid, split_tensor_dict, start_event_loop_in_daemon, unsplit_pixel_values_by_grid, use_adapter, ) if is_peft_available(): from peft import PeftConfig, PeftModel, get_peft_model if is_liger_kernel_available(): from liger_kernel.chunked_loss import LigerFusedLinearGRPOLoss if is_wandb_available(): import wandb if is_trackio_available(): import trackio logger = get_logger(__name__) # A reward function can be a string, interpreted as a model ID and loaded as a pretrained model, a pretrained model, or # a callable that returns a list of floats (the rewards). The callable receives prompts, completions, and additional # arguments from the trainer (refer to the trainer's source for details). To ensure forward compatibility, it should # accept **kwargs. RewardFunc = str | PreTrainedModel | Callable[..., list[float | None]] # What we call a rollout function is a callable that takes prompts (list) and the trainer instance as parameters and # returns a dict of generation results. Those results must include "prompt_ids", "completion_ids", and "logprobs" # fields. Any extra fields (per-completion) are forwarded to the reward functions. RolloutFunc = Callable[[list[str], "GRPOTrainer"], dict[str, Any]] class _SupportsReset(Protocol): def reset(self, **kwargs) -> str | None: ... EnvironmentFactory = Callable[[], _SupportsReset] class GRPOTrainer(_BaseTrainer): """ Trainer for the Group Relative Policy Optimization (GRPO) method. This algorithm was initially proposed in the paper [DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models](https://huggingface.co/papers/2402.03300). Example: ```python from trl import GRPOTrainer from trl.rewards import accuracy_reward from datasets import load_dataset dataset = load_dataset("trl-lib/DeepMath-103K", split="train") trainer = GRPOTrainer( model="Qwen/Qwen2.5-0.5B-Instruct", reward_funcs=accuracy_reward, train_dataset=dataset, ) trainer.train() ``` Args: model (`str` or [`~transformers.PreTrainedModel`] or [`~peft.PeftModel`]): Model to be trained. Can be either: - A string, being the *model id* of a pretrained model hosted inside a model repo on huggingface.co, or a path to a *directory* containing model weights saved using [`~transformers.PreTrainedModel.save_pretrained`], e.g., `'./my_model_directory/'`. The model is loaded using `.from_pretrained` (where `` is derived from the model config) with the keyword arguments in `args.model_init_kwargs`. - A [`~transformers.PreTrainedModel`] object. Only causal language models are supported. - A [`~peft.PeftModel`] object. Only causal language models are supported. reward_funcs (`RewardFunc | list[RewardFunc]`): Reward functions to be used for computing the rewards. To compute the rewards, we call all the reward functions with the prompts and completions and sum the rewards. Can be either: - A single reward function, such as: - A string: The *model ID* of a pretrained model hosted inside a model repo on huggingface.co, or a path to a *directory* containing model weights saved using [`~transformers.PreTrainedModel.save_pretrained`], e.g., `'./my_model_directory/'`. The model is loaded using [`~transformers.AutoModelForSequenceClassification.from_pretrained`] with `num_labels=1` and the keyword arguments in `args.model_init_kwargs`. - A [`~transformers.PreTrainedModel`] object: Only sequence classification models are supported. - A custom reward function: The function is provided with the prompts and the generated completions, plus any additional columns in the dataset. It should return a list of rewards. Custom reward functions can be either synchronous or asynchronous and can also return `None` when the reward is not applicable to those samples. This is useful for multi-task training where different reward functions apply to different types of samples. When a reward function returns `None` for a sample, that reward function is excluded from the reward calculation for that sample. For more details, see [Using a custom reward function](#using-a-custom-reward-function). The trainer's state is also passed to the reward function. The trainer's state is an instance of [`~transformers.TrainerState`] and can be accessed by accessing the `trainer_state` argument to the reward function's signature. - A list of reward functions, where each item can independently be any of the above types. Mixing different types within the list (e.g., a string model ID and a custom reward function) is allowed. args ([`GRPOConfig`], *optional*): Configuration for this trainer. If `None`, a default configuration is used. train_dataset ([`~datasets.Dataset`] or [`~datasets.IterableDataset`]): Dataset to use for training. It must include a column `"prompt"`. Any additional columns in the dataset is ignored. The format of the samples can be either: - [Standard](dataset_formats#standard): Each sample contains plain text. - [Conversational](dataset_formats#conversational): Each sample contains structured messages (e.g., role and content). eval_dataset ([`~datasets.Dataset`], [`~datasets.IterableDataset`] or `dict[str, Dataset | IterableDataset]`): Dataset to use for evaluation. It must meet the same requirements as `train_dataset`. processing_class ([`~transformers.PreTrainedTokenizerBase`], [`~transformers.ProcessorMixin`], *optional*): Processing class used to process the data. The padding side must be set to "left". If `None`, the processing class is loaded from the model's name with [`~transformers.AutoProcessor.from_pretrained`]. A padding token, `tokenizer.pad_token`, must be set. If the processing class has not set a padding token, `tokenizer.eos_token` will be used as the default. reward_processing_classes ([`~transformers.PreTrainedTokenizerBase`] or `list[PreTrainedTokenizerBase]`, *optional*): Processing classes corresponding to the reward functions specified in `reward_funcs`. Can be either: - A single processing class: Used when `reward_funcs` contains only one reward function. - A list of processing classes: Must match the order and length of the reward functions in `reward_funcs`. If set to `None`, or if an element of the list corresponding to a [`~transformers.PreTrainedModel`] is `None`, the tokenizer for the model is automatically loaded using [`~transformers.AutoTokenizer.from_pretrained`]. For elements in `reward_funcs` that are custom reward functions (not [`~transformers.PreTrainedModel`]), the corresponding entries in `reward_processing_classes` are ignored. callbacks (list of [`~transformers.TrainerCallback`], *optional*): List of callbacks to customize the training loop. Will add those to the list of default callbacks detailed in [here](https://huggingface.co/docs/transformers/main_classes/callback). If you want to remove one of the default callbacks used, use the [`~transformers.Trainer.remove_callback`] method. optimizers (`tuple[torch.optim.Optimizer | None, torch.optim.lr_scheduler.LambdaLR | None]`, *optional*, defaults to `(None, None)`): A tuple containing the optimizer and the scheduler to use. Will default to an instance of `AdamW` on your model and a scheduler given by [`~transformers.get_linear_schedule_with_warmup`] controlled by `args`. peft_config ([`~peft.PeftConfig`], *optional*): PEFT configuration used to wrap the model. If `None`, the model is not wrapped. tools (list of `Callable`, *optional*): A list of callable tool functions (sync or async) that the model can invoke during generation. Each tool should be a standard Python function with properly type-hinted arguments and return values, and a Google-style docstring describing its purpose, arguments, and return value. For more details, see: https://huggingface.co/docs/transformers/en/chat_extras#passing-tools. The model uses the function's name, type hints, and docstring to determine how to call it. Ensure that the model's chat template supports tool use and that it has been fine-tuned for tool calling. rollout_func (`RolloutFunc`, *optional*): Function to use for generating completions. It receives the list of prompts allocated to the current process and the trainer instance. It must return a dict with `"prompt_ids"`, `"completion_ids"`, and `"logprobs"` fields, and can optionally return `"logprob_token_ids"` (same shape as `"logprobs"`). Any other fields are forwarded to the reward functions. The function receives the raw per-process prompt slice with no duplication; it is responsible for returning the correct number of completions per prompt (see `num_generations` / `num_generations_eval` on the trainer). This feature is experimental and may change or be removed at any time without prior notice. environment_factory (`EnvironmentFactory`, *optional*): A callable that creates and returns an environment instance. The environment class should define methods that can be invoked as tools during generation. Each method should comply with the same requirements as the `tools` described above. If `environment_factory` is provided, an instance of the environment is created for each generation in the batch, allowing for parallel and independent interactions. The environment must also implement a callable `reset` method that can be used to reset state between generations. The `reset` method should return either `None` or a string: when it returns a string, that string is appended to the last user message before generation. This feature is experimental and may change or be removed at any time without prior notice. """ _tag_names = ["trl", "grpo"] _name = "GRPO" _paper = { "title": "DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models", "id": "2402.03300", # docstyle-ignore "citation": textwrap.dedent("""\ @article{shao2024deepseekmath, title = {{DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models}}, author = {Zhihong Shao and Peiyi Wang and Qihao Zhu and Runxin Xu and Junxiao Song and Mingchuan Zhang and Y. K. Li and Y. Wu and Daya Guo}, year = 2024, eprint = {arXiv:2402.03300}, }"""), } def __init__( self, model: "str | PreTrainedModel | PeftModel", reward_funcs: RewardFunc | list[RewardFunc], args: GRPOConfig | None = None, train_dataset: Dataset | IterableDataset | None = None, eval_dataset: Dataset | IterableDataset | dict[str, Dataset | IterableDataset] | None = None, processing_class: PreTrainedTokenizerBase | ProcessorMixin | None = None, reward_processing_classes: PreTrainedTokenizerBase | list[PreTrainedTokenizerBase] | None = None, callbacks: list[TrainerCallback] | None = None, optimizers: tuple[torch.optim.Optimizer | None, torch.optim.lr_scheduler.LambdaLR | None] = (None, None), peft_config: "PeftConfig | None" = None, tools: list[Callable] | None = None, rollout_func: RolloutFunc | None = None, environment_factory: EnvironmentFactory | None = None, ): # Args if args is None: model_name = model if isinstance(model, str) else get_config_model_id(model.config) model_name = model_name.split("/")[-1] args = GRPOConfig(f"{model_name}-GRPO") # Model if isinstance(model, str): model_init_kwargs = args.model_init_kwargs or {} # Distributed training requires device_map=None ("auto" fails) if args.distributed_state.distributed_type in ["MULTI_GPU", "DEEPSPEED"]: model_init_kwargs["device_map"] = None model = create_model_from_path(model, **model_init_kwargs) else: if args.model_init_kwargs is not None: logger.warning( "You passed `model_init_kwargs` to the `GRPOConfig`, but your model is already instantiated. " "The `model_init_kwargs` will be ignored." ) # Some models (SmolVLM/Idefics3) don't support `logits_to_keep` argument and error out if we pass it # Inspect the forward method before we wrap the model with PEFT self.model_kwarg_keys = ( inspect.signature(model.forward).parameters.keys() if not hasattr(model, "get_base_model") else inspect.signature(model.get_base_model().forward).parameters.keys() ) # Processing class if processing_class is None: processing_class = AutoProcessor.from_pretrained( get_config_model_id(model.config), truncation_side="left", padding_side="left" ) # Handle pad token for processors or tokenizers if isinstance(processing_class, ProcessorMixin): tokenizer = processing_class.tokenizer elif isinstance(processing_class, PreTrainedTokenizerBase): tokenizer = processing_class else: raise TypeError("The `processing_class` must be either a `PreTrainedTokenizerBase` or a `ProcessorMixin`") if tokenizer.pad_token is None: tokenizer.pad_token = tokenizer.eos_token self.pad_token = tokenizer.pad_token self.pad_token_id = tokenizer.pad_token_id self.eos_token_id = tokenizer.eos_token_id if is_peft_available() and is_peft_model(model) and peft_config is not None: raise ValueError( "You passed a `PeftModel` instance together with a `peft_config` to the trainer. Please first merge " "and unload the existing adapter, save the resulting base model, and then pass that base model along " "with the new `peft_config` to the trainer." ) if is_peft_available() and is_peft_model(model) and args.beta != 0.0: # If the model is a PEFT model with a pretrained adapter, we need to create a "ref" adapter that is a copy # of the "default" adapter, so that we can use it as the reference model during GRPO training. model.add_adapter("ref", model.peft_config["default"]) for name, param in model.named_parameters(): if ".default." in name: ref_name = name.replace(".default.", ".ref.") ref_param = model.get_parameter(ref_name) ref_param.data.copy_(param.data) # Create PEFT model if peft_config is not None: model = get_peft_model(model, peft_config) # When using gradient checkpointing with PEFT, we need to enable input gradients. transformers.Trainer normally # handles this, but a bug currently prevents it; see https://github.com/huggingface/transformers/issues/42489 if is_peft_available() and is_peft_model(model) and args.gradient_checkpointing: model.enable_input_require_grads() # When using QLoRA, the PEFT adapter weights are converted to bf16 to follow the recommendations from the # original paper (see https://huggingface.co/papers/2305.14314, paragraph 3). Normally, this can be done by # passing `autocast_adapter_dtype=False` to `get_peft_model`, but this option is not yet supported for # quantized models. See: https://github.com/huggingface/peft/issues/2889 # Non-quantized models do not have the `is_loaded_in_{8,4}bit` attributes, whereas quantized models do if getattr(model, "is_loaded_in_4bit", False) or getattr(model, "is_loaded_in_8bit", False): for param in model.parameters(): if param.requires_grad: param.data = param.data.to(torch.bfloat16) # Reward functions if not isinstance(reward_funcs, list): reward_funcs = [reward_funcs] self.reward_func_names = [] for i, reward_func in enumerate(reward_funcs): if isinstance(reward_func, str): model_init_kwargs = args.model_init_kwargs or {} # Distributed training requires device_map=None ("auto" fails) if args.distributed_state.distributed_type in ["MULTI_GPU", "DEEPSPEED"]: model_init_kwargs["device_map"] = None reward_funcs[i] = AutoModelForSequenceClassification.from_pretrained( reward_func, num_labels=1, **model_init_kwargs ) if isinstance(reward_funcs[i], nn.Module): # Use Module over PretrainedModel for compat w/ compiled models self.reward_func_names.append(get_config_model_id(reward_funcs[i].config).split("/")[-1]) else: self.reward_func_names.append(reward_funcs[i].