modeling_codesage.py

#!/usr/bin/env python
# coding=utf-8
# Copyright Amazon.com, Inc. or its affiliates. All Rights Reserved.

import math
import torch
import torch.utils.checkpoint
from torch import nn
from torch.nn import CrossEntropyLoss, MSELoss, BCEWithLogitsLoss
from transformers.activations import ACT2FN
from transformers.modeling_utils import Conv1D, PreTrainedModel
from transformers.utils import logging
from .config_codesage import CodeSageConfig
from transformers.modeling_outputs import (
    BaseModelOutputWithPooling,
    MaskedLMOutput,
    SequenceClassifierOutput
)

logger = logging.get_logger(__name__)

CODESAGE_PRETRAINED_MODEL_ARCHIVE_LIST = [
    "codesage/codesage-small-v2",
    "codesage/codesage-base-v2",
    "codesage/codesage-large-v2",
    # See all CodeSage models at https://huggingface.co/models?filter=codesage
]


class CodeSageAttention(nn.Module):
    def __init__(self, config):
        super().__init__()

        self.hidden_size = config.hidden_size
        self.num_heads = config.num_attention_heads
        self.head_dim = config.hidden_size // self.num_heads
        if self.head_dim * self.num_heads != config.hidden_size:
            raise ValueError(
                f"`hidden_size` must be divisible by num_heads "
                f"(got `hidden_size`: {config.hidden_size} and `num_heads`: {self.num_heads})."
            )

        self.c_attn = Conv1D(3 * self.hidden_size, self.hidden_size)
        self.c_proj = Conv1D(self.hidden_size, self.hidden_size)

        self.attention_dropout = nn.Dropout(config.attention_dropout_prob)
        self.residual_dropout = nn.Dropout(config.residual_dropout_prob)

    def attn(self, query, key, value, attention_mask=None, head_mask=None):
        attn_weights = torch.matmul(query, key.transpose(-1, -2))
        attn_weights = attn_weights / math.sqrt(self.head_dim)
        if attention_mask is not None:
            attn_weights = attn_weights + attention_mask

        attn_weights = nn.Softmax(dim=-1)(attn_weights)
        attn_weights = self.attention_dropout(attn_weights)
        if head_mask is not None:
            attn_weights = attn_weights * head_mask

        attn_output = torch.matmul(attn_weights, value)
        return attn_output, attn_weights

    def split_heads(self, tensor, num_heads, attn_head_size):
        """
        Splits hidden_size dim into attn_head_size and num_heads
        """
        new_shape = tensor.size()[:-1] + (num_heads, attn_head_size)
        tensor = tensor.view(*new_shape)
        return tensor.permute(0, 2, 1, 3)  # (batch, head, seq_length, head_features)

    def merge_heads(self, tensor, num_heads, attn_head_size):
        """
        Merges attn_head_size dim and num_attn_heads dim into hidden_size
        """
        tensor = tensor.permute(0, 2, 1, 3).contiguous()
        new_shape = tensor.size()[:-2] + (num_heads * attn_head_size,)
        return tensor.view(new_shape)

    def forward(
            self,
            hidden_states,
            attention_mask=None,
            head_mask=None,
            output_attentions=False,
    ):
        query, key, value = self.c_attn(hidden_states).split(self.hidden_size, dim=2)
        query = self.split_heads(query, self.num_heads, self.head_dim)
        key = self.split_heads(key, self.num_heads, self.head_dim)
        value = self.split_heads(value, self.num_heads, self.head_dim)

        attn_output, attn_weights = self.attn(query, key, value, attention_mask, head_mask)

        attn_output = self.merge_heads(attn_output, self.num_heads, self.head_dim)
        attn_output = self.c_proj(attn_output)
        attn_output = self.residual_dropout(attn_output)

        outputs = (attn_output, attn_weights) if output_attentions else (attn_output,)
        return outputs  # a, present, (attentions)


class CodeSageMLP(nn.Module):
    def __init__(self, intermediate_size, config):
        super().__init__()

        self.c_fc = Conv1D(intermediate_size, config.hidden_size)
        self.act = ACT2FN[config.activation_function]
        self.c_proj = Conv1D(config.hidden_size, intermediate_size)
        self.dropout = nn.Dropout(config.residual_dropout_prob)

    def forward(self, hidden_states):
        hidden_states = self.c_fc(hidden_states)
        hidden_states = self.act(hidden_states)
        hidden_states = self.c_proj(hidden_states)
        hidden_states = self.dropout(hidden_states)
        return hidden_states


class CodeSageBlock(nn.Module):
    def __init__(self, config):
        super().__init__()
        hidden_size = config.hidden_size
        inner_dim = config.intermediate_size if config.intermediate_size is not None else 4 * hidden_size
        self.ln_1 = nn.LayerNorm(hidden_size, eps=config.layer_norm_epsilon)
        self.attn = CodeSageAttention(config)
        self.ln_2 = nn.LayerNorm(hidden_size, eps=config.layer_norm_epsilon)
        self.mlp = CodeSageMLP(inner_dim, config)

