Upload custom_modeling.py with huggingface_hub

custom_modeling.py

@@ -0,0 +1,122 @@
+from transformers import PreTrainedModel, AutoModel, AutoConfig
+from transformers.modeling_outputs import TokenClassifierOutput
+import torch
+import torch.nn as nn
+from torchcrf import CRF
+from typing import Optional, Union, Tuple, List
+import os
+import json
+
+class BertCRFPreTrainedModel(PreTrainedModel):
+    """An abstract class to handle weights initialization and a simple interface for downloading and loading pretrained models."""
+    config_class = AutoConfig
+    base_model_prefix = "bert"
+
+    def _init_weights(self, module):
+        """Initialize the weights"""
+        if isinstance(module, nn.Linear):
+            module.weight.data.normal_(mean=0.0, std=self.config.initializer_range if hasattr(self.config, 'initializer_range') else 0.02)
+            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 hasattr(self.config, 'initializer_range') else 0.02)
+            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 BertCRFForTokenClassification(BertCRFPreTrainedModel):
+    def __init__(self, config):
+        super().__init__(config)
+        self.num_labels = config.num_labels
+        self.bert = AutoModel.from_config(config)
+        self.dropout = nn.Dropout(config.hidden_dropout_prob if hasattr(config, 'hidden_dropout_prob') else 0.1)
+        self.classifier = nn.Linear(config.hidden_size, config.num_labels)
+        
+        self.use_crf = config.use_crf if hasattr(config, 'use_crf') else False
+        if self.use_crf:
+            self.crf = CRF(num_tags=self.num_labels, batch_first=True)
+        else:
+            self.crf = None
+            self.loss_fn = nn.CrossEntropyLoss()
+
+        # Initialize weights and apply final processing
+        self.post_init()
+
+    def forward(
+        self,
+        input_ids: Optional[torch.Tensor] = None,
+        attention_mask: Optional[torch.Tensor] = None,
+        token_type_ids: Optional[torch.Tensor] = None,
+        position_ids: Optional[torch.Tensor] = None,
+        head_mask: Optional[torch.Tensor] = None,
+        inputs_embeds: Optional[torch.Tensor] = None,
+        labels: Optional[torch.Tensor] = None,
+        output_attentions: Optional[bool] = None,
+        output_hidden_states: Optional[bool] = None,
+        return_dict: Optional[bool] = None,
+    ) -> Union[Tuple[torch.Tensor], TokenClassifierOutput]:
+        return_dict = return_dict if return_dict is not None else self.config.use_return_dict
+
+        outputs = self.bert(
+            input_ids,
+            attention_mask=attention_mask,
+            token_type_ids=token_type_ids,
+            position_ids=position_ids,
+            head_mask=head_mask,
+            inputs_embeds=inputs_embeds,
+            output_attentions=output_attentions,
+            output_hidden_states=output_hidden_states,
+            return_dict=return_dict,
+        )
+
+        sequence_output = outputs[0]
+        sequence_output = self.dropout(sequence_output)
+        logits = self.classifier(sequence_output)
+
+        loss = None
+        if labels is not None:
+            if self.crf is not None:
+                mask = attention_mask.bool()
+                labels_mask = labels != -100
+                mask = mask & labels_mask
+                loss = -self.crf(logits, labels, mask=mask, reduction='mean')
+            else:
+                loss = self.loss_fn(logits.view(-1, self.num_labels), labels.view(-1))
+
+        if not return_dict:
+            output = (logits,) + outputs[2:]
+            return ((loss,) + output) if loss is not None else output
+
+        return TokenClassifierOutput(
+            loss=loss,
+            logits=logits,
+            hidden_states=outputs.hidden_states if output_hidden_states else None,
+            attentions=outputs.attentions if output_attentions else None,
+        )
+
+    def save_pretrained(self, save_directory: str, **kwargs):
+        """Save model with custom CRF layer"""
+        # Save the config
+        self.config.use_crf = self.use_crf
+        self.config.save_pretrained(save_directory)
+        
+        # Save the model weights
+        super().save_pretrained(save_directory, **kwargs)
+        
+        if self.crf is not None:
+            crf_path = os.path.join(save_directory, "crf.pt")
+            torch.save(self.crf.state_dict(), crf_path)
+
+    @classmethod
+    def from_pretrained(cls, pretrained_model_name_or_path: str, *model_args, **kwargs):
+        """Load model with custom CRF layer"""
+        model = super().from_pretrained(pretrained_model_name_or_path, *model_args, **kwargs)
+        
+        # Load CRF if it exists
+        crf_path = os.path.join(pretrained_model_name_or_path, "crf.pt")
+        if os.path.exists(crf_path) and model.use_crf:
+            model.crf.load_state_dict(torch.load(crf_path))
+        
+        return model