AutoDL 上部署 InternLM-XComposer 踩坑记录

[vansin]

环境配置

cd /root/autodl-tmp
source /etc/network_turbo 
git clone https://github.com/InternLM/InternLM-XComposer.git
unset http_proxy && unset https_proxy

将 pypi 镜像切换为清华源

pip config set global.index-url https://pypi.tuna.tsinghua.edu.cn/simple
pip install transformers
pip install modelscope

下载模型

方法一

import torch
from modelscope import snapshot_download, AutoModel, AutoTokenizer

torch.set_grad_enabled(False)

# init model and tokenizer
model_dir = snapshot_download('Shanghai_AI_Laboratory/internlm-xcomposer-7b')
model = AutoModel.from_pretrained(model_dir, trust_remote_code=True).cuda().eval()

方法二

sudo apt update
sudo apt install git-lfs
git lfs install

cd /root/autodl-tmp
git clone https://www.modelscope.cn/Shanghai_AI_Laboratory/internlm-xcomposer-7b.git

运行 Demo

更改 web_demo.py 中模型的地址，将 48 和 50 行的模型路径改成改成 ``