    def forward(
            self,
            hidden_states,
            attention_mask=None,
            head_mask=None,
            output_attentions=False,
    ):
        residual = hidden_states
        hidden_states = self.ln_1(hidden_states)
        attn_outputs = self.attn(
            hidden_states,
            attention_mask=attention_mask,
            head_mask=head_mask,
            output_attentions=output_attentions
        )
        attn_output = attn_outputs[0]  # output_attn: a, present, (attentions)
        outputs = attn_outputs[1:]
        hidden_states = attn_output + residual

        residual = hidden_states
        hidden_states = self.ln_2(hidden_states)
        feed_forward_hidden_states = self.mlp(hidden_states)
        hidden_states = residual + feed_forward_hidden_states

        outputs = (hidden_states,) + outputs[1:]
        return outputs  # hidden_states, present, (attentions)


class CodeSagePreTrainedModel(PreTrainedModel):
    config_class = CodeSageConfig
    base_model_prefix = "transformer"

    def _init_weights(self, module):
        """Initialize the weights."""
        if isinstance(module, (nn.Linear, Conv1D)):
            module.weight.data.normal_(mean=0.0, std=self.config.initializer_range)
            if module.bias is not None:
                module.bias.data.zero_()
        elif isinstance(module, nn.Embedding):
            module.weight.data.normal_(mean=0.0, std=self.config.initializer_range)
            if module.padding_idx is not None:
                module.weight.data[module.padding_idx].zero_()
        elif isinstance(module, nn.LayerNorm):
            module.bias.data.zero_()
            module.weight.data.fill_(1.0)


class CodeSageModel(CodeSagePreTrainedModel):
    def __init__(self, config):
        super().__init__(config)

        self.wte = nn.Embedding(config.vocab_size, config.hidden_size)
        self.wpe = nn.Embedding(config.max_position_embeddings, config.hidden_size)

        self.drop = nn.Dropout(config.embedding_dropout_prob)
        self.h = nn.ModuleList([CodeSageBlock(config) for _ in range(config.num_hidden_layers)])
        self.ln_f = nn.LayerNorm(config.hidden_size, eps=config.layer_norm_epsilon)

        self.init_weights()

    def get_input_embeddings(self):
        return self.wte

    def set_input_embeddings(self, new_embeddings: torch.Tensor):
        self.wte = new_embeddings

    def forward(
            self,
            input_ids=None,
            attention_mask=None,
            position_ids=None,
            head_mask=None,
            inputs_embeds=None,
            output_attentions=None,
            output_hidden_states=None,
            return_dict=None
    ):
        output_attentions = output_attentions if output_attentions is not None else self.config.output_attentions
        output_hidden_states = (
            output_hidden_states if output_hidden_states is not None else self.config.output_hidden_states
        )
        return_dict = return_dict if return_dict is not None else self.config.use_return_dict

        if input_ids is not None and inputs_embeds is not None:
            raise ValueError("You cannot specify both input_ids and inputs_embeds at the same time")
        if input_ids is not None:
            input_shape = input_ids.size()
        elif inputs_embeds is not None:
            input_shape = inputs_embeds.size()[:-1]
        else:
            raise ValueError("You have to specify either input_ids or inputs_embeds")

        device = input_ids.device if input_ids is not None else inputs_embeds.device
        if position_ids is None:
            position_ids = torch.arange(input_shape[-1], dtype=torch.long, device=device)
            position_ids = position_ids.unsqueeze(0).view(-1, input_shape[-1])
        else:
            position_ids = position_ids.view(-1, input_shape[-1])

        extended_attention_mask = None
        if attention_mask is not None:
            assert attention_mask.dim() == 2
            extended_attention_mask = attention_mask[:, None, None, :]
            extended_attention_mask = extended_attention_mask.to(dtype=self.dtype)  # fp16 compatibility
            extended_attention_mask = (1.0 - extended_attention_mask) * -10000.0

        head_mask = self.get_head_mask(head_mask, self.config.num_hidden_layers)
        if inputs_embeds is None:
            inputs_embeds = self.wte(input_ids)

        position_embeds = self.wpe(position_ids)
        hidden_states = inputs_embeds + position_embeds

        hidden_states = self.drop(hidden_states)
        output_shape = input_shape + (hidden_states.size(-1),)

        all_self_attentions = () if output_attentions else None
        all_hidden_states = () if output_hidden_states else None
        for i, block in enumerate(self.h):
            if output_hidden_states:
                all_hidden_states = all_hidden_states + (hidden_states,)

            outputs = block(
                hidden_states,
                attention_mask=extended_attention_mask,
                head_mask=head_mask[i],
                output_attentions=output_attentions,
            )

            hidden_states = outputs[0]
            if output_attentions:
                all_self_attentions = all_self_attentions + (outputs[1],)

        hidden_states = self.ln_f(hidden_states)
        hidden_states = hidden_states.view(*output_shape)
        if output_hidden_states:
            all_hidden_states = all_hidden_states + (hidden_states,)

