423A35C7/LLaMA-Factory
1
0
Fork 0
mirror of https://github.com/hiyouga/LLaMA-Factory.git synced 2025-08-23 22:32:54 +08:00
Code Issues Packages Projects Releases Wiki Activity

fix #6448

Browse Source 
Former-commit-id: 27198679829fb766c7eef468ae4311fdced695a2
This commit is contained in:
hiyouga 2024-12-27 16:54:39 +00:00
parent 377dfe5665
commit 3bcb4633ca
7 changed files with 29 additions and 27 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

1
src/llamafactory/cli.py
View File

@ -121,5 +121,6 @@ def main():
else: else:
raise NotImplementedError(f"Unknown command: {command}.") raise NotImplementedError(f"Unknown command: {command}.")
if __name__ == "__main__": if __name__ == "__main__":
main() main()

6
src/llamafactory/data/collator.py
View File

@ -105,7 +105,10 @@ class MultiModalDataCollatorForSeq2Seq(DataCollatorForSeq2Seq):
fake_messages = [{"role": "user", "content": IMAGE_PLACEHOLDER}] fake_messages = [{"role": "user", "content": IMAGE_PLACEHOLDER}]
fake_images = [Image.new("RGB", (64, 64), (255, 255, 255))] fake_images = [Image.new("RGB", (64, 64), (255, 255, 255))]
fake_messages = self.template.mm_plugin.process_messages(fake_messages, fake_images, [], self.processor) fake_messages = self.template.mm_plugin.process_messages(fake_messages, fake_images, [], self.processor)
fake_input_ids = self.processor.tokenizer.encode(fake_messages[0]["content"], add_special_tokens=False) fake_input_ids = self.tokenizer.encode(fake_messages[0]["content"], add_special_tokens=False)
fake_input_ids, _ = self.template.mm_plugin.process_token_ids(
fake_input_ids, None, fake_images, [], self.tokenizer, self.processor
)
if self.tokenizer.padding_side == "right": if self.tokenizer.padding_side == "right":
features[0]["input_ids"] = features[0]["input_ids"] + fake_input_ids features[0]["input_ids"] = features[0]["input_ids"] + fake_input_ids
features[0]["attention_mask"] = features[0]["attention_mask"] + [0] * len(fake_input_ids) features[0]["attention_mask"] = features[0]["attention_mask"] + [0] * len(fake_input_ids)
@ -116,6 +119,7 @@ class MultiModalDataCollatorForSeq2Seq(DataCollatorForSeq2Seq):
features[0]["labels"] = [IGNORE_INDEX] * len(fake_input_ids) + features[0]["labels"] features[0]["labels"] = [IGNORE_INDEX] * len(fake_input_ids) + features[0]["labels"]
batch_images = fake_images batch_images = fake_images
batch_imglens[0] = 1
batch_input_ids[0] = features[0]["input_ids"] batch_input_ids[0] = features[0]["input_ids"]
mm_inputs = self.template.mm_plugin.get_mm_inputs( mm_inputs = self.template.mm_plugin.get_mm_inputs(

10
src/llamafactory/train/dpo/trainer.py
View File

@ -79,6 +79,7 @@ class CustomDPOTrainer(DPOTrainer):
self.simpo_gamma = finetuning_args.simpo_gamma self.simpo_gamma = finetuning_args.simpo_gamma
Trainer.__init__(self, model=model, **kwargs) Trainer.__init__(self, model=model, **kwargs)
self.model_accepts_loss_kwargs = False # overwrite trainer's default behavior
if not hasattr(self, "accelerator"): if not hasattr(self, "accelerator"):
raise AttributeError("Please update `transformers`.") raise AttributeError("Please update `transformers`.")
@ -274,15 +275,14 @@ class CustomDPOTrainer(DPOTrainer):
self, model: "PreTrainedModel", inputs: Dict[str, "torch.Tensor"], return_outputs: bool = False, **kwargs self, model: "PreTrainedModel", inputs: Dict[str, "torch.Tensor"], return_outputs: bool = False, **kwargs
) -> Union["torch.Tensor", Tuple["torch.Tensor", List["torch.Tensor"]]]: ) -> Union["torch.Tensor", Tuple["torch.Tensor", List["torch.Tensor"]]]:
r""" r"""
Fixes the loss value for transformers 4.46.0. Fixes the loss value. See https://github.com/huggingface/transformers/pull/35438 for details.
https://github.com/huggingface/transformers/blob/v4.46.0/src/transformers/trainer.py#L3605
""" """
loss = super().compute_loss(model, inputs, return_outputs) loss = super().compute_loss(model, inputs, return_outputs)
if is_transformers_version_equal_to_4_46() and kwargs.pop("num_items_in_batch", False): if is_transformers_version_equal_to_4_46() and kwargs.get("num_items_in_batch"):
if return_outputs: if return_outputs:
return (loss[0] / self.args.gradient_accumulation_steps, *loss[1:]) loss = (loss[0] / self.args.gradient_accumulation_steps, *loss[1:])
else: else:
return loss / self.args.gradient_accumulation_steps loss = loss / self.args.gradient_accumulation_steps
return loss return loss

