This commit is contained in:
2026-03-30 23:27:58 +08:00
parent 1aa8d011e8
commit dbd934cdda
2 changed files with 4 additions and 1 deletions
+3
View File
@@ -6,6 +6,9 @@ COPY requirements.txt /workspace/requirements.txt
RUN pip install --no-cache-dir --retries 20 --timeout 600 -i https://pypi.tuna.tsinghua.edu.cn/simple -r /workspace/requirements.txt \
|| pip install --no-cache-dir --retries 20 --timeout 600 -i https://mirrors.aliyun.com/pypi/simple -r /workspace/requirements.txt \
|| pip install --no-cache-dir --retries 20 --timeout 600 -r /workspace/requirements.txt
RUN pip install --no-cache-dir --retries 20 --timeout 600 -i https://pypi.tuna.tsinghua.edu.cn/simple --upgrade --pre transformers \
|| pip install --no-cache-dir --retries 20 --timeout 600 -i https://mirrors.aliyun.com/pypi/simple --upgrade --pre transformers \
|| pip install --no-cache-dir --retries 20 --timeout 600 --upgrade --pre transformers
COPY app /workspace/app
+1 -1
View File
@@ -124,7 +124,7 @@ curl -X POST "http://localhost:8001/v1/chat/completions" \
## 常见故障排查
- 报错 `model type ... Transformers does not recognize this architecture` 时,说明当前模型与镜像内依赖不兼容,建议更换模型或升级镜像版本。
- 报错 `model type ... Transformers does not recognize this architecture` 时,先确认已使用当前 Dockerfile 重建镜像(其中会升级 `transformers`),并执行 `docker compose build --no-cache` 后再启动。
- 报错 `model config (gptq) does not match quantization argument (gptq_marlin)` 时,将该模型配置改为 `dtype=float16` 且 `quantization=gptq`。
- 报错 `RPC call to sample_tokens timed out` 或出现 `GPU core dump` 时,先下调模型配置为更稳参数:`ctx=32768`、`max_num_seqs=4`、`max_tokens=2048`、`gpu_util=0.90`,并开启 `enforce_eager=true`。
- 若模型目录存在但仍加载失败,检查挂载路径是否为 `/opt/model:/opt/model:ro`,并确认容器内可见模型文件。