[Bug]: Fail to start vllm under /workspace

[Yikun]

Your current environment

In PR #661 , we change pip install from to pip install -e to help developers

Error like:

ERROR: cann't import name 'PoolingParams' from 'vllm'

ModuleNotFoundError: No module named 'vllm.benchmarks.serve'

🐛 Describe the bug

• Fail to start vllm under /workspace: [Bug]: Fail to start vllm under /workspace #725

  • Workaround:

    • 1. cd ~ or cd /, before you execute the vllm serve or launch openai server
    • 2. **(Update 20250430 9:00 AM) You can ** Use docker pull redownload v0.8.4rc2 image to refresh image

  • Resolved: [Build][Bugfix] Fix source code path to avoid reference error #726

    • we need follow vLLM, should move all source code into /vllm-workspace/, and keep /workspace as workdir:
      [1] https://github.com/vllm-project/vllm/blob/main/docker/Dockerfile#L264C7-L264C23
      [2] https://github.com/vllm-project/vllm/blob/main/docker/Dockerfile#L52

[MengqingCao]

vllm serve xxxx failed with ModuleNotFoundError: No module named 'vllm.benchmarks.serve' is also casued by the wrong inference path. will fix it asap.

we can inference normally under the other path:

root@537c0e4a8232:/# vllm serve facebook/opt-125m
INFO 04-29 11:21:45 [__init__.py:30] Available plugins for group vllm.platform_plugins:
INFO 04-29 11:21:45 [__init__.py:32] name=ascend, value=vllm_ascend:register
INFO 04-29 11:21:45 [__init__.py:34] all available plugins for group vllm.platform_plugins will be loaded.
INFO 04-29 11:21:45 [__init__.py:36] set environment variable VLLM_PLUGINS to control which plugins to load.
INFO 04-29 11:21:45 [__init__.py:44] plugin ascend loaded.
INFO 04-29 11:21:45 [__init__.py:230] Platform plugin ascend is activated
INFO 04-29 11:21:48 [__init__.py:30] Available plugins for group vllm.general_plugins:
INFO 04-29 11:21:48 [__init__.py:32] name=ascend_enhanced_model, value=vllm_ascend:register_model
INFO 04-29 11:21:48 [__init__.py:34] all available plugins for group vllm.general_plugins will be loaded.
INFO 04-29 11:21:48 [__init__.py:36] set environment variable VLLM_PLUGINS to control which plugins to load.
INFO 04-29 11:21:48 [__init__.py:44] plugin ascend_enhanced_model loaded.
INFO 04-29 11:21:48 [patch_tritonplaceholder.py:33] Triton not installed or not compatible; certain GPU-related functions will not be available.
WARNING 04-29 11:21:48 [patch_tritonplaceholder.py:46] Triton is not installed. Using dummy decorators. Install it via `pip install triton` to enable kernel compilation.
INFO 04-29 11:21:48 [patch_tritonplaceholder.py:71] Triton module has been replaced with a placeholder.
WARNING 04-29 11:21:48 [_custom_ops.py:21] Failed to import from vllm._C with ModuleNotFoundError("No module named 'vllm._C'")
WARNING 04-29 11:21:49 [registry.py:380] Model architecture DeepSeekMTPModel is already registered, and will be overwritten by the new model class vllm_ascend.models.deepseek_mtp:CustomDeepSeekMTP.
WARNING 04-29 11:21:49 [registry.py:380] Model architecture Qwen2VLForConditionalGeneration is already registered, and will be overwritten by the new model class vllm_ascend.models.qwen2_vl:CustomQwen2VLForConditionalGeneration.
WARNING 04-29 11:21:49 [registry.py:380] Model architecture DeepseekV2ForCausalLM is already registered, and will be overwritten by the new model class vllm_ascend.models.deepseek_v2:CustomDeepseekV2ForCausalLM.
WARNING 04-29 11:21:49 [registry.py:380] Model architecture DeepseekV3ForCausalLM is already registered, and will be overwritten by the new model class vllm_ascend.models.deepseek_v2:CustomDeepseekV3ForCausalLM