10
src/llamafactory/train/kto/trainer.py
View File

@ -77,6 +77,7 @@ class CustomKTOTrainer(KTOTrainer):
self.ftx_gamma = finetuning_args.pref_ftx self.ftx_gamma = finetuning_args.pref_ftx
Trainer.__init__(self, model=model, **kwargs) Trainer.__init__(self, model=model, **kwargs)
self.model_accepts_loss_kwargs = False # overwrite trainer's default behavior
if not hasattr(self, "accelerator"): if not hasattr(self, "accelerator"):
raise AttributeError("Please update `transformers`.") raise AttributeError("Please update `transformers`.")
@ -252,15 +253,14 @@ class CustomKTOTrainer(KTOTrainer):
self, model: "PreTrainedModel", inputs: Dict[str, "torch.Tensor"], return_outputs: bool = False, **kwargs self, model: "PreTrainedModel", inputs: Dict[str, "torch.Tensor"], return_outputs: bool = False, **kwargs
) -> Union["torch.Tensor", Tuple["torch.Tensor", List["torch.Tensor"]]]: ) -> Union["torch.Tensor", Tuple["torch.Tensor", List["torch.Tensor"]]]:
r""" r"""
Fixes the loss value for transformers 4.46.0. Fixes the loss value. See https://github.com/huggingface/transformers/pull/35438 for details.
https://github.com/huggingface/transformers/blob/v4.46.0/src/transformers/trainer.py#L3605
""" """
loss = super().compute_loss(model, inputs, return_outputs) loss = super().compute_loss(model, inputs, return_outputs)
if is_transformers_version_equal_to_4_46() and kwargs.pop("num_items_in_batch", False): if is_transformers_version_equal_to_4_46() and kwargs.get("num_items_in_batch"):
if return_outputs: if return_outputs:
return (loss[0] / self.args.gradient_accumulation_steps, *loss[1:]) loss = (loss[0] / self.args.gradient_accumulation_steps, *loss[1:])
else: else:
return loss / self.args.gradient_accumulation_steps loss = loss / self.args.gradient_accumulation_steps
return loss return loss

12
src/llamafactory/train/pt/trainer.py
View File

@ -19,7 +19,7 @@ import torch
from transformers import Trainer from transformers import Trainer
from typing_extensions import override from typing_extensions import override
from ...extras.packages import is_transformers_version_equal_to_4_46, is_transformers_version_greater_than from ...extras.packages import is_transformers_version_greater_than
from ..callbacks import SaveProcessorCallback from ..callbacks import SaveProcessorCallback
from ..trainer_utils import create_custom_optimizer, create_custom_scheduler from ..trainer_utils import create_custom_optimizer, create_custom_scheduler
@ -78,15 +78,13 @@ class CustomTrainer(Trainer):
self, model: "PreTrainedModel", inputs: Dict[str, "torch.Tensor"], return_outputs: bool = False, **kwargs self, model: "PreTrainedModel", inputs: Dict[str, "torch.Tensor"], return_outputs: bool = False, **kwargs
) -> Union["torch.Tensor", Tuple["torch.Tensor", List["torch.Tensor"]]]: ) -> Union["torch.Tensor", Tuple["torch.Tensor", List["torch.Tensor"]]]:
r""" r"""
Fixes the loss value for transformers 4.46.0. Fixes the loss value. See https://github.com/huggingface/transformers/pull/35438 for details.
https://github.com/huggingface/transformers/blob/v4.46.0/src/transformers/trainer.py#L3605
""" """
loss = super().compute_loss(model, inputs, return_outputs, **kwargs) loss = super().compute_loss(model, inputs, return_outputs, **kwargs)
if is_transformers_version_equal_to_4_46() and not getattr(self, "model_accepts_loss_kwargs", False): if kwargs.get("num_items_in_batch") and not getattr(self, "model_accepts_loss_kwargs", False):
# other model should not scale the loss
if return_outputs: if return_outputs:
return (loss[0] / self.args.gradient_accumulation_steps, *loss[1:]) loss = (loss[0] / self.args.gradient_accumulation_steps, *loss[1:])
else: else:
return loss / self.args.gradient_accumulation_steps loss = loss / self.args.gradient_accumulation_steps
return loss return loss

