From d5e5911031b5d756ecfcc3e60fd27163536ddca2 Mon Sep 17 00:00:00 2001 From: SZCJW <792430652@qq.com> Date: Tue, 31 Mar 2026 00:27:03 +0800 Subject: [PATCH] x --- Dockerfile | 3 --- README.md | 1 + 2 files changed, 1 insertion(+), 3 deletions(-) diff --git a/Dockerfile b/Dockerfile index ac76c9f..e08a27c 100644 --- a/Dockerfile +++ b/Dockerfile @@ -6,9 +6,6 @@ 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 diff --git a/README.md b/README.md index 23ef15c..50107d7 100644 --- a/README.md +++ b/README.md @@ -125,6 +125,7 @@ curl -X POST "http://localhost:8001/v1/chat/completions" \ ## 常见故障排查 - 报错 `Model architectures ['Qwen3_5MoeForConditionalGeneration'] are not supported for now` 或 `The Transformers implementation ... is not compatible with vLLM` 时,说明当前 vLLM 栈与该模型架构不兼容,需切换到兼容模型或改用其他推理后端。 +- 报错 `StrictDataclassClassValidationError` 且包含 `validate_rope` / `unsupported operand type(s) for -=: 'set' and 'list'` 时,移除 Dockerfile 中对 `transformers --upgrade --pre` 的强制升级,使用镜像内置依赖重建。 - 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。