__name__) self.reward_funcs = reward_funcs # Reward weights if args.reward_weights is not None: if len(args.reward_weights) != len(reward_funcs): raise ValueError( f"Number of reward weights ({len(args.reward_weights)}) must match number of reward " f"functions ({len(reward_funcs)})" ) self.reward_weights = torch.tensor(args.reward_weights, dtype=torch.float32) else: self.reward_weights = torch.ones(len(reward_funcs), dtype=torch.float32) # Reward processing class if reward_processing_classes is None: reward_processing_classes = [None] * len(reward_funcs) elif not isinstance(reward_processing_classes, list): reward_processing_classes = [reward_processing_classes] if len(reward_processing_classes) != len(reward_funcs): raise ValueError( f"The number of reward processing classes ({len(reward_processing_classes)}) must match the number of " f"reward functions ({len(reward_funcs)})." ) for i, (reward_processing_class, reward_func) in enumerate( zip(reward_processing_classes, reward_funcs, strict=True) ): if isinstance(reward_func, PreTrainedModel): if reward_processing_class is None: reward_processing_class = AutoTokenizer.from_pretrained(get_config_model_id(reward_func.config)) if reward_processing_class.pad_token_id is None: reward_processing_class.pad_token = reward_processing_class.eos_token # The reward model computes the reward for the latest non-padded token in the input sequence. # So it's important to set the pad token ID to the padding token ID of the processing class. reward_func.config.pad_token_id = reward_processing_class.pad_token_id reward_processing_classes[i] = reward_processing_class self.reward_processing_classes = reward_processing_classes # Rollout function if rollout_func is not None and os.environ.get("TRL_EXPERIMENTAL_SILENCE", "0") != "1": warnings.warn( "You are using 'rollout_func', which is an experimental feature. This API may change or be removed at " "any time without prior notice. Silence this warning by setting environment variable " "TRL_EXPERIMENTAL_SILENCE=1.", UserWarning, stacklevel=2, ) self.rollout_func = rollout_func if environment_factory is not None and os.environ.get("TRL_EXPERIMENTAL_SILENCE", "0") != "1": warnings.warn( "You are using 'environment_factory', which is an experimental feature. This API may change or be " "removed at any time without prior notice. Silence this warning by setting environment variable " "TRL_EXPERIMENTAL_SILENCE=1.", UserWarning, stacklevel=2, ) # Tools if tools: if not Version(transformers.__version__) >= Version("5.0.0"): raise ImportError( "Using tools with GRPOTrainer requires transformers version 5.0.0 or higher. Please upgrade " "transformers with `pip install --upgrade transformers` to use this feature." ) if environment_factory: if not Version(transformers.__version__) >= Version("5.2.0"): raise ImportError( "Using `environment_factory` with GRPOTrainer requires transformers version 5.2.0 or higher. " "Please install transformers from the main branch with `pip install " "git+https://github.com/huggingface/transformers.git@main` to use this feature." ) if tools or environment_factory: if not is_jmespath_available(): raise ImportError( "Using tools with GRPOTrainer requires the jmespath library for response parsing. Please install " "it with `pip install jmespath` to use this feature." ) # Create the environments and extract their methods to be used as tools. We create one environment per rollout generation_batch_size = args.per_device_train_batch_size * args.steps_per_generation if environment_factory is not None: self.environments = [environment_factory() for _ in range(generation_batch_size)] environment_methods = [[] for _ in range(generation_batch_size)] for i, environment in enumerate(self.environments): has_reset = False for name, member in inspect.getmembers(environment, predicate=inspect.ismethod): if name == "reset": has_reset = True elif not name.startswith("_"): environment_methods[i].append(member) if not has_reset: raise ValueError( "Each environment instance returned by `environment_factory` must define a callable `reset` " ) else: self.environments = None tools = tools or [] self._sync_tool_dicts = [{} for _ in range(generation_batch_size)] self._async_tool_dicts = [{} for _ in range(generation_batch_size)] for i in range(generation_batch_size): for tool in tools + (environment_methods[i] if self.environments is not None else []): if inspect.iscoroutinefunction(tool): self._async_tool_dicts[i][tool.__name__] = tool else: self._sync_tool_dicts[i][tool.__name__] = tool self.tools = tools + (environment_methods[0] if self.environments is not None else []) # Check for async functions to start an event loop on a daemon thread self._has_async_funcs = any(inspect.iscoroutinefunction(func) for func in self.reward_funcs + self.tools) if self._has_async_funcs: self.async_loop_thread, self.async_loop, self.async_loop_ready_event = start_event_loop_in_daemon( name="GRPOTrainer-AsyncLoop" ) # wait until the event loop is running in the daemon thread self.async_loop_ready_event.wait() atexit.register(shutdown_event_loop_in_daemon, self.async_loop_thread, self.async_loop) # At the time of initial implementation, most tokenizers do not have built-in support for response schemas. # While waiting for broader adoption, we provide this utility function to manually set the response schema for # known chat templates. # We need `getattr`` until the base class sets a default None value for response_schema if self.tools and not getattr(processing_class, "response_schema", None): processing_class = add_response_schema(processing_class) # In multi-turn training, the chat template *must* be prefix-preserving. If the tokenizer's original template # isn't, we replace it at initialization with a training-safe, prefix-preserving template. if self.tools: self.chat_template = get_training_chat_template(processing_class) else: self.chat_template = None # Training arguments self.max_completion_length = args.max_completion_length # = |o_i| in the GRPO paper self.num_generations = args.num_generations # = G in the GRPO paper self.max_tool_calling_iterations = args.max_tool_calling_iterations or sys.maxsize self.num_generations_eval = args.num_generations_eval or self.num_generations self.chat_template_kwargs = args.chat_template_kwargs or {} self.temperature = args.temperature self.top_p = args.top_p self.top_k = args.top_k self.min_p = args.min_p self.repetition_penalty = args.repetition_penalty self.use_transformers_paged = args.use_transformers_paged self.pad_to_multiple_of = args.pad_to_multiple_of self.use_vllm = args.use_vllm self.vllm_mode = args.vllm_mode self.vllm_gpu_memory_utilization = args.vllm_gpu_memory_utilization # only applies to colocation mode self.vllm_tensor_parallel_size = args.vllm_tensor_parallel_size # only applies to colocation mode self.vllm_importance_sampling_correction = args.vllm_importance_sampling_correction self.vllm_importance_sampling_mode = args.vllm_importance_sampling_mode self.vllm_importance_sampling_cap = args.vllm_importance_sampling_cap self.use_liger_kernel = args.use_liger_kernel self.loss_type = args.loss_type self.multi_objective_aggregation = args.multi_objective_aggregation self.scale_rewards = args.scale_rewards self.importance_sampling_level = args.importance_sampling_level self.off_policy_mask_threshold = args.off_policy_mask_threshold if self.use_liger_kernel and self.off_policy_mask_threshold is not None: raise ValueError("Liger kernel does not support off-policy sequence masking yet.") self.mask_truncated_completions = args.mask_truncated_completions self.top_entropy_quantile = args.top_entropy_quantile if self.use_liger_kernel and self.top_entropy_quantile < 1.0: raise NotImplementedError( "Liger Kernels don't currently support masking token positions based on entropy." ) if self.use_liger_kernel and self.importance_sampling_level not in ("token", "sequence"): raise ValueError( f"Unknown importance sampling level: {self.importance_sampling_level}. " "Possible values are 'token' and 'sequence'." ) # Datasets self.shuffle_dataset = args.shuffle_dataset if train_dataset is None: raise ValueError("`train_dataset` is required") elif ( isinstance(train_dataset, IterableDataset) or isinstance(eval_dataset, IterableDataset) or ( isinstance(eval_dataset, dict) and any(isinstance(ds, IterableDataset) for ds in eval_dataset.values()) ) ): # See https://github.com/huggingface/trl/issues/3213 raise NotImplementedError( "Iterable datasets are not yet supported in GRPOTrainer. Please use a standard dataset instead." ) if args.loss_type == "luspo" and args.importance_sampling_level != "sequence": logger.warning( "When using `'luspo'` loss, `importance_sampling_level` should be set to `'sequence'` to mirror the " "paper's setup." ) if args.loss_type == "vespo" and args.importance_sampling_level != "token": logger.warning( "VESPO computes sequence-level importance weights internally. `importance_sampling_level` should be " "set to `'token'` (the default)." ) if self.loss_type == "vespo" and self.use_vllm and self.vllm_importance_sampling_correction: if self.vllm_importance_sampling_mode not in ["token_truncate", "token_mask"]: raise ValueError( f"VESPO loss requires `vllm_importance_sampling_mode` to be either 'token_truncate' or " f"'token_mask'. Got: {self.vllm_importance_sampling_mode}." ) # Multi-step self.num_iterations = args.num_iterations # = 𝜇 in the GRPO paper self.epsilon_low = args.epsilon self.epsilon_high = args.epsilon_high if args.epsilon_high is not None else args.epsilon # Tracks the number of iterations (forward + backward passes), including those within a grad accum cycle self._step = 0 # Buffer the batch to reuse generated outputs across multiple updates. For more details, see # `_get_train_sampler` and `_prepare_inputs`. self._buffered_inputs = None # Transformers explicitly set use_reentrant=True in the past to silence a PyTorch warning, but the default was # never updated once PyTorch switched to recommending use_reentrant=False. Until that change lands upstream # (see https://github.com/huggingface/transformers/pull/43203) and is released (most likely in 5.0.0), we # default to the recommended non-reentrant behavior here, while preserving any user-provided value. if args.gradient_checkpointing and Version(transformers.__version__) < Version("5.0.0"): args.gradient_checkpointing_kwargs = args.gradient_checkpointing_kwargs or {} args.gradient_checkpointing_kwargs.setdefault("use_reentrant", False) super().