From bc77d133339cabbf350201f84f54997c48497902 Mon Sep 17 00:00:00 2001 From: =?UTF-8?q?=E9=83=91=E7=89=B9=E9=A9=B9?= Date: Mon, 14 Jul 2025 11:28:38 +0800 Subject: [PATCH] Modify the installation process of Xtuner in the README_npu and README_npu_zh-CN files. --- ecosystem/README_npu.md | 39 ++++++++++++++++++++++++++++------- ecosystem/README_npu_zh-CN.md | 39 ++++++++++++++++++++++++++++------- 2 files changed, 62 insertions(+), 16 deletions(-) diff --git a/ecosystem/README_npu.md b/ecosystem/README_npu.md index 7b0e03a..401bf7b 100644 --- a/ecosystem/README_npu.md +++ b/ecosystem/README_npu.md @@ -72,18 +72,39 @@ source /usr/local/Ascend/ascend-toolkit/set_env.sh ### Installing Xtuner ```shell -git clone https://github.com/InternLM/xtuner.git +git clone -b v0.2.0rc0 https://github.com/InternLM/xtuner.git cd xtuner ``` -Modify `requirements/runtime.txt` with the following changes: +Modify `requirements/runtime.txt` to the following dependencies: ```text -bitsandbytes==0.42.0 -torchvision==0.19.0 +pyyaml +datasets>=3.2.0 +einops +loguru +mmengine==0.10.6 +openpyxl +peft>=0.14.0 +scikit-image +scipy +SentencePiece +tiktoken +torch==2.6 +torchvision==0.21.0 +transformers==4.48.0 +transformers_stream_generator +decorator numpy==1.26.4 ``` +Modify `requirements/deepspeed.txt` to the following dependencies: + +```text +deepspeed==0.16.2 +mpi4py +``` + Use the following command for installation: ```shell @@ -92,7 +113,7 @@ pip install -e '.[all]' **Note**: -- The default installation version of `torch` is the latest version. Please pay attention to match it with the version of `torch_npu`. +- Please note that the versions of `torch`, `torch_npu` and `torchvision` must match. ### LoRA Fine-tuning @@ -103,10 +124,12 @@ xtuner copy-cfg internlm2_5_chat_7b_qlora_oasst1_e3 . mv internlm2_5_chat_7b_qlora_oasst1_e3_copy.py internlm3_8b_instruct_lora_oasst1_e10.py ``` -The modifications to the configuration file `internlm3_8b_instruct_lora_oasst1_e10.py` are as follows: +Download the model weights `internlm3-8b-instruction` and dataset `oasst1` by yourself，the modifications to the configuration file `internlm3_8b_instruct_lora_oasst1_e10.py` are as follows: ```python -pretrained_model_name_or_path = 'internlm/internlm3-8b-instruct' +pretrained_model_name_or_path = 'path/to/internlm3-8b-instruct' + +data_path = 'path/to/oasst1' max_epochs = 10 @@ -137,7 +160,7 @@ Run the following commands to start single-machine eight-card fine-tuning: NPROC_PER_NODE=8 xtuner train internlm3_8b_instruct_lora_oasst1_e10.py --deepspeed deepspeed_zero2 ``` -The fine-tuning results are saved in the directory `./work_dirs/internlm3_8b_instruct_lora_oasst1_e10/iter_xxx.pth`. +The fine-tuning results are saved in the directory `./work_dirs/internlm3_8b_instruct_lora_oasst1_e10`. The comparison of loss between NPU and GPU is as follows: ![xtuner_training_loss](../assets/npu/xtuner_training_loss_compare.png) diff --git a/ecosystem/README_npu_zh-CN.md b/ecosystem/README_npu_zh-CN.md index 0e34679..6453224 100644 --- a/ecosystem/README_npu_zh-CN.md +++ b/ecosystem/README_npu_zh-CN.md @@ -72,18 +72,39 @@ source /usr/local/Ascend/ascend-toolkit/set_env.sh ### 安装 Xtuner ```shell -git clone https://github.com/InternLM/xtuner.git +git clone -b v0.2.0rc0 https://github.com/InternLM/xtuner.git cd xtuner ``` -修改`requirements/runtime.txt`，修改点如下： +修改`requirements/runtime.txt`，修改为如下依赖： ```text -bitsandbytes==0.42.0 -torchvision==0.19.0 +pyyaml +datasets>=3.2.0 +einops +loguru +mmengine==0.10.6 +openpyxl +peft>=0.14.0 +scikit-image +scipy +SentencePiece +tiktoken +torch==2.6 +torchvision==0.21.0 +transformers==4.48.0 +transformers_stream_generator +decorator numpy==1.26.4 ``` +修改`requirements/deepspeed.txt`,修改为如下依赖： + +```text +deepspeed==0.16.2 +mpi4py +``` + 使用以下命令进行安装： ```shell @@ -92,7 +113,7 @@ pip install -e '.[all]' **注意**: -- 默认安装`torch`为最新版，请注意与`torch_npu`版本相匹配 +- 请注意`torch`与`torch_npu`、`torchvision`版本要相匹配 ### LoRA 微调 @@ -103,10 +124,12 @@ xtuner copy-cfg internlm2_5_chat_7b_qlora_oasst1_e3 . mv internlm2_5_chat_7b_qlora_oasst1_e3_copy.py internlm3_8b_instruct_lora_oasst1_e10.py ``` -`internlm3_8b_instruct_lora_oasst1_e10.py`配置文件的修改点如下： +自行下载模型权重`internlm3-8b-instruct`以及数据集`oasst1`，`internlm3_8b_instruct_lora_oasst1_e10.py`配置文件的修改点如下： ```python -pretrained_model_name_or_path = 'internlm/internlm3-8b-instruct' +pretrained_model_name_or_path = 'path/to/internlm3-8b-instruct' + +data_path = 'path/to/oasst1' max_epochs = 10 @@ -137,7 +160,7 @@ randomness = dict(seed=123, deterministic=True) NPROC_PER_NODE=8 xtuner train internlm3_8b_instruct_lora_oasst1_e10.py --deepspeed deepspeed_zero2 ``` -微调后结果保存在`./work_dirs/internlm3_8b_instruct_lora_oasst1_e10/iter_xxx.pth`，NPU与GPU的loss对比如下： +微调后结果保存在`./work_dirs/internlm3_8b_instruct_lora_oasst1_e10/`，NPU与GPU的loss对比如下： ![xtuner_training_loss](../assets/npu/xtuner_training_loss_compare.png)