diff --git a/README.md b/README.md index db981f9f..324940e3 100644 --- a/README.md +++ b/README.md @@ -5,7 +5,7 @@ [![GitHub contributors](https://img.shields.io/github/contributors/hiyouga/LLaMA-Factory?color=orange)](https://github.com/hiyouga/LLaMA-Factory/graphs/contributors) [![GitHub workflow](https://github.com/hiyouga/LLaMA-Factory/actions/workflows/tests.yml/badge.svg)](https://github.com/hiyouga/LLaMA-Factory/actions/workflows/tests.yml) [![PyPI](https://img.shields.io/pypi/v/llamafactory)](https://pypi.org/project/llamafactory/) -[![Citation](https://img.shields.io/badge/citation-561-green)](https://scholar.google.com/scholar?cites=12620864006390196564) +[![Citation](https://img.shields.io/badge/citation-614-green)](https://scholar.google.com/scholar?cites=12620864006390196564) [![Docker Pulls](https://img.shields.io/docker/pulls/hiyouga/llamafactory)](https://hub.docker.com/r/hiyouga/llamafactory/tags) [![Twitter](https://img.shields.io/twitter/follow/llamafactory_ai)](https://twitter.com/llamafactory_ai) @@ -55,7 +55,7 @@ Choose your path: - **Colab (free)**: https://colab.research.google.com/drive/1eRTPn37ltBbYsISy9Aw2NuI2Aq5CQrD9?usp=sharing - **Local machine**: Please refer to [usage](#getting-started) - **PAI-DSW (free trial)**: https://gallery.pai-ml.com/#/preview/deepLearning/nlp/llama_factory -- **Alaya NeW (cloud GPU deal)**: https://docs.alayanew.com/docs/documents/newActivities/llamafactory/?utm_source=LLaMA-Factory +- **Alaya NeW (cloud GPU deal)**: https://docs.alayanew.com/docs/documents/useGuide/LLaMAFactory/mutiple/?utm_source=LLaMA-Factory > [!NOTE] > Except for the above links, all other websites are unauthorized third-party websites. Please carefully use them. diff --git a/README_zh.md b/README_zh.md index 216d9b6c..85d3c823 100644 --- a/README_zh.md +++ b/README_zh.md @@ -5,7 +5,7 @@ [![GitHub contributors](https://img.shields.io/github/contributors/hiyouga/LLaMA-Factory?color=orange)](https://github.com/hiyouga/LLaMA-Factory/graphs/contributors) [![GitHub workflow](https://github.com/hiyouga/LLaMA-Factory/actions/workflows/tests.yml/badge.svg)](https://github.com/hiyouga/LLaMA-Factory/actions/workflows/tests.yml) [![PyPI](https://img.shields.io/pypi/v/llamafactory)](https://pypi.org/project/llamafactory/) -[![Citation](https://img.shields.io/badge/citation-561-green)](https://scholar.google.com/scholar?cites=12620864006390196564) +[![Citation](https://img.shields.io/badge/citation-614-green)](https://scholar.google.com/scholar?cites=12620864006390196564) [![Docker Pulls](https://img.shields.io/docker/pulls/hiyouga/llamafactory)](https://hub.docker.com/r/hiyouga/llamafactory/tags) [![Twitter](https://img.shields.io/twitter/follow/llamafactory_ai)](https://twitter.com/llamafactory_ai) @@ -41,7 +41,7 @@ -👋 加入我们的[微信群](assets/wechat.jpg)、[NPU 用户群](assets/wechat_npu.jpg)或 [Alaya NeW 算力优惠群](assets/wechat_alaya.png)。 +👋 加入我们的[微信群](assets/wechat.jpg)、[NPU 用户群](assets/wechat_npu.jpg)或 [九章智算云算力优惠群](assets/wechat_alaya.png)。 \[ [English](README.md) | 中文 \] @@ -57,7 +57,7 @@ https://github.com/user-attachments/assets/43b700c6-a178-41db-b1f8-8190a5d3fcfc - **Colab（免费）**：https://colab.research.google.com/drive/1d5KQtbemerlSDSxZIfAaWXhKr30QypiK?usp=sharing - **本地机器**：请见[如何使用](#如何使用) - **PAI-DSW（免费试用）**：https://gallery.pai-ml.com/#/preview/deepLearning/nlp/llama_factory -- **Alaya NeW（算力优惠活动）**：https://docs.alayanew.com/docs/documents/newActivities/llamafactory/?utm_source=LLaMA-Factory +- **九章智算云（算力优惠活动）**：https://docs.alayanew.com/docs/documents/useGuide/LLaMAFactory/mutiple/?utm_source=LLaMA-Factory > [!NOTE] > 除上述链接以外的其他网站均为未经许可的第三方网站，请小心甄别。 diff --git a/src/llamafactory/hparams/finetuning_args.py b/src/llamafactory/hparams/finetuning_args.py index cd403e8c..43596864 100644 --- a/src/llamafactory/hparams/finetuning_args.py +++ b/src/llamafactory/hparams/finetuning_args.py @@ -204,7 +204,12 @@ class RLHFArguments: ) ld_alpha: Optional[float] = field( default=None, - metadata={"help": "α parameter from the LD-DPO paper, which controls the weighting of the verbose token log-probabilities in responses"}, + metadata={ + "help": ( + "Alpha parameter from the LD-DPO paper, which controls the weighting of" + " the verbose token log-probabilities in responses." + ) + }, ) diff --git a/src/llamafactory/train/dpo/trainer.py b/src/llamafactory/train/dpo/trainer.py index 44587264..63822e88 100644 --- a/src/llamafactory/train/dpo/trainer.py +++ b/src/llamafactory/train/dpo/trainer.py @@ -188,8 +188,9 @@ class CustomDPOTrainer(DPOTrainer): batch = nested_detach(batch, clone=True) # avoid error all_logits: torch.Tensor = model(**batch, return_dict=True, use_cache=False).logits.to(torch.float32) - all_logps, valid_length = get_batch_logps(logits=all_logits, labels=batch["labels"], - ld_alpha=(self.ld_alpha if not is_ref_model else None)) + all_logps, valid_length = get_batch_logps( + logits=all_logits, labels=batch["labels"], ld_alpha=(self.ld_alpha if not is_ref_model else None) + ) if self.loss_type in ["ipo", "orpo", "simpo"]: all_logps = all_logps / valid_length @@ -219,8 +220,9 @@ class CustomDPOTrainer(DPOTrainer): ref_context = nullcontext() with torch.no_grad(), ref_context: - reference_chosen_logps, reference_rejected_logps, *_ = self.concatenated_forward(ref_model, batch, - is_ref_model=True) + reference_chosen_logps, reference_rejected_logps, *_ = self.concatenated_forward( + ref_model, batch, is_ref_model=True + ) return reference_chosen_logps, reference_rejected_logps diff --git a/src/llamafactory/train/trainer_utils.py b/src/llamafactory/train/trainer_utils.py index 688c7035..7d35bbeb 100644 --- a/src/llamafactory/train/trainer_utils.py +++ b/src/llamafactory/train/trainer_utils.py @@ -585,7 +585,10 @@ def create_custom_scheduler( def get_batch_logps( - logits: "torch.Tensor", labels: "torch.Tensor", label_pad_token_id: int = IGNORE_INDEX, ld_alpha: Optional[float] = None + logits: "torch.Tensor", + labels: "torch.Tensor", + label_pad_token_id: int = IGNORE_INDEX, + ld_alpha: Optional[float] = None, ) -> tuple["torch.Tensor", "torch.Tensor"]: r"""Compute the log probabilities of the given labels under the given logits.