__init__( model=model, args=args, data_collator=identity, # No data collation is needed in GRPO train_dataset=train_dataset, eval_dataset=eval_dataset, processing_class=processing_class, callbacks=callbacks, optimizers=optimizers, # In Trainer, `training_step` scales the loss by `gradient_accumulation_steps` only if `compute_loss_func` # is None. For DAPO, loss scaling instead depends on the total number of completions tokens across the # global accumulated batch. To control scaling ourselves, we must disable Trainer’s built-in scaling. The # simplest (though a bit hacky) way is to set `compute_loss_func` to any non-None value, which bypasses # that behavior without rewriting `training_step`. compute_loss_func="non-None value to disable scaling", ) # Reference model self.beta = args.beta if self.beta == 0.0: # If beta is 0.0, the reference model is not needed self.ref_model = None elif is_peft_model(model): # If PEFT is used, the reference model is not needed since the adapter can be disabled # to revert to the initial model. self.ref_model = None else: # For deepspeed, fsdp or non-distributed models, create a reference model from scratch model_init_kwargs = args.model_init_kwargs or {} # Distributed training requires device_map=None ("auto" fails) if self.args.distributed_state.distributed_type in ["MULTI_GPU", "DEEPSPEED"]: model_init_kwargs["device_map"] = None self.ref_model = create_model_from_path(get_config_model_id(self.model.config), **model_init_kwargs) # Disable dropout in the models if args.disable_dropout: disable_dropout_in_model(model) if self.ref_model is not None: disable_dropout_in_model(self.ref_model) # Cast LM Head To FP32 if args.cast_lm_head_to_fp32: def _cast_lm_head_to_fp32(target_model: PreTrainedModel): """Cast lm_head to fp32 while preserving embedding output dtype if tied.""" def cast_inputs_to_fp32(module, inputs): # Preserve other positional args and kwargs untouched if not inputs: return inputs return (inputs[0].to(torch.float32),) + inputs[1:] original_dtype_local = target_model.lm_head.weight.dtype target_model.lm_head = target_model.lm_head.float() target_model.lm_head.register_forward_pre_hook(cast_inputs_to_fp32) if target_model.config.tie_word_embeddings: def cast_outputs_to_original_dtype(module, args, output): return output.to(original_dtype_local) # Only cast activations; weights are now fp32 (intentional for numerical stability of logits) target_model.model.embed_tokens.register_forward_hook(cast_outputs_to_original_dtype) _cast_lm_head_to_fp32(model) if self.ref_model is not None: _cast_lm_head_to_fp32(self.ref_model) # Liger loss if self.use_liger_kernel: if not is_liger_kernel_available(): raise ImportError( "Liger is required to use `use_liger_kernel` as the GRPO loss. Run `pip install liger-kernel`." ) # redirect the model.module forward to the model forward to ensure pre-forward hooks are called self._forward_redirection = _ForwardRedirection() self.liger_grpo_loss = LigerFusedLinearGRPOLoss( beta=self.beta, epsilon_low=self.epsilon_low, epsilon_high=self.epsilon_high, temperature=self.temperature, use_ref_model=self.beta != 0.0, loss_type=self.loss_type, max_completion_length=self.max_completion_length, importance_sampling_level=self.importance_sampling_level, ) # Initialize the metrics self._metrics = {"train": defaultdict(list), "eval": defaultdict(list)} self._total_train_tokens = 0 self._current_train_step_time = 0.0 self.log_completions = args.log_completions self.log_unique_prompts = args.log_unique_prompts self.num_completions_to_print = args.num_completions_to_print # Keep logs sized to the generation batch to record only outputs from the latest model update. self._logs = { "images": deque(maxlen=args.generation_batch_size), "prompt": deque(maxlen=args.generation_batch_size), "completion": deque(maxlen=args.generation_batch_size), "rewards": defaultdict(lambda: deque(maxlen=args.generation_batch_size)), "advantages": deque(maxlen=args.generation_batch_size), "extra": defaultdict(lambda: deque(maxlen=args.generation_batch_size)), } # Buffers for user-logged data from reward functions, flushed after gathering self._pending_extra_logs = defaultdict(list) self._pending_metrics = defaultdict(list) # Ensure each process receives a unique seed to prevent duplicate completions when generating with # transformers if num_generations exceeds per_device_train_batch_size. We could skip it if we use vLLM, but # it's safer to set it in all cases. set_seed(args.seed, device_specific=True) if self.use_vllm: # Initialize vLLM generation backend self.vllm_generation = VLLMGeneration( model=self.model, accelerator=self.accelerator, is_fsdp_enabled=self.is_fsdp_enabled, processing_class=self.processing_class, # vLLM configuration mode=args.vllm_mode, structured_outputs_regex=args.vllm_structured_outputs_regex, # Server mode configuration server_base_url=args.vllm_server_base_url, server_host=args.vllm_server_host, server_port=args.vllm_server_port, group_port=args.vllm_group_port, server_timeout=args.vllm_server_timeout, # Colocate mode configuration tensor_parallel_size=args.vllm_tensor_parallel_size, gpu_memory_utilization=args.vllm_gpu_memory_utilization, max_model_length=args.vllm_max_model_length, max_num_seqs=args.per_device_train_batch_size * args.vllm_tensor_parallel_size * args.steps_per_generation, enable_sleep_mode=args.vllm_enable_sleep_mode, model_impl=args.vllm_model_impl, # Generation configuration repetition_penalty=self.repetition_penalty, temperature=self.temperature, top_p=self.top_p, top_k=self.top_k, min_p=self.min_p, max_completion_length=self.max_completion_length, logprobs=0, # we only need the generated token logprobs for the importance sampling correction generation_kwargs=args.generation_kwargs, ) self._last_loaded_step = -1 # tag to avoid useless loading during grad accumulation else: generation_kwargs = { "max_new_tokens": self.max_completion_length, "do_sample": True, "pad_token_id": tokenizer.pad_token_id, "bos_token_id": tokenizer.bos_token_id, "eos_token_id": tokenizer.eos_token_id, "temperature": self.temperature, "top_p": self.top_p, "top_k": self.top_k, "min_p": self.min_p, "repetition_penalty": self.repetition_penalty, "cache_implementation": args.cache_implementation, } if args.generation_kwargs is not None: generation_kwargs.update(args.generation_kwargs) self.generation_config = GenerationConfig(**generation_kwargs, disable_compile=True) # Keep training-specific generation kwargs to overwrite model's original generation config self.generation_kwargs = generation_kwargs # Gradient accumulation requires scaled loss. Normally, loss scaling in the parent class depends on whether the # model accepts loss-related kwargs. Since we compute our own loss, this check is irrelevant. We set # self.model_accepts_loss_kwargs to False to enable scaling. self.model_accepts_loss_kwargs = False # Add tags to the model self.model.add_model_tags(self._tag_names) if self.ref_model is not None: if self.is_deepspeed_enabled: self.ref_model = prepare_deepspeed(self.ref_model, self.accelerator) elif self.is_fsdp_enabled: self.ref_model = prepare_fsdp(self.ref_model, self.accelerator) else: self.ref_model = self.accelerator.prepare_model(self.ref_model, evaluation_mode=True) if args.sync_ref_model: if self.beta == 0.0: raise ValueError( "You passed `sync_ref_model=True` while `beta=0.0`, which means the reference model is not used " "during training. Consequently, GRPOTrainer does not create a `ref_model` instance, and there is " "nothing to synchronize. Please set `sync_ref_model=False`, or set `beta` to a non-zero value." ) if is_peft_model(model): raise NotImplementedError( "You passed `sync_ref_model=True` while using a PEFT model, which is currently not supported. " "With PEFT, GRPOTrainer does not keep a separate reference model in memory; instead, it recovers " "reference behavior by temporarily disabling the adapter. As a result, there is no standalone " "`ref_model` instance to synchronize. Use `sync_ref_model=False`, or opt for full fine-tuning if " "you need a synced reference model. If you need `sync_ref_model` to work with PEFT, please open a " "feature request at https://github.com/huggingface/trl/issues." ) self.add_callback(SyncRefModelCallback(ref_model=self.ref_model, accelerator=self.accelerator)) for i, reward_func in enumerate(self.reward_funcs): if isinstance(reward_func, PreTrainedModel): if self.is_deepspeed_enabled: self.reward_funcs[i] = prepare_deepspeed(reward_func, self.accelerator) else: # set device placement to True to make `prepare_model` move `reward_func` to device when using fsdp self.reward_funcs[i] = self.accelerator.prepare_model( reward_func, evaluation_mode=True, device_placement=True ) if self.accelerator.is_main_process and self.log_completions: os.makedirs(os.path.join(self.args.output_dir, "completions"), exist_ok=True) if self.args.log_completions_hub_repo is not None: repo_id = self.args.log_completions_hub_repo create_repo(repo_id, private=self.args.hub_private_repo, repo_type="dataset", exist_ok=True) template_path = pkg_resources.files("trl").joinpath("templates/completions_dataset_card.md") card_data = DatasetCardData( pretty_name="TRL Completion logs", tags=["trl", "trl-logs", "completions"], ) card = DatasetCard.from_template( card_data=card_data, template_path=str(template_path), repo_id=repo_id, hub_model_id=self.args.hub_model_id, ) card.push_to_hub(repo_id) self.commit_scheduler = CommitScheduler( repo_id=repo_id, repo_type="dataset", folder_path=f"{self.args.output_dir}/completions", every=2, # minutes allow_patterns=["*.parquet"], ) def _set_signature_columns_if_needed(self): # If `self.args.remove_unused_columns` is True, non-signature columns are removed. # By default, this method sets `self._signature_columns` to the model's expected inputs (usually, "input_ids" # and "attention_mask"). In GRPOTrainer, we preprocess data, so using the model's signature columns doesn't # work. Instead, we set them to the columns expected by the `training_step` method, hence the override. if self._signature_columns is None: self._signature_columns = ["prompt", "image", "images"] # This method overrides `Trainer.get_train_dataloader` to support our custom batching strategy. # Instead of returning a standard per-step batch (i.e., `per_device_batch_size), our dataloader loads an # *generation* batch (i.e., `per_device_batch_size × steps_per_generation`). This allows us to generate completions # once every steps_per_generation step—rather than once per accumulation step—which is significantly more # efficient. The only change from the original implementation is multiplying the batch size by # `steps_per_generation`. Thus, `_prepare_inputs` is called with this *generation* batch, and it handles the # splitting internally. # Maintenance note: This method is a copy-paste of the original `Trainer.get_train_dataloader` with only one line # modification. def get_train_dataloader(self): return self._get_dataloader( dataset=self.train_dataset, description="Training", batch_size=self._train_batch_size * self.args.steps_per_generation, # < this is the change sampler_fn=self._get_train_sampler, is_training=True, ) def _get_train_sampler(self, dataset: Dataset | None = None) -> Sampler: # Returns a sampler that # 1. ensures each prompt is repeated across multiple processes. This guarantees that identical prompts are # distributed to different GPUs, allowing rewards to be computed and normalized correctly within each prompt # group. Using the same seed across processes ensures consistent prompt assignment, preventing discrepancies # in group formation. # 2. repeats the batch multiple times to allow reusing generations across multiple updates. Refer to # _prepare_inputs to see how the generations are stored and reused. # In the following figure, the values are the prompt indices. The first row shows the first sampled batch, the # second row shows the second sampled batch, and so on. # # | GPU 0 | GPU 1 | # # global_step step <-───> num_generations=2 # <-───────> per_device_train_batch_size=3 # grad_accum ▲ ▲ 0 0 0 0 1 1 2 2 <- Generate for the first `steps_per_generation` (prompts 0 to 11); store the completions; use the first slice to compute the loss # =2 ▼ | 0 1 3 3 4 4 5 5 <- Take the stored generations and use the second slice to compute the loss # | # | 1 2 6 6 7 7 8 8 <- Take the stored generations and use the third slice to compute the loss # steps_per_gen=4 ▼ 1 3 9 9 10 10 11 11 <- Take the stored generations and use the fourth slice to compute the loss # # 2 4 12 12 13 13 14 14 <- Generate for the second `steps_per_generation` (prompts 12 to 23); store the completions; use the first slice to compute the loss # 2 5 15 15 16 16 17 17 <- Take the stored generations and use the second slice to compute the loss # ... if dataset is None: dataset = self.train_dataset return RepeatSampler( data_source=dataset, mini_repeat_count=self.num_generations, batch_size=self.args.generation_batch_size // self.num_generations, repeat_count=self.num_iterations * self.args.steps_per_generation, shuffle=self.shuffle_dataset, seed=self.args.seed, ) def _get_eval_sampler(self, eval_dataset) -> Sampler: # See _get_train_sampler for an explanation of the sampler. return RepeatSampler( data_source=eval_dataset, mini_repeat_count=self.num_generations_eval, seed=self.args.seed, ) @profiling_decorator def _get_last_hidden_state( self, unwrapped_model, input_ids, attention_mask, logits_to_keep, pixel_values=None, image_grid_thw=None, pixel_attention_mask=None, image_sizes=None, pixel_position_ids=None, ): if is_peft_model(unwrapped_model): unwrapped_model = unwrapped_model.base_model.model # Build model inputs - check if the model supports logits_to_keep (some models and VLMs don't) model_inputs = {"input_ids": input_ids, "attention_mask": attention_mask} # For Qwen models: if image_grid_thw is not None and pixel_values is not None: model_inputs["image_grid_thw"] = image_grid_thw # For Gemma, SmolVLM2, LLaVa-Next etc.: if pixel_values is not None: model_inputs["pixel_values"] = pixel_values # For SmolVLM2 if pixel_attention_mask is not None: model_inputs["pixel_attention_mask"] = pixel_attention_mask # For LLaVa-Next if image_sizes is not None: model_inputs["image_sizes"] = image_sizes if pixel_position_ids is not None: model_inputs["pixel_position_ids"] = pixel_position_ids # Only add logits_to_keep if the model supports it if "logits_to_keep" in self.model_kwarg_keys: # We add 1 to `logits_to_keep` because the last logits of the sequence is later excluded model_inputs["logits_to_keep"] = logits_to_keep + 1 model_inputs["use_cache"] = False # only used in generation; set False to suppress warnings last_hidden_state = unwrapped_model.model(**model_inputs).last_hidden_state # Exclude the last value: it corresponds to the next token pred last_hidden_state = last_hidden_state[:, :-1, :] # (B, L-1, H) # Only keep the last logits_to_keep. For model that support logits_to_keep, this is a no-op. last_hidden_state = last_hidden_state[:, -logits_to_keep:, :] # (B, logits_to_keep, H) return last_hidden_state def get_high_entropy_mask(self, entropies: torch.Tensor, mask: torch.Tensor, threshold: float) -> torch.Tensor: """ Returns a binary mask identifying tokens whose entropy exceeds a given quantile threshold. Args: entropies (`torch.Tensor`): Tensor of shape (batch_size, seq_len) with per-token entropy values. mask (`torch.Tensor`): Binary mask of the same shape as `entropies`, where `1` indicates valid tokens and `0` padding. threshold (`float`): Quantile threshold between `0.0` and `1.0` to select