update liger_kernel docs

README.md

@@ -51,7 +51,7 @@ You can contact us and communicate with us by adding our group:
 
 
 ## 📝 Introduction
-🍲 ms-swift is an official framework provided by the ModelScope community for fine-tuning and deploying large language models and multi-modal large models. It currently supports the training (pre-training, fine-tuning, human alignment), inference, evaluation, quantization, and deployment of 500+ large models and 200+ multi-modal large models. These large language models (LLMs) include models such as Qwen3, Qwen3-MoE, Qwen2.5, InternLM3, GLM4, Mistral, DeepSeek-R1, Yi1.5, TeleChat2, Baichuan2, and Gemma2. The multi-modal LLMs include models such as Qwen2.5-VL, Qwen2-Audio, Llama3.4, Llava, InternVL2.5, MiniCPM-V-2.6, GLM4v, Xcomposer2.5, Yi-VL, DeepSeek-VL2, Phi3.5-Vision, and GOT-OCR2.
+🍲 ms-swift is an official framework provided by the ModelScope community for fine-tuning and deploying large language models and multi-modal large models. It currently supports the training (pre-training, fine-tuning, human alignment), inference, evaluation, quantization, and deployment of 500+ large models and 200+ multi-modal large models. These large language models (LLMs) include models such as Qwen3, Qwen3-MoE, Qwen2.5, InternLM3, GLM4, Mistral, DeepSeek-R1, Yi1.5, TeleChat2, Baichuan2, and Gemma2. The multi-modal LLMs include models such as Qwen2.5-VL, Qwen2-Audio, Llama3.4, Llava, InternVL3, MiniCPM-V-2.6, GLM4v, Xcomposer2.5, Yi-VL, DeepSeek-VL2, Phi3.5-Vision, and GOT-OCR2.
 
 🍔 Additionally, ms-swift incorporates the latest training technologies, including lightweight techniques such as LoRA, QLoRA, Llama-Pro, LongLoRA, GaLore, Q-GaLore, LoRA+, LISA, DoRA, FourierFt, ReFT, UnSloth, and Liger, as well as human alignment training methods like DPO, GRPO, RM, PPO, KTO, CPO, SimPO, and ORPO. ms-swift supports acceleration of inference, evaluation, and deployment modules using vLLM and LMDeploy, and it supports model quantization with technologies like GPTQ, AWQ, and BNB. Furthermore, ms-swift offers a Gradio-based Web UI and a wealth of best practices.
 

README_CN.md

@@ -49,7 +49,7 @@
 <img src="asset/discord_qr.jpg" width="200" height="200">  |  <img src="asset/wechat.png" width="200" height="200">
 
 ## 📝 简介
-🍲 ms-swift是魔搭社区提供的大模型与多模态大模型微调部署框架，现已支持500+大模型与200+多模态大模型的训练（预训练、微调、人类对齐）、推理、评测、量化与部署。其中大模型包括：Qwen3、Qwen3-MoE、Qwen2.5、InternLM3、GLM4、Mistral、DeepSeek-R1、Yi1.5、TeleChat2、Baichuan2、Gemma2等模型，多模态大模型包括：Qwen2.5-VL、Qwen2-Audio、Llama4、Llava、InternVL2.5、MiniCPM-V-2.6、GLM4v、Xcomposer2.5、Yi-VL、DeepSeek-VL2、Phi3.5-Vision、GOT-OCR2等模型。
+🍲 ms-swift是魔搭社区提供的大模型与多模态大模型微调部署框架，现已支持500+大模型与200+多模态大模型的训练（预训练、微调、人类对齐）、推理、评测、量化与部署。其中大模型包括：Qwen3、Qwen3-MoE、Qwen2.5、InternLM3、GLM4、Mistral、DeepSeek-R1、Yi1.5、TeleChat2、Baichuan2、Gemma2等模型，多模态大模型包括：Qwen2.5-VL、Qwen2-Audio、Llama4、Llava、InternVL3、MiniCPM-V-2.6、GLM4v、Xcomposer2.5、Yi-VL、DeepSeek-VL2、Phi3.5-Vision、GOT-OCR2等模型。
 
