Update modeling_custom_seq2seq_llm.py

modeling_custom_seq2seq_llm.py

@@ -1228,33 +1228,33 @@ class CustomSeq2SeqLLM(PreTrainedModel):
             torch_filepath = os.path.join(save_directory, "pytorch_model.bin")
             torch.save(cpu_state_dict, torch_filepath)
 
-    @classmethod
-    def from_pretrained(cls, pretrained_model_name_or_path, *model_args, **kwargs):
-        config = kwargs.pop("config", None)
-        state_dict = kwargs.pop("state_dict", None)
-
-        if config is None:
-            config = cls.config_class.from_pretrained(pretrained_model_name_or_path, **kwargs)
-
-        model = cls(config)
-
-        if state_dict is None:
-            # Try loading safetensors first
-            safe_filepath = os.path.join(pretrained_model_name_or_path, "model.safetensors")
-            if os.path.exists(safe_filepath):
-                from safetensors.torch import load_file
-                state_dict = load_file(safe_filepath)
-            else:
-                # Fall back to PyTorch format
-                torch_filepath = os.path.join(pretrained_model_name_or_path, "pytorch_model.bin")
-                state_dict = torch.load(torch_filepath, map_location="cpu")
-
-        # Handle shared weights
-        if config.tie_word_embeddings and "lm_head.weight" not in state_dict:
-            state_dict["lm_head.weight"] = state_dict["shared.weight"]
-
-        model.load_state_dict(state_dict)
-        return model
+    # @classmethod
+    # def from_pretrained(cls, pretrained_model_name_or_path, *model_args, **kwargs):
+    #     config = kwargs.pop("config", None)
+    #     state_dict = kwargs.pop("state_dict", None)
+
+    #     if config is None:
+    #         config = cls.config_class.from_pretrained(pretrained_model_name_or_path, **kwargs)
+
+    #     model = cls(config)
+
+    #     if state_dict is None:
+    #         # Try loading safetensors first
+    #         safe_filepath = os.path.join(pretrained_model_name_or_path, "model.safetensors")
+    #         if os.path.exists(safe_filepath):
+    #             from safetensors.torch import load_file
+    #             state_dict = load_file(safe_filepath)
+    #         else:
+    #             # Fall back to PyTorch format
+    #             torch_filepath = os.path.join(pretrained_model_name_or_path, "pytorch_model.bin")
+    #             state_dict = torch.load(torch_filepath, map_location="cpu")
+
+    #     # Handle shared weights
+    #     if config.tie_word_embeddings and "lm_head.weight" not in state_dict:
+    #         state_dict["lm_head.weight"] = state_dict["shared.weight"]
+
+    #     model.load_state_dict(state_dict)
+    #     return model
 
 class CustomEncoder(nn.Module):
     def __init__(self, config):