high-entropy tokens. Returns: `torch.Tensor`: Boolean mask of shape (batch_size, seq_len), where `True` indicates tokens with entropy >= threshold and `False` otherwise. """ local = entropies[mask.bool()].float() # Use a negative pad_value as a sentinel because entropy values are always >= 0. # This guarantees that the sentinel cannot collide with any real entropy value. pad_value = -1e9 # Pad across processes so that every rank has the same tensor length padded = self.accelerator.pad_across_processes(local, dim=0, pad_index=pad_value) gathered = self.accelerator.gather(padded) # Drop sentinel values (safe because no entropy can be negative) gathered = gathered[gathered != pad_value] if gathered.numel() == 0: return torch.zeros_like(entropies, dtype=torch.bool) entropy_threshold = torch.quantile(gathered, threshold) masked_entropies = entropies * mask.float() entropy_mask = masked_entropies >= entropy_threshold return entropy_mask & mask.bool() # ensure padding tokens are always masked out @profiling_decorator def _get_per_token_logps_and_entropies( self, model, input_ids, attention_mask, logits_to_keep, batch_size=None, compute_entropy=False, pixel_values=None, image_grid_thw=None, num_images=None, pixel_attention_mask=None, image_sizes=None, token_type_ids=None, mm_token_type_ids=None, pixel_position_ids=None, ) -> dict[str, torch.Tensor | None]: """Compute log-probs and (optionally) entropies for each token.""" batch_size = batch_size or input_ids.size(0) # Chunk inputs into smaller batches to reduce memory peak all_logps = [] all_entropies = [] for start in range(0, input_ids.size(0), batch_size): input_ids_batch = input_ids[start : start + batch_size] attention_mask_batch = attention_mask[start : start + batch_size] # Build model inputs - check if the model supports logits_to_keep (some models and VLMs don't) model_inputs = {"input_ids": input_ids_batch, "attention_mask": attention_mask_batch} if image_grid_thw is not None and pixel_values is not None: rows_per_image = image_grid_thw.prod(dim=-1) rows_per_sample = torch.split(rows_per_image, num_images) rows_per_sample = torch.stack([s.sum() for s in rows_per_sample]) cum_rows = torch.cat([torch.tensor([0], device=rows_per_sample.device), rows_per_sample.cumsum(0)]) row_start, row_end = cum_rows[start].item(), cum_rows[start + batch_size].item() model_inputs["pixel_values"] = pixel_values[row_start:row_end] cum_imgs = torch.tensor([0] + num_images).cumsum(0) img_start, img_end = cum_imgs[start], cum_imgs[start + batch_size] model_inputs["image_grid_thw"] = image_grid_thw[img_start:img_end] elif pixel_values is not None: model_inputs["pixel_values"] = pixel_values[start : start + batch_size] if pixel_attention_mask is not None: model_inputs["pixel_attention_mask"] = pixel_attention_mask[start : start + batch_size] if image_sizes is not None: model_inputs["image_sizes"] = image_sizes[start : start + batch_size] if token_type_ids is not None: model_inputs["token_type_ids"] = token_type_ids[start : start + batch_size] if mm_token_type_ids is not None: model_inputs["mm_token_type_ids"] = mm_token_type_ids[start : start + batch_size] if pixel_position_ids is not None: model_inputs["pixel_position_ids"] = pixel_position_ids[start : start + batch_size] # Only add logits_to_keep if the model supports it if "logits_to_keep" in self.model_kwarg_keys: # We add 1 to `logits_to_keep` because the last logits of the sequence is later excluded model_inputs["logits_to_keep"] = logits_to_keep + 1 model_inputs["use_cache"] = False # only used in generation; set False to suppress warnings logits = model(**model_inputs).logits # Exclude the last value: it corresponds to the next token pred logits = logits[:, :-1, :] # (B, L-1, H) # Only keep the last logits_to_keep. For model that support logits_to_keep, this is a no-op. logits = logits[:, -logits_to_keep:, :] # (B, logits_to_keep, H) # Divide logits by sampling temperature. # See https://huggingface.co/blog/the_n_implementation_details_of_rlhf_with_ppo#policy-training-implementation-details logits.div_(self.temperature) completion_ids = input_ids_batch[:, -logits_to_keep:] logps = selective_log_softmax(logits, completion_ids) # compute logprobs all_logps.append(logps) if compute_entropy: with torch.no_grad(): entropies = entropy_from_logits(logits) all_entropies.append(entropies) logps = torch.cat(all_logps, dim=0) entropies = torch.cat(all_entropies, dim=0) if compute_entropy else None return logps, entropies def training_step(self, model, inputs, num_items_in_batch): time_before = time.perf_counter() output = super().training_step(model, inputs, num_items_in_batch) self._step += 1 time_after = time.perf_counter() self._current_train_step_time += time_after - time_before if self._step % self.current_gradient_accumulation_steps == 0: self._metrics["train"]["step_time"].append(self._current_train_step_time) self._current_train_step_time = 0.0 return output @profiling_decorator def _prepare_inputs(self, generation_batch: dict[str, torch.Tensor | Any]) -> dict[str, torch.Tensor | Any]: # Prepares inputs for model training/evaluation by managing completion generation and batch handling. # During training: # - Receives the local generation batch (Per-GPU batch size × steps per generation) # from the modified training dataloader instead of the standard local batch # - Generates completions once for the entire generation batch and splits it into batches of size # `per_device_train_batch_size` # - Buffers these completions and returns the appropriate slice for the current accumulation step # - Optimizes by regenerating completions only periodically (every steps_per_generation * num_iterations) # During evaluation: # - The input is treated as a standard local batch (no accumulation, no multiple iterations) # - Completions are generated for each batch without buffering or reuse # Returns a single local batch in both cases. mode = "train" if self.model.training else "eval" if mode == "train": generate_every = self.args.steps_per_generation * self.num_iterations if self._step % generate_every == 0 or self._buffered_inputs is None: # self._buffered_inputs=None can occur when resuming from a checkpoint generation_batch = self._generate_and_score_completions(generation_batch) generation_batch = split_pixel_values_by_grid(generation_batch) generation_batch = shuffle_sequence_dict(generation_batch) generation_batches = split_tensor_dict(generation_batch, self.args.steps_per_generation) self._buffered_inputs = [unsplit_pixel_values_by_grid(batch) for batch in generation_batches] inputs = self._buffered_inputs[self._step % self.args.steps_per_generation] else: # In evaluation, there is neither batch grouping for generation, nor multiple iterations, hence # local generation batch == local eval batch inputs = self._generate_and_score_completions(generation_batch) return inputs def _log_completion_extra(self, column: str, values: list): """ Log extra columns to the completions table. Called from reward functions via the `log_extra` kwarg. Args: column (`str`): Name of the column to add. values (`list`): Values for the column, one per sample in the batch. """ self._pending_extra_logs[column].extend(values) def _log_metric(self, name: str, value: float): """ Log a scalar metric from a reward function. Called via the `log_metric` kwarg. Values are averaged over each logging step and reported alongside built-in metrics like `kl` and `entropy`. Args: name (`str`): Name of the metric. value (`float`): Scalar value for this batch. """ self._pending_metrics[name].append(value) @profiling_decorator def _calculate_rewards(self, inputs, prompts, completions, completion_ids_list): device = self.accelerator.device rewards_per_func = torch.zeros(len(prompts), len(self.reward_funcs), device=device) # Repeat all input columns (but "prompt", "completion", and "completion_ids") to match the num of generations keys = [key for key in inputs[0] if key not in ["prompt", "completion", "completion_ids"]] reward_kwargs = {key: [example[key] for example in inputs] for key in keys} # This allows for dynamic reward shaping based on training progress. reward_kwargs["trainer_state"] = self.state # Allow reward functions to log extra columns to the completions table. reward_kwargs["log_extra"] = self._log_completion_extra # Allow reward functions to log additional scalar metrics. reward_kwargs["log_metric"] = self._log_metric async_funcs_info = [] # async custom functions for asyncio.gather for i, (reward_func, reward_processing_class, reward_func_name) in enumerate( zip(self.reward_funcs, self.reward_processing_classes, self.reward_func_names, strict=True) ): if isinstance(reward_func, nn.Module): # Module (no PretrainedModel) for compat with compiled models with profiling_context(self, reward_func_name): if is_conversational(inputs[0]): messages = [{"messages": p + c} for p, c in zip(prompts, completions, strict=True)] texts = [ apply_chat_template(x, reward_processing_class, **self.chat_template_kwargs)["text"] for x in messages ] else: texts = [p + c for p, c in zip(prompts, completions, strict=True)] reward_inputs = reward_processing_class( text=texts, return_tensors="pt", padding=True, padding_side="right", add_special_tokens=False ) reward_inputs = super()._prepare_inputs(reward_inputs) with torch.inference_mode(): rewards_per_func[:, i] = reward_func(**reward_inputs).logits[:, 0] # Shape (B*G,) elif inspect.iscoroutinefunction(reward_func): # Separate async reward funcs to run them in parallel later async_funcs_info.append((i, reward_func, reward_func_name)) else: # Run synchronous reward function with profiling_context(self, reward_func_name): if self.environments is not None: reward_kwargs["environments"] = self.environments output_reward_func = reward_func( prompts=prompts, completions=completions, completion_ids=completion_ids_list, **reward_kwargs ) # Convert None values to NaN output_reward_func = [reward if reward is not None else torch.nan for reward in output_reward_func] rewards_per_func[:, i] = torch.tensor(output_reward_func, dtype=torch.float32, device=device) # Execute async custom functions in parallel using asyncio.gather if async_funcs_info: async def _invoke_async(index, func, func_name): with profiling_context(self, func_name): output = await func( prompts=prompts, completions=completions, completion_ids=completion_ids_list, **reward_kwargs ) output = [r if r is not None else torch.nan for r in output] return index, output async def _run_async_funcs(): coros = [_invoke_async(i, func, func_name) for (i, func, func_name) in async_funcs_info] return await asyncio.gather(*coros) async_results = asyncio.run_coroutine_threadsafe(_run_async_funcs(), self.async_loop).result() for idx, output_reward_func in async_results: rewards_per_func[:, idx] = torch.tensor(output_reward_func, dtype=torch.float32, device=device) # If all reward functions return None for a given row, issue a detailed warning if torch.isnan(rewards_per_func).all(dim=1).any(): nan_row_idx = torch.isnan(rewards_per_func).all(dim=1).nonzero(as_tuple=True)[0][0] row_reward_kwargs = { key: value[nan_row_idx] for key, value in reward_kwargs.items() if key not in ("trainer_state", "log_extra", "log_metric") } row_reward_kwargs["prompt"] = prompts[nan_row_idx] row_reward_kwargs["completion"] = completions[nan_row_idx] logger.warning( f"All reward functions returned None for the following kwargs:\n{row_reward_kwargs}\n" "Please ensure that at least one reward function returns a valid reward." ) # Gather the reward per function: this part is crucial, because the rewards are normalized per group and the # completions may be distributed across processes rewards_per_func = gather(rewards_per_func) return rewards_per_func def _tokenize_prompts(self, prompts: list): """Tokenize prompts and extract images/multimodal fields for generation.""" if is_conversational({"prompt": prompts[0]}): # Extract images from messages for VLM support images = [] has_images = False for prompt in prompts: prompt_images = [] for message in prompt: if isinstance(message["content"], list): for part in message["content"]: if part["type"] == "image": prompt_images.append(part["image"]) has_images = True images.append(prompt_images if prompt_images else None) images = images if has_images else None # We pass padding=True to work around a bug introduced in transformers 5.2.0 in some processors # (e.g. Qwen2.5-VL) that crash on batched unpadded input. We then unpad input_ids using attention_mask. # See: https://github.com/huggingface/transformers/issues/44514 tokenized = self.processing_class.apply_chat_template( conversation=prompts, tools=self.tools or None, # `or None`: Llama bug: it renders tool boilerplate for tools=[] chat_template=self.chat_template, add_generation_prompt=True, tokenize=True, return_dict=True, padding=True, **self.chat_template_kwargs, ) # Unpad input_ids: remove padding tokens using attention_mask to get per-sequence lists prompt_ids = [ [tok for tok, m in zip(ids, mask, strict=True) if m] for ids, mask in zip(tokenized["input_ids"], tokenized["attention_mask"], strict=True) ] # For VLMs, the processor returns extra multimodal fields (pixel_values, image_grid_thw, etc.) multimodal_fields = {k: v for k, v in tokenized.items() if k not in ("input_ids", "attention_mask")} else: prompt_ids = self.processing_class(text=prompts)["input_ids"] images = None multimodal_fields = {} return prompt_ids, images, multimodal_fields def _generate_single_turn(self, prompt_ids, images, multimodal_fields): device = self.accelerator.device mode = "train" if self.model.training else "eval" # Generate completions using either vLLM or regular generation if self.use_vllm: # Sync weights if training step changed if self.state.global_step != self._last_loaded_step: with