update patcher

src/llamafactory/model/model_utils/checkpointing.py

@@ -78,9 +78,7 @@ def _fp32_forward_post_hook(
     return output.to(torch.float32)
 
 
-def prepare_model_for_training(
-    model: "PreTrainedModel", model_args: "ModelArguments", output_layer_name: str = "lm_head"
-) -> None:
+def prepare_model_for_training(model: "PreTrainedModel", model_args: "ModelArguments") -> None:
     r"""
     Includes:
         (1) cast the layernorm in fp32
@@ -104,8 +102,8 @@ def prepare_model_for_training(
             setattr(model.config, "use_cache", False)  # turn off when gradient checkpointing is enabled
             logger.info("Gradient checkpointing enabled.")
 
-    if hasattr(model, output_layer_name) and model_args.upcast_lmhead_output:
-        logger.info("Upcasting lm_head outputs in float32.")
-        output_layer = getattr(model, output_layer_name)
+    if model_args.upcast_lmhead_output:
+        output_layer = model.get_output_embeddings()
         if isinstance(output_layer, torch.nn.Linear) and output_layer.weight.dtype != torch.float32:
+            logger.info("Upcasting lm_head outputs in float32.")
             output_layer.register_forward_hook(_fp32_forward_post_hook)

src/llamafactory/model/patcher.py

@@ -152,6 +152,10 @@ def patch_valuehead_model(model: "AutoModelForCausalLMWithValueHead") -> None:
         if isinstance(self.pretrained_model, PreTrainedModel):
             return self.pretrained_model.get_input_embeddings()
 
+    def get_output_embeddings(self: "AutoModelForCausalLMWithValueHead") -> torch.nn.Module:
+        if isinstance(self.pretrained_model, PreTrainedModel):
+            return self.pretrained_model.get_output_embeddings()
+
     def create_or_update_model_card(self: "AutoModelForCausalLMWithValueHead", output_dir: str) -> None:
         if isinstance(self.pretrained_model, PeftModel):
             self.pretrained_model.create_or_update_model_card(output_dir)
@@ -160,4 +164,5 @@ def patch_valuehead_model(model: "AutoModelForCausalLMWithValueHead") -> None:
     setattr(model, "_keys_to_ignore_on_save", ignore_modules)
     setattr(model, "tie_weights", MethodType(tie_weights, model))
     setattr(model, "get_input_embeddings", MethodType(get_input_embeddings, model))
+    setattr(model, "get_output_embeddings", MethodType(get_output_embeddings, model))
     setattr(model, "create_or_update_model_card", MethodType(create_or_update_model_card, model))

tests/model/model_utils/test_checkpointing.py

@@ -70,5 +70,5 @@ def test_upcast_lmhead_output():
     tokenizer_module = load_tokenizer(model_args)
     model = load_model(tokenizer_module["tokenizer"], model_args, finetuning_args, is_trainable=True)
     inputs = torch.randn((1, 16), dtype=torch.float16, device=get_current_device())
-    outputs: "torch.Tensor" = model.lm_head(inputs)
+    outputs: "torch.Tensor" = model.get_output_embeddings()(inputs)
     assert outputs.dtype == torch.float32