 🍔 除此之外，ms-swift汇集了最新的训练技术，包括LoRA、QLoRA、Llama-Pro、LongLoRA、GaLore、Q-GaLore、LoRA+、LISA、DoRA、FourierFt、ReFT、UnSloth、和Liger等轻量化训练技术，以及DPO、GRPO、RM、PPO、KTO、CPO、SimPO、ORPO等人类对齐训练方法。ms-swift支持使用vLLM和LMDeploy对推理、评测和部署模块进行加速，并支持使用GPTQ、AWQ、BNB等技术对大模型进行量化。ms-swift还提供了基于Gradio的Web-UI界面及丰富的最佳实践。
 

docs/source/Instruction/命令行参数.md

@@ -164,6 +164,7 @@
 - train_dataloader_shuffle: CPT/SFT训练的dataloader是否随机，默认为True。该参数对IterableDataset无效。IterableDataset采用顺序的方式读取。
 - 🔥neftune_noise_alpha: neftune添加的噪声系数, 默认为0，通常可以设置为5、10、15。
 - 🔥use_liger_kernel: 是否启用[Liger](https://github.com/linkedin/Liger-Kernel)内核加速训练并减少显存消耗。默认为False。示例shell参考[这里](https://github.com/modelscope/ms-swift/blob/main/examples/train/liger)。
+  - 注意：liger_kernel不支持device_map，请使用DDP/DeepSpeed进行多卡训练。
 - average_tokens_across_devices: 是否在设备之间进行token数平均。如果设置为True，将使用all_reduce同步`num_tokens_in_batch`以进行精确的损失计算。默认为False。
 - max_grad_norm: 梯度裁剪。默认为1.。
 - push_to_hub: 推送checkpoint到hub。默认为False。

docs/source_en/Instruction/Command-line-parameters.md

@@ -168,6 +168,7 @@ Other important parameters:
 - train_dataloader_shuffle: Specifies whether the dataloader for CPT/SFT training is shuffled, with the default set to True. This parameter is not applicable to IterableDataset, as IterableDataset reads in a sequential manner.
 - 🔥neftune_noise_alpha: Coefficient of noise added by neftune, default is 0. Usually can be set to 5, 10, 15.
 - 🔥use_liger_kernel: Whether to enable the [Liger](https://github.com/linkedin/Liger-Kernel) kernel to accelerate training and reduce GPU memory consumption. Defaults to False. Example shell script can be found [here](https://github.com/modelscope/ms-swift/blob/main/examples/train/liger).
+  - Note: liger_kernel does not support device_map. Please use DDP/DeepSpeed for multi-GPU training.
 - average_tokens_across_devices: Whether to average the number of tokens across devices. If set to True, `num_tokens_in_batch` will be synchronized using all_reduce for accurate loss calculation. Default is False.
 - max_grad_norm: Gradient clipping. Default is 1.
 - push_to_hub: Push checkpoint to hub. Default is False.

swift/trainers/arguments.py

@@ -11,7 +11,7 @@
 from transformers.training_args import TrainingArguments as HfTrainingArguments
 from transformers.training_args_seq2seq import Seq2SeqTrainingArguments as HfSeq2SeqTrainingArguments
 
-from swift.utils import get_dist_setting, get_logger, is_liger_available, use_torchacc
+from swift.utils import get_dist_setting, get_logger, is_liger_available, is_mp, use_torchacc
 from .optimizers.galore import GaLoreConfig
 
 logger = get_logger()
@@ -87,6 +87,10 @@ def _init_liger(self):
             assert is_liger_available(), 'use_liger_kernel requires liger_kernels, try `pip install liger-kernel`'
 
     def __post_init__(self):
+        if is_mp() and self.use_liger_kernel:
+            raise ValueError('liger_kernel does not support device_map. '
+                             'Please use DDP/DeepSpeed for multi-GPU training.')
+
         from swift.llm.argument.base_args.model_args import ModelArguments
         if use_torchacc():
             self.dataloader_drop_last = True