profiling_context(self, "sync_weights"): self.vllm_generation.sync_weights() self._last_loaded_step = self.state.global_step # Generate using vLLM with raw token IDs num_generations = self.num_generations if mode == "train" else self.num_generations_eval _, completion_ids, logprobs, _ = self.vllm_generation.generate( prompts=prompt_ids, images=images, num_generations=num_generations, profiler=profiling_context(self, "vLLM.generate"), ) # vLLM returns per-token top-k logprobs; keep only the top-1 (sampled token) logprob logprobs = [[lp[0] for lp in seq] for seq in logprobs] elif self.use_transformers_paged: with ( profiling_context(self, "transformers.generate_batch"), unwrap_model_for_generation( self.model_wrapped, self.accelerator, gather_deepspeed3_params=self.args.ds3_gather_for_generation ) as unwrapped_model, torch.no_grad(), FSDP.summon_full_params(self.model_wrapped, recurse=False) if self.is_fsdp_enabled else nullcontext(), ): # Cast to the appropriate dtype based on training configuration if self.args.bf16: unwrapped_model.to(torch.bfloat16) elif self.args.fp16: unwrapped_model.to(torch.float16) if self.args.cast_lm_head_to_fp32: unwrapped_model.lm_head.to(torch.float32) with torch.inference_mode(): # Continuous batching API expects 'inputs' arg only all_outputs = unwrapped_model.generate_batch( prompt_ids, generation_config=self.generation_config, progress_bar=False ) unwrapped_model.train() # restore training mode, as generate_batch forces eval mode completion_ids = [output.generated_tokens for output in all_outputs.values()] logprobs = None # not used in this case else: # Regular generation path: left-pad token IDs into tensors prompt_tensors = [torch.tensor(ids) for ids in prompt_ids] padded_ids = pad(prompt_tensors, padding_value=self.pad_token_id, padding_side="left") attention_mask = pad([torch.ones_like(t) for t in prompt_tensors], padding_value=0, padding_side="left") generate_inputs = {"input_ids": padded_ids, "attention_mask": attention_mask} # For VLMs, include multimodal fields as tensors (pixel_values, image_grid_thw, etc.) for k, v in multimodal_fields.items(): if isinstance(v, torch.Tensor): generate_inputs[k] = v elif isinstance(v, list) and v and isinstance(v[0], list): # Per-token field (e.g., token_type_ids): left-pad like input_ids generate_inputs[k] = pad([torch.tensor(x) for x in v], padding_value=0, padding_side="left") else: generate_inputs[k] = torch.tensor(np.array(v)) generate_inputs = super()._prepare_inputs(generate_inputs) with ( profiling_context(self, "transformers.generate"), unwrap_model_for_generation( self.model_wrapped, self.accelerator, gather_deepspeed3_params=self.args.ds3_gather_for_generation, generation_kwargs=self.generation_kwargs, # Override model.generation_config with generation_kwargs to fix transformers#42762 ) as unwrapped_model, torch.no_grad(), FSDP.summon_full_params(self.model_wrapped, recurse=False) if self.is_fsdp_enabled else nullcontext(), ): prompt_completion_ids = unwrapped_model.generate( **generate_inputs, generation_config=self.generation_config ) # Compute prompt length and extract completion ids prompt_length = generate_inputs["input_ids"].size(1) completion_ids = prompt_completion_ids[:, prompt_length:] # Mask everything after the first EOS token is_eos = completion_ids == self.eos_token_id eos_idx = torch.full((is_eos.size(0),), is_eos.size(1), dtype=torch.long, device=device) eos_idx[is_eos.any(dim=1)] = is_eos.int().argmax(dim=1)[is_eos.any(dim=1)] sequence_indices = torch.arange(is_eos.size(1), device=device).expand(is_eos.size(0), -1) completion_mask = (sequence_indices <= eos_idx.unsqueeze(1)).int() completion_ids = [ c[m].tolist() for c, m in zip(completion_ids.cpu(), completion_mask.bool().cpu(), strict=True) ] logprobs = None # not used in this case return completion_ids, logprobs def _get_tool_suffix_ids(self, tool_messages): """Get token IDs for tool result formatting by using a minimal dummy conversation.""" dummy_messages = [{"role": "user", "content": "dummy"}, {"role": "assistant", "content": "dummy"}] prefix_ids = self.processing_class.apply_chat_template( dummy_messages, add_generation_prompt=False, chat_template=self.chat_template, return_dict=False, **self.chat_template_kwargs, ) full_ids = self.processing_class.apply_chat_template( dummy_messages + tool_messages, add_generation_prompt=True, chat_template=self.chat_template, return_dict=False, **self.chat_template_kwargs, ) # Some chat templates (notably Qwen3/Qwen3.5) render "...<|im_end|>\n" after an assistant/tool block. # When we compute `suffix_ids` by slicing `full_ids`, we must align the slicing boundary to # EOS (not EOS + newline). last_eos_idx = max(i for i, tok_id in enumerate(prefix_ids) if tok_id == self.eos_token_id) prefix_ids = prefix_ids[: last_eos_idx + 1] if full_ids[: len(prefix_ids)] != prefix_ids: raise ValueError("Unexpected tokenization: the EOS-trimmed prefix IDs are not a prefix of the full IDs.") return full_ids[len(prefix_ids) :] def _tool_call_loop(self, prompts, prompt_ids, completion_ids, completions, logprobs, images, multimodal_fields): # Tool execution loop: execute tools, then regenerate completions with tool results appended to the prompt tool_calls = [completion[0].get("tool_calls") for completion in completions] idxs_with_tool = [idx for idx, tool_call in enumerate(tool_calls) if tool_call] tool_calls = [tool_calls[idx] for idx in idxs_with_tool] tool_mask = [[1] * len(ids) for ids in completion_ids] # 0 for tool result tokens, 1 elsewhere tool_call_count = 0 tool_failure_count = 0 iteration_num = 0 while idxs_with_tool and iteration_num < self.max_tool_calling_iterations: prompt_completion_tools = [prompts[i] for i in idxs_with_tool] # select only prompts that need tool calls # Call the tools, and build the new prompt for generation for idx in range(len(idxs_with_tool)): idx_with_tool = idxs_with_tool[idx] tool_call_list = tool_calls[idx] prompt_completion_tool = prompt_completion_tools[idx] sync_tool_dict = self._sync_tool_dicts[idx_with_tool] async_tool_dict = self._async_tool_dicts[idx_with_tool] # Append the last assistant message (which triggered tool_calls) to the prompt prompt_completion_tool.append(completions[idx_with_tool][-1]) async_coros = [] tool_call_results = [] for tool_call in tool_call_list: tool_call_count += 1 if tool_call["type"] == "function": function = tool_call["function"] name = function["name"] try: if name in sync_tool_dict: tool_call_results.append((name, sync_tool_dict[name](**function["arguments"]))) elif name in async_tool_dict: async_coros.append((name, async_tool_dict[name](**function["arguments"]))) else: raise ValueError(f"Tool {name} not found.") except Exception as e: tool_failure_count += 1 result = {"error": str(e)} tool_call_results.append((name, result)) else: tool_failure_count += 1 name = tool_call.get("name", "unknown") tool_call_results.append((name, {"error": f"Unsupported tool call type: {tool_call['type']}"})) if async_coros: async def _run_async_tools(async_coros): coros = [coro for _, coro in async_coros] results = await asyncio.gather(*coros, return_exceptions=True) return [(name, result) for (name, _), result in zip(async_coros, results, strict=False)] async_results = asyncio.run_coroutine_threadsafe( _run_async_tools(async_coros), self.async_loop ).result() for name, result in async_results: if isinstance(result, Exception): tool_failure_count += 1 tool_call_results.append((name, {"error": str(result)})) else: tool_call_results.append((name, result)) for name, result in tool_call_results: tool_message = {"role": "tool", "name": name, "content": str(result)} prompt_completion_tool.append(tool_message) completions[idx_with_tool].append(tool_message) # Build token IDs by concatenation: prompt + completion + tool_suffix. prompt_completion_tool_ids = [] for idx in range(len(idxs_with_tool)): idx_with_tool = idxs_with_tool[idx] # Extract trailing tool messages from completions tool_messages = [] for message in reversed(completions[idx_with_tool]): if message["role"] == "tool": tool_messages.insert(0, message) else: break suffix_ids = self._get_tool_suffix_ids(tool_messages) prompt_completion_tool_ids.append( prompt_ids[idx_with_tool] + completion_ids[idx_with_tool] + suffix_ids ) # Filter samples whose length exceeds max allowed length. This is important, because both # vLLM and transformers will error out if the input is longer than the model's max length. if self.use_vllm and self.vllm_mode == "colocate": max_model_len = self.vllm_generation.llm.llm_engine.model_config.max_model_len elif self.use_vllm and self.vllm_mode == "server": max_model_len = self.model.config.max_position_embeddings elif not self.use_vllm: max_model_len = self.model.config.max_position_embeddings else: raise NotImplementedError( f"Unsupported mode detected: use_vllm={self.use_vllm}, vllm_mode={self.vllm_mode}" ) overlong = [len(pct) >= max_model_len for pct in prompt_completion_tool_ids] for idx in range(len(idxs_with_tool)): idx_with_tool = idxs_with_tool[idx] if overlong[idx]: prompt_length = len(prompt_ids[idx_with_tool]) ct = prompt_completion_tool_ids[idx][prompt_length : prompt_length + self.max_completion_length] completion_ids[idx_with_tool] = ct tool_mask[idx_with_tool] += [1] * (len(ct) - len(tool_mask[idx_with_tool])) if logprobs is not None: logprobs[idx_with_tool] += [0.0] * (len(ct) - len(logprobs[idx_with_tool])) # Keep only non-overlong items for further processing idxs_with_tool = [idx for idx, o in zip(idxs_with_tool, overlong, strict=True) if not o] prompt_completion_tools = [pct for pct, o in zip(prompt_completion_tools, overlong, strict=True) if not o] prompt_completion_tool_ids = [ pct for pct, o in zip(prompt_completion_tool_ids, overlong, strict=True) if not o ] if not idxs_with_tool: break # all overlong, exit tool loop # Filter images and multimodal fields to match the current subset (index into full batch) loop_images = [images[i] for i in idxs_with_tool] if images else None loop_multimodal_fields = ( {k: [v[i] for i in idxs_with_tool] for k, v in multimodal_fields.items()} if multimodal_fields else {} ) # Generate new completions after tool execution (using concatenated IDs, no re-tokenization) post_tool_ids, post_tool_logprobs = self._generate_single_turn( prompt_completion_tool_ids, loop_images, loop_multimodal_fields ) # Truncate so that pct[len(prompt_ids[idx]) :] + post_tool does not exceed max_completion_length for idx in range(len(idxs_with_tool)): idx_with_tool = idxs_with_tool[idx] prompt_len = len(prompt_ids[idx_with_tool]) completion_tool_ids = prompt_completion_tool_ids[idx][prompt_len:] excess_length = len(completion_tool_ids) + len(post_tool_ids[idx]) - self.max_completion_length if excess_length > 0: # If exceeding max length, truncate post_tool_ids post_tool_ids[idx] = post_tool_ids[idx][:-excess_length] if logprobs is not None: post_tool_logprobs[idx] = post_tool_logprobs[idx][:-excess_length] excess_length = len(completion_tool_ids) + len(post_tool_ids[idx]) - self.max_completion_length if excess_length > 0: # If still exceeding max length, truncate completion_tool_ids as well prompt_completion_tool_ids[idx] = prompt_completion_tool_ids[idx][:-excess_length] # Update tool_mask: the tool result should be 0 and the post-tool 1 for idx in range(len(idxs_with_tool)): idx_with_tool = idxs_with_tool[idx] prompt_completion_tool_length = len(prompt_completion_tool_ids[idx]) prompt_length = len(prompt_ids[idx_with_tool]) completion_length = len(completion_ids[idx_with_tool]) post_tool_length = len(post_tool_ids[idx]) tool_length = prompt_completion_tool_length - prompt_length - completion_length tool_mask[idx_with_tool] += [0] * tool_length + [1] * post_tool_length if logprobs is not None: logprobs[idx_with_tool] += [0.0] * tool_length + post_tool_logprobs[idx] # Update completion_ids with the new completions (after tool execution) for idx in range(len(idxs_with_tool)): idx_with_tool = idxs_with_tool[idx] prompt_length = len(prompt_ids[idx_with_tool]) pct = prompt_completion_tool_ids[idx] # = prompt-completion-tool completion_ids[idx_with_tool] = pct[prompt_length:] + post_tool_ids[idx] # Decode post-tool completions post_tool_completions = [ parse_response(self.processing_class, ids) if ids else {} for ids in post_tool_ids ] # Add post-tool completions to the existing completions for idx in range(len(idxs_with_tool)): idx_with_tool = idxs_with_tool[idx] if post_tool_completions[idx]: # {} if post-tool completions completely truncated completions[idx_with_tool].append(post_tool_completions[idx]) # Check for further tool calls tool_calls = [completion.get("tool_calls") for completion in post_tool_completions] idxs_with_tool = [idx for idx, tool_call in zip(idxs_with_tool, tool_calls, strict=True) if tool_call] tool_calls = [tool_call for tool_call in tool_calls if tool_call] iteration_num += 1 return tool_mask, completions, completion_ids, logprobs, tool_call_count, tool_failure_count def _generate(self, prompts: list): device = self.accelerator.device mode = "train" if self.model.training else "eval" # Copy the prompts to avoid modifying the original list prompts = copy.deepcopy(prompts) if self.rollout_func is not None: # Keep vLLM weights in sync for custom rollouts that rely on vLLM utilities. if self.use_vllm and self.state.global_step != self._last_loaded_step: with profiling_context(self, "sync_weights"): self.vllm_generation.sync_weights() self._last_loaded_step = self.state.global_step # Pass prompts to rollout_func preserving structured messages. # Chat templating must happen inside rollout_func, at the backend boundary, so that # multimodal content (images, typed content blocks) is not lost before rollout logic runs. output = self.rollout_func(prompts, self) required_keys = {"prompt_ids", "completion_ids", "logprobs"} missing_keys = required_keys - output.keys() if missing_keys: missing_keys_list = sorted(missing_keys) raise ValueError(f"rollout_func must return keys {missing_keys_list} in its output dict.") extra_fields = {k: v for k, v in output.items() if k not in required_keys} prompt_ids, completion_ids, logprobs = output["prompt_ids"], output["completion_ids"], output["logprobs"] else: prompt_ids, images, multimodal_fields = self._tokenize_prompts(prompts) completion_ids, logprobs = self._generate_single_turn(prompt_ids, images, multimodal_fields) extra_fields = {} # Decode completions. It's important to use `parse_response` when possible, because it handles tool calls. if is_conversational({"prompt": prompts[0]}): if ( Version(transformers.