5
src/llamafactory/train/rm/trainer.py
View File

@ -52,6 +52,7 @@ class PairwiseTrainer(Trainer):
kwargs["processing_class"] = kwargs.pop("tokenizer") kwargs["processing_class"] = kwargs.pop("tokenizer")
super().__init__(**kwargs) super().__init__(**kwargs)
self.model_accepts_loss_kwargs = False # overwrite trainer's default behavior
self.finetuning_args = finetuning_args self.finetuning_args = finetuning_args
self.can_return_loss = True # override property to return eval_loss self.can_return_loss = True # override property to return eval_loss
self.add_callback(FixValueHeadModelCallback) self.add_callback(FixValueHeadModelCallback)
@ -107,8 +108,8 @@ class PairwiseTrainer(Trainer):
loss = -torch.nn.functional.logsigmoid(chosen_scores.float() - rejected_scores.float()).mean() loss = -torch.nn.functional.logsigmoid(chosen_scores.float() - rejected_scores.float()).mean()
if is_transformers_version_equal_to_4_46() and kwargs.pop("num_items_in_batch", False): if is_transformers_version_equal_to_4_46() and kwargs.get("num_items_in_batch"):
loss /= self.args.gradient_accumulation_steps # fixes the loss value for transformers 4.46.0 loss /= self.args.gradient_accumulation_steps # fixes the loss value for transformers 4.46.0-4.46.1
if return_outputs: if return_outputs:
return loss, (loss, chosen_scores, rejected_scores) return loss, (loss, chosen_scores, rejected_scores)

12
src/llamafactory/train/sft/trainer.py
View File

@ -27,7 +27,7 @@ from typing_extensions import override
from ...extras import logging from ...extras import logging
from ...extras.constants import IGNORE_INDEX from ...extras.constants import IGNORE_INDEX
from ...extras.packages import is_transformers_version_equal_to_4_46, is_transformers_version_greater_than from ...extras.packages import is_transformers_version_greater_than
from ..callbacks import SaveProcessorCallback from ..callbacks import SaveProcessorCallback
from ..trainer_utils import create_custom_optimizer, create_custom_scheduler from ..trainer_utils import create_custom_optimizer, create_custom_scheduler
@ -93,16 +93,14 @@ class CustomSeq2SeqTrainer(Seq2SeqTrainer):
self, model: "PreTrainedModel", inputs: Dict[str, "torch.Tensor"], return_outputs: bool = False, **kwargs self, model: "PreTrainedModel", inputs: Dict[str, "torch.Tensor"], return_outputs: bool = False, **kwargs
) -> Union["torch.Tensor", Tuple["torch.Tensor", List["torch.Tensor"]]]: ) -> Union["torch.Tensor", Tuple["torch.Tensor", List["torch.Tensor"]]]:
r""" r"""
Fixes the loss value for transformers 4.46.0. Fixes the loss value. See https://github.com/huggingface/transformers/pull/35438 for details.
https://github.com/huggingface/transformers/blob/v4.46.0/src/transformers/trainer.py#L3605
""" """
loss = super().compute_loss(model, inputs, return_outputs, **kwargs) loss = super().compute_loss(model, inputs, return_outputs, **kwargs)
if is_transformers_version_equal_to_4_46() and not getattr(self, "model_accepts_loss_kwargs", False): if kwargs.get("num_items_in_batch") and not getattr(self, "model_accepts_loss_kwargs", False):
# other model should not scale the loss
if return_outputs: if return_outputs:
return (loss[0] / self.args.gradient_accumulation_steps, *loss[1:]) loss = (loss[0] / self.args.gradient_accumulation_steps, *loss[1:])
else: else:
return loss / self.args.gradient_accumulation_steps loss = loss / self.args.gradient_accumulation_steps
return loss return loss