Keep image generic; move startup settings into run.sh/compose

2026-10-26 18:53:54 +08:00
parent 9f523e208a
commit d0e260c322
4 changed files with 25 additions and 17 deletions
--- a/README.md
+++ b/README.md
@@ -78,6 +78,11 @@ MODEL_HOST_DIR=/mnt/disk0/models/LLM-Research/Llama-4-Maverick-17B-128E-Instruct
    GPUs  : 0,1,2,...,15
 ```

+> 镜像本身不设置 `ENTRYPOINT` / `CMD` / `ENV` / `WORKDIR` / `EXPOSE`，
+> 启动方式、模型路径、端口、TP/PP、GPU 和挂载全部由 `run.sh` 或 `docker-compose.yml` 传入。
+> 如果直接 `docker run`，需要显式指定：
+> `--entrypoint /usr/local/bin/llama4-entrypoint.sh --workdir /opt/llama4`。
+
 ### 2.1 自动检测规则

 `detect_model.py` 读取 `config.json`：