        pooled_output = None  # max-pooled output
        if attention_mask is not None:
            pooled_output = (hidden_states * attention_mask[:, :, None]).sum(1) / attention_mask.sum(1)[:, None]

        if not return_dict:
            return tuple(
                v
                for v in [hidden_states, pooled_output, all_hidden_states, all_self_attentions]
                if v is not None
            )

        return BaseModelOutputWithPooling(
            last_hidden_state=hidden_states,
            pooler_output=pooled_output,
            hidden_states=all_hidden_states,
            attentions=all_self_attentions
        )


class CodeSageForMaskedLM(CodeSagePreTrainedModel):
    _tied_weights_keys = ["lm_head.weight"]

    def __init__(self, config):
        super().__init__(config)
        self.transformer = CodeSageModel(config)
        self.lm_head = nn.Linear(config.hidden_size, config.vocab_size, bias=False)

        self.init_weights()

    def get_output_embeddings(self):
        return self.lm_head

    def set_output_embeddings(self, new_embeddings):
        self.lm_head = new_embeddings

    def forward(
            self,
            input_ids=None,
            attention_mask=None,
            position_ids=None,
            head_mask=None,
            inputs_embeds=None,
            labels=None,
            output_attentions=None,
            output_hidden_states=None,
            return_dict=None
    ):
        return_dict = return_dict if return_dict is not None else self.config.use_return_dict

        transformer_outputs = self.transformer(
            input_ids,
            attention_mask=attention_mask,
            position_ids=position_ids,
            head_mask=head_mask,
            inputs_embeds=inputs_embeds,
            output_attentions=output_attentions,
            output_hidden_states=output_hidden_states,
            return_dict=return_dict
        )
        hidden_states = transformer_outputs[0]
        lm_logits = self.lm_head(hidden_states)

        masked_lm_loss = None
        if labels is not None:
            loss_fct = CrossEntropyLoss()
            masked_lm_loss = loss_fct(lm_logits.view(-1, lm_logits.size(-1)), labels.view(-1))

        if not return_dict:
            output = (lm_logits,) + transformer_outputs[1:]
            return ((masked_lm_loss,) + output) if masked_lm_loss is not None else output

        return MaskedLMOutput(
            loss=masked_lm_loss,
            logits=lm_logits,
            hidden_states=transformer_outputs.hidden_states,
            attentions=transformer_outputs.attentions,
        )


class CodeSageForSequenceClassification(CodeSagePreTrainedModel):

    def __init__(self, config):
        super().__init__(config)
        self.num_labels = config.num_labels
        self.config = config

        self.transformer = CodeSageModel(config)
        classifier_dropout = (
            config.classifier_dropout 
            if hasattr(config, 'classifier_dropout') and config.classifier_dropout is not None 
            else config.residual_dropout_prob
        )
        self.dropout = nn.Dropout(classifier_dropout)
        self.classifier = nn.Linear(config.hidden_size, config.num_labels)

        # Initialize weights and apply final processing
        self.post_init()

    def forward(
            self,
            input_ids=None,
            attention_mask=None,
            position_ids=None,
            head_mask=None,
            inputs_embeds=None,
            labels=None,
            output_attentions=None,
            output_hidden_states=None,
            return_dict=None,
    ):
        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
        assert attention_mask is not None, "attention_mask is needed to perform max-pooling"

        outputs = self.transformer(
            input_ids,
            attention_mask=attention_mask,
            position_ids=position_ids,
            head_mask=head_mask,
            inputs_embeds=inputs_embeds,
            output_attentions=output_attentions,
            output_hidden_states=output_hidden_states,
            return_dict=return_dict,
        )

        pooled_output = outputs[1]
        pooled_output = self.dropout(pooled_output)
        logits = self.classifier(pooled_output)

        loss = None
        if labels is not None:
            if self.config.problem_type is None:
                if self.num_labels == 1:
                    self.config.problem_type = "regression"
                elif self.num_labels > 1 and (labels.dtype == torch.long or labels.dtype == torch.int):
                    self.config.problem_type = "single_label_classification"
                else:
                    self.config.problem_type = "multi_label_classification"

            if self.config.problem_type == "regression":
                loss_fct = MSELoss()
                if self.num_labels == 1:
                    loss = loss_fct(logits.squeeze(), labels.squeeze())
                else:
                    loss = loss_fct(logits, labels)
            elif self.config.problem_type == "single_label_classification":
                loss_fct = CrossEntropyLoss()
                loss = loss_fct(logits.view(-1, self.num_labels), labels.view(-1))
            elif self.config.problem_type == "multi_label_classification":
                loss_fct = BCEWithLogitsLoss()
                loss = loss_fct(logits, labels)

        if not return_dict:
            output = (logits,) + outputs[2:]
            return ((loss,) + output) if loss is not None else output

        return SequenceClassifierOutput(
            loss=loss,
            logits=logits,
            hidden_states=outputs.hidden_states,
            attentions=outputs.attentions,
        )