目标
1、使用开源的大模型服务搭建属于自己的模型服务；
2、调优自己的大模型；
选型
采用通义千问模型，https://github.com/QwenLM/Qwen
步骤
1、下载模型文件
开源模型库：https://www.modelscope.cn/models
mkdir -p /data/qwen
cd /data/qwen
git clone --depth 1 https://www.modelscope.cn/qwen/Qwen-14B-Chat.git
# 小内存机器下载1.8B参数的，14B需要几十内存
# git clone --depth 1 https://www.modelscope.cn/qwen/Qwen-1_8B-Chat.git
2、下载使用docker 镜像
docker pull qwenllm/qwen3、启动脚本
https://github.com/QwenLM/Qwen/blob/main/docker/docker_web_demo.sh
# 修改如下内容
IMAGE_NAME=qwenllm/qwen
QWEN_CHECKPOINT_PATH=/data/qwen/Qwen-14B-Chat
PORT=8000
CONTAINER_NAME=qwen4、运行
访问http://localhost:8080 即可
sh docker_web_demo.sh输出如下，可以查看容器日志是否报错。
Successfully started web demo. Open '...' to try!
Run `docker logs ...` to check demo status.
Run `docker rm -f ...` to stop and remove the demo.效果
文档参考
https://github.com/QwenLM/Qwen/blob/main/README_CN.md
常见问题
1、运行报错
去掉docker_web_demo.sh中--gpus all
docker: Error response from daemon: could not select device driver "" with capabilities: [[gpu]].
2、Error while deserializing header: HeaderTooLarge
先安装yum install git-lfs 在下载模型文件，模型是git大文件管理，需要git-lfs的支持。
Traceback (most recent call last):
File "web_demo.py", line 209, in <module>
main()
File "web_demo.py", line 203, in main
model, tokenizer, config = _load_model_tokenizer(args)
File "web_demo.py", line 50, in _load_model_tokenizer
model = AutoModelForCausalLM.from_pretrained(
File "/usr/local/lib/python3.8/dist-packages/transformers/models/auto/auto_factory.py", line 511, in from_pretrained
return model_class.from_pretrained(
File "/usr/local/lib/python3.8/dist-packages/transformers/modeling_utils.py", line 3091, in from_pretrained
) = cls._load_pretrained_model(
File "/usr/local/lib/python3.8/dist-packages/transformers/modeling_utils.py", line 3456, in _load_pretrained_model
state_dict = load_state_dict(shard_file)
File "/usr/local/lib/python3.8/dist-packages/transformers/modeling_utils.py", line 458, in load_state_dict
with safe_open(checkpoint_file, framework="pt") as f:
safetensors_rust.SafetensorError: Error while deserializing header: HeaderTooLarge
3、Cannot allocate memory
内存不足，可以尝试选择1_8B小参数的模型。