__version__) >= Version("5.0.0") # parse_response added in v5 and isinstance(self.processing_class, PreTrainedTokenizerBase) # doesn't work with processors and hasattr(self.processing_class, "response_schema") # attribute not set by default for now and self.processing_class.response_schema is not None # only works if the tokenizer has a schema ): completions = [[parse_response(self.processing_class, ids)] for ids in completion_ids] else: contents = self.processing_class.batch_decode(completion_ids, skip_special_tokens=True) completions = [[{"role": "assistant", "content": content}] for content in contents] else: completions = self.processing_class.batch_decode(completion_ids, skip_special_tokens=True) # Extract tool calls from the completions and (possibly) execute them if self.tools: ( tool_mask, completions, completion_ids, logprobs, tool_call_count, tool_failure_count, ) = self._tool_call_loop( prompts, prompt_ids, completion_ids, completions, logprobs, images, multimodal_fields ) else: # Support custom env_mask from rollout_func (e.g., for environment feedback masking) # Internally treated as tool_mask - marks model tokens (1) vs external tokens (0) tool_mask = extra_fields.pop("env_mask", None) # Get completion length per sequence, used for logging prompt_lengths = torch.tensor([len(ids) for ids in prompt_ids], device=device) if tool_mask is not None: # count only model-generated tokens (tool_mask=1) completion_lengths = torch.tensor([sum(mask) for mask in tool_mask], device=device) else: completion_lengths = torch.tensor([len(ids) for ids in completion_ids], device=device) agg_prompt_lengths = self.accelerator.gather(prompt_lengths) agg_completion_lengths = self.accelerator.gather(completion_lengths) total_prompt_tokens = agg_prompt_lengths.sum() total_completion_tokens = agg_completion_lengths.sum() # = num_items_in_batch, required for the DAPO loss # Log the metrics if mode == "train": self.state.num_input_tokens_seen += (total_prompt_tokens + total_completion_tokens).item() self._metrics[mode]["num_tokens"] = [self.state.num_input_tokens_seen] # Log completion lengths, mean, min, max self._metrics[mode]["completions/mean_length"].append(agg_completion_lengths.float().mean().item()) self._metrics[mode]["completions/min_length"].append(agg_completion_lengths.float().min().item()) self._metrics[mode]["completions/max_length"].append(agg_completion_lengths.float().max().item()) # Identify sequences that terminated with EOS and log their lengths eos_and_pad = [self.eos_token_id, self.pad_token_id] is_truncated = torch.tensor([ids[-1] not in eos_and_pad for ids in completion_ids], device=device) agg_is_truncated = self.accelerator.gather(is_truncated) self._metrics[mode]["completions/clipped_ratio"].append(agg_is_truncated.float().mean().item()) term_completion_lengths = agg_completion_lengths[~agg_is_truncated] if len(term_completion_lengths) == 0: # edge case where no terminated sequences are found term_completion_lengths = torch.zeros(1, device=device) self._metrics[mode]["completions/mean_terminated_length"].append(term_completion_lengths.float().mean().item()) self._metrics[mode]["completions/min_terminated_length"].append(term_completion_lengths.float().min().item()) self._metrics[mode]["completions/max_terminated_length"].append(term_completion_lengths.float().max().item()) if self.tools: agg_tool_call_count = self.accelerator.gather(torch.tensor(tool_call_count, device=device)).sum() tool_call_frequency = (agg_tool_call_count / len(agg_prompt_lengths)).item() self._metrics[mode]["tools/call_frequency"].append(tool_call_frequency) agg_tool_failure_count = self.accelerator.gather(torch.tensor(tool_failure_count, device=device)).sum() failure_frequency = ( (agg_tool_failure_count / agg_tool_call_count).item() if agg_tool_call_count > 0 else 0.0 ) self._metrics[mode]["tools/failure_frequency"].append(failure_frequency) return ( prompt_ids, completion_ids, tool_mask, completions, total_completion_tokens, logprobs, extra_fields, ) def _generate_and_score_completions( self, inputs: list[dict[str, torch.Tensor | Any]] ) -> dict[str, torch.Tensor | Any]: device = self.accelerator.device mode = "train" if self.model.training else "eval" prompts = [x["prompt"] for x in inputs] if self.environments: for prompt, environment, reset_kwargs in zip(prompts, self.environments, inputs, strict=True): observation = environment.reset(**reset_kwargs) if observation is None: continue prompt[-1]["content"] += observation if "images" in inputs[0]: images = [example.get("images") for example in inputs] elif "image" in inputs[0]: images = [[example.get("image")] if example.get("image") is not None else None for example in inputs] else: images = None # Transformers requires at least one image in the batch, otherwise it throws an error if images is not None and all(img_list == [] for img_list in images): images = None # If the prompts are conversational and the inputs contain images, we need to convert the prompts from # [{"role": "user", "content": "What color is the sky?"}] to # [{"role": "user", "content": [{"type": "image", "image": }, {"type": "text", "text": "What color is the sky?"}]}] if images is not None: if not is_conversational(inputs[0]): raise ValueError( "Multimodal training requires conversational prompts. It looks like the dataset contains " "non-conversational inputs, likely because a chat template was applied before passing the dataset " "to the trainer. Please provide the raw conversational prompts and let the trainer apply the chat " "template internally." ) prompts = [ prepare_multimodal_messages(prompt, image_list) for prompt, image_list in zip(prompts, images, strict=True) ] ( prompt_ids_list, completion_ids_list, tool_mask_list, completions, num_items_in_batch, sampling_per_token_logps_list, extra_fields, ) = self._generate(prompts) # Convert lists of token IDs to padded tensors prompt_ids = [torch.tensor(ids) for ids in prompt_ids_list] prompt_mask = [torch.ones_like(ids, dtype=torch.long) for ids in prompt_ids] prompt_ids = pad( prompt_ids, padding_value=self.pad_token_id, padding_side="left", pad_to_multiple_of=self.pad_to_multiple_of, ).to(device=device) prompt_mask = pad( prompt_mask, padding_value=0, padding_side="left", pad_to_multiple_of=self.pad_to_multiple_of ).to(device=device) completion_ids = [torch.tensor(ids) for ids in completion_ids_list] completion_mask = [torch.ones_like(ids, dtype=torch.long) for ids in completion_ids] completion_ids = pad( completion_ids, padding_value=self.pad_token_id, padding_side="right", pad_to_multiple_of=self.pad_to_multiple_of, ).to(device=device) completion_mask = pad( completion_mask, padding_value=0, padding_side="right", pad_to_multiple_of=self.pad_to_multiple_of ).to(device=device) if sampling_per_token_logps_list is not None: sampling_per_token_logps = [torch.tensor(logps) for logps in sampling_per_token_logps_list] sampling_per_token_logps = pad( sampling_per_token_logps, padding_value=0.0, padding_side="right", pad_to_multiple_of=self.pad_to_multiple_of, ).to(device=device) else: sampling_per_token_logps = None if tool_mask_list is not None: tool_mask = [torch.tensor(mask) for mask in tool_mask_list] tool_mask = pad( tool_mask, padding_value=1, padding_side="right", pad_to_multiple_of=self.pad_to_multiple_of ).to(device=device) else: tool_mask = None # If mask_truncated_completions is enabled, zero out truncated completions for attention and loss masking if self.mask_truncated_completions: eos_and_pad = [self.eos_token_id, self.pad_token_id] is_truncated = torch.tensor([ids[-1] not in eos_and_pad for ids in completion_ids_list], device=device) # Mask completion_mask for attention masking completion_mask = completion_mask * (~is_truncated).unsqueeze(1).int() # Also mask tool_mask for consistency in multi-turn training if tool_mask is not None: tool_mask = tool_mask * (~is_truncated).unsqueeze(1).int() # Concatenate prompt_mask with completion_mask for logit computation prompt_completion_ids = torch.cat([prompt_ids, completion_ids], dim=1) # (B, P+C) attention_mask = torch.cat([prompt_mask, completion_mask], dim=1) # (B, P+C) logits_to_keep = completion_ids.size(1) # we only need to compute the logits for the completion tokens batch_size = self.args.per_device_train_batch_size if mode == "train" else self.args.per_device_eval_batch_size num_images = [len(img_list) for img_list in images] if images is not None else None # Get forward_kwargs for models with multimodal inputs if images is not None: prompts_text = [ apply_chat_template( {"prompt": prompt}, self.processing_class, tools=self.tools, **self.chat_template_kwargs )["prompt"] for prompt in prompts ] prompt_inputs = self.processing_class(images=images, text=prompts_text, padding=True, return_tensors="pt") prompt_inputs = super()._prepare_inputs(prompt_inputs) forward_kwargs = {k: v for k, v in prompt_inputs.items() if k not in ["input_ids", "attention_mask"]} else: forward_kwargs = {} # If token_type_ids are used, extend them with zeros for the completion part if "token_type_ids" in forward_kwargs: token_type_ids = forward_kwargs["token_type_ids"] if self.pad_to_multiple_of is not None: # Needed only with pad_to_multiple_of: otherwise prompt_ids and token_type_ids must have equal len padding_size = prompt_ids.size(1) - token_type_ids.size(1) if padding_size > 0: token_type_ids = torch.cat( [token_type_ids.new_zeros((token_type_ids.size(0), padding_size)), token_type_ids], dim=1 ) forward_kwargs["token_type_ids"] = torch.cat( [token_type_ids, token_type_ids.new_zeros(completion_ids.shape)], dim=1 ) # If mm_token_type_ids are used, extend them with zeros for the completion part if "mm_token_type_ids" in forward_kwargs: mm_token_type_ids = forward_kwargs["mm_token_type_ids"] if self.pad_to_multiple_of is not None: # Needed only with pad_to_multiple_of: otherwise prompt_ids and mm_token_type_ids must have equal len padding_size = prompt_ids.size(1) - mm_token_type_ids.size(1) if padding_size > 0: mm_token_type_ids = torch.cat( [mm_token_type_ids.new_zeros((mm_token_type_ids.size(0), padding_size)), mm_token_type_ids], dim=1, ) forward_kwargs["mm_token_type_ids"] = torch.cat( [mm_token_type_ids, mm_token_type_ids.new_zeros(completion_ids.shape)], dim=1 ) # When gradient checkpointing is enabled with use_reentrant=True (non default), calling the model inside a # torch.no_grad() block triggers a harmless PyTorch warning ("None of the inputs have requires_grad=True"). # Temporarily disable checkpointing to avoid this warning during inference. with torch.no_grad(), disable_gradient_checkpointing(self.model, self.args.gradient_checkpointing_kwargs): # If the generation and optimization steps are misaligned—i.e., if generation does not occur at the end of # a full optimizer step (when gradient_accumulation_steps is not a multiple of generate_every)—then the # samples may come from an earlier version of the model. In that case, we need to track old_per_token_logps # for importance sampling. If the steps are aligned, importance sampling isn't necessary and we set # old_per_token_logps to None. # When using vLLM, we always compute old_per_token_logps for importance sampling, it was shown that the # distribution mismatch between vLLM and the training model can be large and harm the training. generate_every = self.args.steps_per_generation * self.num_iterations # generation frequency if self.args.gradient_accumulation_steps % generate_every != 0 or ( self.use_vllm and self.vllm_importance_sampling_correction ): old_per_token_logps, _ = self._get_per_token_logps_and_entropies( self.model, prompt_completion_ids, attention_mask, logits_to_keep, batch_size, num_images=num_images, **forward_kwargs, # may contain pixel_values, image_grid_thw, pixel_attention_mask, image_sizes, pixel_position_ids ) else: old_per_token_logps = None # Compute the importance sampling ratio when using vLLM, to correct for potential distribution mismatch if self.use_vllm and self.vllm_importance_sampling_correction: mask = completion_mask if tool_mask is None else completion_mask * tool_mask per_token_logps_diff = (old_per_token_logps - sampling_per_token_logps) * mask sequence_level_is = self.vllm_importance_sampling_mode in ["sequence_mask", "sequence_truncate"] if sequence_level_is: per_sequence_logps_diff = per_token_logps_diff.sum(dim=-1, keepdim=True) logps_diff = per_sequence_logps_diff else: logps_diff = per_token_logps_diff vllm_importance_sampling_ratio = torch.exp(logps_diff) # vllm_importance_sampling_ratio.shape: # token_* modes: (B, T) (per-token ratio) # sequence_* modes: (B, 1) (per-sequence ratio) if self.vllm_importance_sampling_mode in ["sequence_truncate", "token_truncate"]: vllm_importance_sampling_ratio = torch.clamp( vllm_importance_sampling_ratio, max=self.vllm_importance_sampling_cap ) elif self.vllm_importance_sampling_mode in ["sequence_mask", "token_mask"]: vllm_importance_sampling_ratio = vllm_importance_sampling_ratio.masked_fill( vllm_importance_sampling_ratio > self.vllm_importance_sampling_cap, value=0.0 ) else: raise ValueError( f"Unknown vLLM importance sampling level: {self.vllm_importance_sampling_mode}. Possible values are 'token_truncate', 'token_mask', 'sequence_truncate', and 'sequence_mask'." ) # Compute the per-token log probabilities for the reference model if self.beta != 0.0: if self.ref_model is not None: ref_per_token_logps, _ = self._get_per_token_logps_and_entropies( self.ref_model, prompt_completion_ids, attention_mask, logits_to_keep, batch_size=batch_size, num_images=num_images, **forward_kwargs, # may contain pixel_values, image_grid_thw, pixel_attention_mask, image_sizes, pixel_position_ids ) else: # When training a PEFT adapter, how we obtain the reference depends on the setup: # - New adapter: disabling adapters yields the base model. # - Re-training an existing adapter: an initial copy is loaded under the name "ref". model = self.accelerator.unwrap_model(self.model) with use_adapter(model, adapter_name="ref" if "ref" in model.peft_config else None): ref_per_token_logps, _ = self._get_per_token_logps_and_entropies( self.model, prompt_completion_ids, attention_mask, logits_to_keep, batch_size=batch_size, num_images=num_images, **forward_kwargs, # may contain pixel_values, image_grid_thw, pixel_attention_mask, image_sizes, pixel_position_ids ) else: ref_per_token_logps = None # Decode prompts_text = self.processing_class.batch_decode(prompt_ids, skip_special_tokens=True) completions_text = self.processing_class.batch_decode(completion_ids, skip_special_tokens=True) # Merge extra_fields from rollout_func into inputs for reward functions if extra_fields: for i, inp in enumerate(inputs): for key, values in extra_fields.items(): if isinstance(values, list) and i < len(values): inp[key] = values[i] elif not isinstance(values, list): inp[key] = values # Calculate rewards for each reward function. rewards_per_func aggregates rewards across all processes. This is # important because rewards will be normalized per group, and completions are distributed. We will later slice # rewards_per_func to extract each process's subset. rewards_per_func = self._calculate_rewards(inputs, prompts, completions, completion_ids_list) num_generations = self.num_generations if mode == "train" else self.num_generations_eval if self.multi_objective_aggregation == "sum_then_normalize": # Apply weights to each reward function's output and sum rewards = (rewards_per_func * self.reward_weights.to(device).unsqueeze(0)).nansum(dim=1) mean_grouped_rewards = rewards.view(-1, num_generations).mean(dim=1) mean_grouped_rewards = mean_grouped_rewards.repeat_interleave(num_generations, dim=0) if self.scale_rewards in ["group", "none"]: # If self.scale_rewards = "none", we'll only use std_rewards to check for zero std for logging if num_generations > 1: std_rewards = rewards.view(-1, num_generations).std(dim=1) std_rewards = std_rewards.repeat_interleave(num_generations, dim=0) else: # doesn't occur during training, but could occur in eval when num_generations_eval=1 std_rewards = torch.zeros_like(rewards) elif self.scale_rewards == "batch": # Compute global std if rewards.numel() > 1: std_rewards = rewards.std().expand_as(rewards) else: # doesn't occur during training, but could occur in eval when num_generations_eval=batch_size=1 std_rewards = torch.zeros_like(rewards) else: raise ValueError( f"Invalid value for scale_rewards: {self.scale_rewards}. Must be one of 'batch', 'group', or 'none'." ) advantages = rewards - mean_grouped_rewards if self.scale_rewards != "none": advantages = advantages / (std_rewards + 1e-4) is_std_zero = torch.isclose(std_rewards, torch.zeros_like(std_rewards)) # for logging elif self.multi_objective_aggregation == "normalize_then_sum": grouped = rewards_per_func.view(-1, num_generations, len(self.reward_funcs)) mean_k = torch.nanmean(grouped, dim=1, keepdim=True) std_k = nanstd(grouped, dim=1, keepdim=True) if num_generations > 1 else torch.zeros_like(mean_k) reward_k = (grouped - mean_k) / (std_k + 1e-4) reward_k = reward_k.view(-1, len(self.reward_funcs)) rewards = (reward_k * self.reward_weights.to(device).unsqueeze(0)).nansum(dim=1) std_rewards = rewards.std().expand_as(rewards) if rewards.numel() > 1 else torch.zeros_like(rewards) advantages = (rewards - rewards.mean()) / (std_rewards + 1e-4) is_std_zero = torch.isclose(std_rewards, torch.zeros_like(std_rewards)) # for logging else: raise ValueError( f"Invalid multi_objective_aggregation: {self.multi_objective_aggregation}. Must be " "'sum_then_normalize' or 'normalize_then_sum'." ) # Slice to keep only the local part of the data process_slice = slice( self.accelerator.process_index * len(prompts), (self.accelerator.process_index + 1) * len(prompts), ) all_process_advantages = advantages.clone() # keep the aggregated advantages for logging advantages = advantages[process_slice] # Calculate mean reward per function, but only for samples where the function was applied (non-NaN values) for i, reward_func_name in enumerate(self.reward_func_names): mean_rewards = torch.nanmean(rewards_per_func[:, i]).item() self._metrics[mode][f"rewards/{reward_func_name}/mean"].append(mean_rewards) std_func_rewards = nanstd(rewards_per_func[:, i]).item() self._metrics[mode][f"rewards/{reward_func_name}/std"].append(std_func_rewards) rewards = (rewards_per_func * self.reward_weights.to(rewards_per_func.device).unsqueeze(0)).nansum(dim=1) self._metrics[mode]["reward"].append(rewards.mean().item()) self._metrics[mode]["reward_std"].append(rewards.std().item()) self._metrics[mode]["frac_reward_zero_std"].append(is_std_zero.float().mean().item()) # Log prompt and completion texts self._logs["prompt"].extend(gather_object(prompts_text)) self._logs["completion"].extend(gather_object(completions_text)) for i, name in enumerate(self.reward_func_names): self._logs["rewards"][name].extend(rewards_per_func[:, i].tolist()) self._logs["advantages"].extend(all_process_advantages.tolist()) # Flush user-logged extra columns (from log_extra), gathering across processes. # Keys must be sorted so that all ranks call gather_object in the same order, otherwise values # get mis-attributed across columns (dict insertion order may differ between processes). for column in sorted(self._pending_extra_logs): self._logs["extra"][column].extend(gather_object(self._pending_extra_logs[column])) self._pending_extra_logs.clear() # Flush user-logged metrics (from log_metric), averaging across processes. # Keys must be sorted so that all ranks call accelerator.gather in the same order, otherwise values # get mis-attributed across metrics (dict insertion order may differ between processes). for name in sorted(self._pending_metrics): values = self._pending_metrics[name] local_mean = sum(values) / len(values) global_mean = self.accelerator.gather(torch.tensor(local_mean, device=device)).mean().item() self._metrics[mode][name].append(global_mean) self._pending_metrics.clear() if images is not None: self._logs["images"].extend(gather_object(images)) if self.use_vllm and self.vllm_importance_sampling_correction: delta = torch.abs(old_per_token_logps - sampling_per_token_logps) mask = completion_mask.bool() if tool_mask is None else (completion_mask * tool_mask).bool() delta = delta[mask] mean_delta = torch.mean(delta) if delta.numel() > 0 else torch.tensor(0.0, device=device) max_delta = torch.max(delta) if delta.numel() > 0 else torch.tensor(0.0, device=device) self._metrics[mode]["sampling/sampling_logp_difference/mean"].append( self.accelerator.gather(mean_delta).mean().item() ) self._metrics[mode]["sampling/sampling_logp_difference/max"].append( self.accelerator.gather(max_delta).max().item() ) if sequence_level_is: flat_is_ratio = vllm_importance_sampling_ratio.flatten() else: flat_is_ratio = vllm_importance_sampling_ratio[mask] min_importance_sampling_ratio = ( torch.min(flat_is_ratio) if flat_is_ratio.numel() > 0 else torch.tensor(0.0, device=device) ) mean_importance_sampling_ratio = ( torch.mean(flat_is_ratio) if flat_is_ratio.numel() > 0 else torch.tensor(0.0, device=device) ) max_importance_sampling_ratio = ( torch.max(flat_is_ratio) if flat_is_ratio.numel() > 0 else torch.tensor(0.0, device=device) ) self._metrics[mode]["sampling/importance_sampling_ratio/min"].append( nanmin(self.accelerator.gather(min_importance_sampling_ratio)).item() ) self._metrics[mode]["sampling/importance_sampling_ratio/mean"].append( self.accelerator.gather(mean_importance_sampling_ratio).nanmean().item() ) self._metrics[mode]["sampling/importance_sampling_ratio/max"].append( nanmax(self.accelerator.gather(max_importance_sampling_ratio)).item() ) output = { "prompt_ids": prompt_ids, "prompt_mask": prompt_mask, "completion_ids": completion_ids, "completion_mask": completion_mask, "advantages": advantages, "num_items_in_batch": num_items_in_batch, } if old_per_token_logps is not None: output["old_per_token_logps"] = old_per_token_logps if self.use_vllm and self.vllm_importance_sampling_correction: output["importance_sampling_ratio"] = vllm_importance_sampling_ratio if sampling_per_token_logps is not None: output["sampling_per_token_logps"] = sampling_per_token_logps if ref_per_token_logps is not None: output["ref_per_token_logps"] = ref_per_token_logps if "pixel_values" in forward_kwargs: output["pixel_values"] = forward_kwargs["pixel_values"] if "image_grid_thw" in forward_kwargs: output["image_grid_thw"] = forward_kwargs["image_grid_thw"] if "pixel_attention_mask" in forward_kwargs: output["pixel_attention_mask"] = forward_kwargs["pixel_attention_mask"] if "image_sizes" in forward_kwargs: output["image_sizes"] = forward_kwargs["image_sizes"] if "token_type_ids" in forward_kwargs: output["token_type_ids"] = forward_kwargs["token_type_ids"] if "mm_token_type_ids" in forward_kwargs: output["mm_token_type_ids"] = forward_kwargs["mm_token_type_ids"] if "pixel_position_ids" in forward_kwargs: output["pixel_position_ids"] = forward_kwargs["pixel_position_ids"] if images is not None: output["num_images"] = num_images if tool_mask is not None: output["tool_mask"] = tool_mask return output def compute_liger_loss(self, unwrapped_model, inputs): # Compute the per-token log probabilities for the model prompt_ids, prompt_mask = inputs["prompt_ids"], inputs["prompt_mask"] completion_ids, completion_mask = inputs["completion_ids"], inputs["completion_mask"] input_ids = torch.cat([prompt_ids, completion_ids], dim=1) attention_mask = torch.cat([prompt_mask, completion_mask], dim=1) logits_to_keep = completion_ids.size(1) # we only need to compute the logits for the completion tokens # Get the last hidden state of the model last_hidden_state = self._get_last_hidden_state( unwrapped_model, input_ids, attention_mask, logits_to_keep, inputs.get("pixel_values"), inputs.get("image_grid_thw"), inputs.get("pixel_attention_mask"), inputs.get("image_sizes"), inputs.get("pixel_position_ids"), ) # Apply tool_mask (from env_mask) for loss computation in multi-turn training scenarios loss_mask = completion_mask if "tool_mask" not in inputs else completion_mask * inputs["tool_mask"] # Compute loss and metrics using liger grpo loss loss, metrics = self.liger_grpo_loss( _input=last_hidden_state, lin_weight=unwrapped_model.lm_head.weight, selected_token_ids=completion_ids, # The attention_mask parameter in liger loss is actually used as a loss mask (not model attention) attention_mask=loss_mask, advantages=inputs["advantages"], bias=unwrapped_model.lm_head.bias, old_per_token_logps=inputs.get("old_per_token_logps"), ref_per_token_logps=inputs.get("ref_per_token_logps"), vllm_is_ratio=inputs.get("importance_sampling_ratio"), ) # Extract metrics from the liger_grpo_loss output # KL divergence is the first metric when beta is non-zero mean_kl = metrics[0] if self.beta != 0.0 else None clip_ratio = metrics[-1] mode = "train" if self.model.training else "eval" if self.beta != 0.0: self._metrics[mode]["kl"].append(self.accelerator.gather(mean_kl).mean().item()) self._metrics[mode]["clip_ratio"].append(self.accelerator.gather(clip_ratio).mean().item()) normalizer = self.current_gradient_accumulation_steps if mode == "train" else 1.0 # no accum in eval return loss / normalizer @profiling_decorator def compute_loss(self, model, inputs, return_outputs=False, num_items_in_batch=None): if return_outputs: raise ValueError("The GRPOTrainer does not support returning outputs") if self.use_liger_kernel: # Compute the loss using the liger grpo loss unwrapped_model = self.accelerator.unwrap_model(model) return self._forward_redirection(model, unwrapped_model, self.compute_liger_loss, unwrapped_model, inputs) else: return self._compute_loss(model, inputs) @staticmethod def get_off_policy_mask( advantages: torch.Tensor, per_token_logps: torch.Tensor, sampling_per_token_logps: torch.Tensor, mask: torch.Tensor, off_policy_threshold: float, ) -> torch.Tensor: """ Computes the Off-Policy Sequence Mask from DeepSeek-V3.2 paper. Returns a (B, 1) tensor where 1.0 indicates "Keep" and 0.0 indicates "Drop". """ # forward KL div: log(pi_old) - log(pi_theta) kl_div = sampling_per_token_logps - per_token_logps.detach() # Sequence-level Mean KL (ignoring prompt+padding) seq_kl_sum = (kl_div * mask).sum(dim=1, keepdim=True) avg_seq_kl = seq_kl_sum / mask.sum(dim=1, keepdim=True).clamp(min=1.0) # Keep if (Advantage >= 0) OR (KL <= delta) is_pos_adv = advantages >= 0 is_low_kl = avg_seq_kl <= off_policy_threshold return (is_pos_adv | is_low_kl).to(dtype=mask.dtype) # (B, 1) @staticmethod @torch.no_grad() def get_gamma_weights( advantages: torch.Tensor, log_ratio_per_token: torch.Tensor, mask: torch.Tensor, importance_sampling_ratio: torch.Tensor | None, # (B, T) k_pos: float = 2.0, lambda_pos: float = 3.0, k_neg: float = 3.0, lambda_neg: float = 2.0, ) -> torch.Tensor: """ Computes the Gamma weights for the VESPO loss. For reference: φ(w) = e^λ × w^k × e^{-λw} is the gamma weighting (normalized so φ(1)=1) with w = sequence-level importance sampling ratio note: we will compute φ(w) in log space φ(w) is detached via @torch.no_grad(), only acts as gradient scaling coefficient VESPO loss = -φ(w) × A × log_prob, gradient naturally gives φ(w) × A × ∇log π """ # reducing clamp range directly to log(1e-8) ~ -18.42, to avoid recomputing log_w=log(w.clamp(min=1e-8)) later # This is solely for matching truthfully the original implementation, otherwise keeping -20 could be fine. lower_clamp = math.log(1e-8) # Sequence-level log ratio Σ log(π_θ/π_old) (not a mean like for `log_importance_weights`) log_ratio_clamped = torch.clamp(log_ratio_per_token, -20.0, 20.0) seq_log_ratio = torch.sum(log_ratio_clamped * mask, dim=-1, keepdim=True) # (B, 1) # Apply token-level TIS or MIS correction (in log space) if importance_sampling_ratio is not None: log_is_ratio = torch.clamp(torch.log(importance_sampling_ratio), lower_clamp, 20.0) # log(w) = log(π_θ/π_old) + log(π_old/π_sampler) seq_log_ratio += torch.sum(log_is_ratio, dim=-1, keepdim=True) log_w_seq = torch.clamp(seq_log_ratio, lower_clamp, 20.0) w_seq = torch.exp(log_w_seq) # compute k and lambda based on advantage sign is_nonneg_adv = advantages >= 0 k_seq = torch.where(is_nonneg_adv, k_pos, k_neg) lambda_seq = torch.where(is_nonneg_adv, lambda_pos, lambda_neg).clamp(min=1e-4) # log(φ(w)) = λ + k × log(w) - λ × w log_phi = lambda_seq + k_seq * log_w_seq - lambda_seq * w_seq phi_seq = torch.exp(log_phi).nan_to_num(nan=0.0, posinf=0.0, neginf=0.0) return phi_seq # (B, 1) def _compute_loss(self, model, inputs): # Compute the per-token log probabilities for the model prompt_ids, prompt_mask = inputs["prompt_ids"], inputs["prompt_mask"] completion_ids, completion_mask = inputs["completion_ids"], inputs["completion_mask"] input_ids = torch.cat([prompt_ids, completion_ids], dim=1) attention_mask = torch.cat([prompt_mask, completion_mask], dim=1) logits_to_keep = completion_ids.size(1) # we only need to compute the logits for the completion tokens mask = completion_mask if "tool_mask" not in inputs else completion_mask * inputs["tool_mask"] # Compute the per_token_logps and the entropy at each position in the completion per_token_logps, entropies = self._get_per_token_logps_and_entropies( model, input_ids, attention_mask, logits_to_keep, compute_entropy=True, pixel_values=inputs.get("pixel_values"), image_grid_thw=inputs.get("image_grid_thw"), num_images=inputs.get("num_images"), pixel_attention_mask=inputs.get("pixel_attention_mask"), image_sizes=inputs.get("image_sizes"), token_type_ids=inputs.get("token_type_ids"), mm_token_type_ids=inputs.get("mm_token_type_ids"), pixel_position_ids=inputs.get("pixel_position_ids"), ) if self.top_entropy_quantile < 1.0: entropy_mask = self.get_high_entropy_mask(entropies, mask, 1 - self.top_entropy_quantile) else: entropy_mask = None # Compute the loss advantages = inputs["advantages"] # In the base GRPO implementation, advantages are expected to have shape (B,). To support subclasses that # provide advantages with shape (B, T) (e.g., MiniLLM), we *conditionally* unsqueeze the tensor. if advantages.dim() == 1: advantages = advantages.unsqueeze(1) # When num_iterations == 1 and steps_per_generation <= gradient_accumulation_steps, # old_per_token_logps == per_token_logps. In this case we can skip its computation # (see _generate_and_score_completions) and instead use per_token_logps.detach(). # The exception is when using vLLM, where we always compute old_per_token_logps # for importance sampling old_per_token_logps = inputs.get("old_per_token_logps") old_per_token_logps = per_token_logps.detach() if old_per_token_logps is None else old_per_token_logps if self.off_policy_mask_threshold is not None: # OPSM should use inference-time logprobs to detect both sources of off-policyness: # 1. Drift from gradient updates (always present) # 2. Drift from training-inference mismatch (when using vLLM) # When using vLLM, prioritize sampling_per_token_logps, otherwise use old_per_token_logps sampling_per_token_logps = inputs.get("sampling_per_token_logps", old_per_token_logps) off_policy_mask = self.get_off_policy_mask( advantages=advantages, per_token_logps=per_token_logps, sampling_per_token_logps=sampling_per_token_logps, mask=mask, off_policy_threshold=self.off_policy_mask_threshold, ) log_ratio = per_token_logps - old_per_token_logps if self.importance_sampling_level == "token": log_importance_weights = log_ratio elif self.importance_sampling_level == "sequence": log_importance_weights = (log_ratio * mask).sum(-1) / mask.sum(-1).clamp(min=1.0) log_importance_weights = log_importance_weights.unsqueeze(-1) else: raise ValueError( f"Unknown importance sampling level: {self.importance_sampling_level}. Possible values are 'token' " "and 'sequence'." ) coef_1 = torch.exp(log_importance_weights) # Compute the KL divergence between the model and the reference model if self.beta != 0.0: ref_per_token_logps = inputs["ref_per_token_logps"] per_token_kl = ( torch.exp(ref_per_token_logps - per_token_logps) - (ref_per_token_logps - per_token_logps) - 1 ) # Importance sampling correction for the KL divergence if self.args.use_bias_correction_kl: per_token_kl = per_token_kl * coef_1 # From here, log_importance_weights (and all subsequent tensors, coef_1, coef_2, etc.) shape depends on # importance_sampling_level: "token" level: (B, T); "sequence" level: (B, 1) if self.loss_type == "cispo": clamped_ratios = torch.clamp(coef_1, max=self.epsilon_high).detach() per_token_loss = -clamped_ratios * advantages * per_token_logps elif self.loss_type in ["grpo", "bnpo", "dr_grpo", "dapo", "luspo"]: coef_2 = torch.clamp(coef_1, 1 - self.epsilon_low, 1 + self.epsilon_high) # Two-sided clipping if self.args.delta is not None: coef_1 = torch.clamp(coef_1, max=self.args.delta) per_token_loss1 = coef_1 * advantages per_token_loss2 = coef_2 * advantages per_token_loss = -torch.min(per_token_loss1, per_token_loss2) elif self.loss_type == "sapo": temperatures = torch.where(advantages > 0, self.args.sapo_temperature_pos, self.args.sapo_temperature_neg) soft_coef_1 = torch.sigmoid(temperatures * (coef_1 - 1)) * 4 / temperatures per_token_loss = -soft_coef_1 * advantages elif self.loss_type == "vespo": phi_seq = self.get_gamma_weights( advantages=advantages, log_ratio_per_token=log_ratio, mask=mask, importance_sampling_ratio=inputs.get("importance_sampling_ratio"), k_pos=self.args.vespo_k_pos, lambda_pos=self.args.vespo_lambda_pos, k_neg=self.args.vespo_k_neg, lambda_neg=self.args.vespo_lambda_neg, ) per_token_loss = -phi_seq * advantages * per_token_logps else: raise ValueError(f"Unknown loss type: {self.loss_type}") if self.off_policy_mask_threshold is not None: per_token_loss = per_token_loss * off_policy_mask if entropy_mask is not None: per_token_loss = per_token_loss * entropy_mask if self.use_vllm and self.vllm_importance_sampling_correction and self.loss_type != "vespo": per_token_loss = per_token_loss * inputs["importance_sampling_ratio"] if self.beta != 0.0: per_token_loss = per_token_loss + self.beta * per_token_kl mode = "train" if self.model.training else "eval" if self.loss_type in ["grpo", "sapo"]: loss = ((per_token_loss * mask).sum(-1) / mask.sum(-1).clamp(min=1.0)).mean() normalizer = self.current_gradient_accumulation_steps if mode == "train" else 1.0 # no accum in eval loss = loss / normalizer elif self.loss_type == "bnpo": loss = (per_token_loss * mask).sum() / mask.sum().clamp(min=1.0) normalizer = self.current_gradient_accumulation_steps if mode == "train" else 1.0 # no accum in eval loss = loss / normalizer elif self.loss_type == "dr_grpo": loss = (per_token_loss * mask).sum() / (per_token_loss.size(0) * self.max_completion_length) normalizer = self.current_gradient_accumulation_steps if mode == "train" else 1.0 # no accum in eval loss = loss / normalizer elif self.loss_type in ["cispo", "dapo", "vespo"]: normalizer = inputs["num_items_in_batch"] / self.accelerator.num_processes loss = (per_token_loss * mask).sum() / normalizer elif self.loss_type == "luspo": # Unless importance_sampling_level="token" (not recommended here), per_token_loss is expected to be (B, 1) loss = (per_token_loss * mask.sum(1, keepdim=True)).mean() normalizer = self.current_gradient_accumulation_steps if mode == "train" else 1.0 loss = loss / normalizer else: raise ValueError(f"Unknown loss type: {self.loss_type}") # Log the metrics completion_token_count = mask.sum().clamp(min=1.0) def masked_batch_mean(x): if x.shape[1] == 1: # when importance_sampling_level == "sequence" return x.mean() else: return (x * mask).sum() / completion_token_count if self.beta != 0.0: mean_kl = masked_batch_mean(per_token_kl) self._metrics[mode]["kl"].append(self.accelerator.gather(mean_kl).nanmean().item()) mean_entropy = masked_batch_mean(entropies) self._metrics[mode]["entropy"].append(self.accelerator.gather(mean_entropy).nanmean().item()) if self.loss_type in ["grpo", "bnpo", "dr_grpo", "dapo", "luspo"]: # Compute the clipped probability ratios is_low_clipped = (coef_1 < 1 - self.epsilon_low) & (advantages < 0) is_high_clipped = (coef_1 > 1 + self.epsilon_high) & (advantages > 0) is_region_clipped = is_low_clipped | is_high_clipped low_clip = masked_batch_mean(is_low_clipped.float()) high_clip = masked_batch_mean(is_high_clipped.float()) clip_ratio = masked_batch_mean(is_region_clipped.float()) gathered_low_clip = self.accelerator.gather(low_clip) self._metrics[mode]["clip_ratio/low_mean"].append(gathered_low_clip.nanmean().item()) self._metrics[mode]["clip_ratio/low_min"].append(nanmin(gathered_low_clip).item()) gathered_high_clip = self.accelerator.gather(high_clip) self._metrics[mode]["clip_ratio/high_mean"].append(gathered_high_clip.nanmean().item()) self._metrics[mode]["clip_ratio/high_max"].append(nanmax(gathered_high_clip).item()) gathered_clip_ratio = self.accelerator.gather(clip_ratio) self._metrics[mode]["clip_ratio/region_mean"].append(gathered_clip_ratio.nanmean().item()) elif self.loss_type == "cispo": is_cispo_clipped = (coef_1 > self.epsilon_high) & (advantages > 0) cispo_clip_ratio = masked_batch_mean(is_cispo_clipped.float()) gathered_cispo_clip_ratio = self.accelerator.gather(cispo_clip_ratio) self._metrics[mode]["cispo_clip_ratio"].append(gathered_cispo_clip_ratio.nanmean().item()) elif self.loss_type == "vespo": gathered_phi_seq = self.accelerator.gather(phi_seq) self._metrics[mode]["vespo/phi_seq_mean"].append(gathered_phi_seq.nanmean().item()) return loss # During eval, Trainer calls prediction_step. If no labels are present in the inputs, it only runs forward and # returns logits. We override prediction_step to force compute_loss, because this trainer doesn't involve labels. def prediction_step(self, model, inputs, prediction_loss_only, ignore_keys: list[str] | None = None): inputs = self._prepare_inputs(inputs) with torch.no_grad(): with self.compute_loss_context_manager(): loss = self.compute_loss(model, inputs) loss = loss.mean().detach() return loss, None, None def log(self, logs: dict[str, float], start_time: float | None = None) -> None: mode = "train" if self.model.training else "eval" # Average the metrics metrics = {} for key, val in self._metrics[mode].items(): # Filter out NaN values before averaging. A reward function that returns None for all samples # in a batch produces NaN for that batch's metric. With logging_steps > 1, a naive sum()/len() # would let a single NaN contaminate valid data from other batches. Only return None when no # valid values remain (e.g. JSON loggers crash on float NaN). valid = [v for v in val if not math.isnan(v)] metrics[key] = sum(valid) / len(valid) if valid else None # This method can be called both in training and evaluation. When called in evaluation, the keys in `logs` # start with "eval_". We need to add the prefix "eval_" to the keys in `metrics` to match the format. if mode == "eval": metrics = {f"eval_{key}": val for key, val in metrics.items()} logs = {**logs, **metrics} super().log(logs, start_time) self._metrics[mode].clear() if self.accelerator.is_main_process and self.log_completions: if is_rich_available(): print_prompt_completions_sample( self._logs["prompt"], self._logs["completion"], self._logs["rewards"], self._logs["advantages"], self.state.global_step, self.num_completions_to_print, ) logging_backends = [] if self.args.report_to and "wandb" in self.args.report_to and wandb.run is not None: logging_backends.append(wandb) if self.args.report_to and "trackio" in self.args.report_to: logging_backends.append(trackio) table = { "step": [self.state.global_step] * len(self._logs["prompt"]), "prompt": self._logs["prompt"], "completion": self._logs["completion"], **self._logs["rewards"], **self._logs["extra"], "advantage": self._logs["advantages"], } df_base = pd.DataFrame(table) df_base.to_parquet( os.path.join( self.args.output_dir, "completions", f"completions_{self.state.global_step:05d}.parquet", ) ) images_raw = self._logs["images"] or [] for logging_backend in logging_backends: if images_raw: images = [] for image_list in self._logs["images"]: images.append([logging_backend.Image(image) for image in image_list]) df = pd.concat( [df_base, pd.Series(images, name="image")], axis=1, copy=False, ) else: df = df_base if self.log_unique_prompts: df = df.drop_duplicates(subset=["prompt"]) logging_backend.log({"completions": logging_backend.Table(dataframe=df)}) # Ensure the model card is saved along with the checkpoint def _save_checkpoint(self, model, trial): if self.args.hub_model_id is None: model_name = Path(self.args.output_dir).name else: model_name = self.args.hub_model_id.split("/")[-1] self.create_model_card(model_name=model_name) super()._save_checkpoint(model, trial)