Daniele
Daniele
[Installation]: No module named 'vllm._version' from vllm.version import __version__ as VLLM_VERSION
Duplicate of #9421
I'm seeing the same issue while running LORA tests on ROCm (MI300X): `pytest -v -s lora`: ``` + torch==2.7.0.dev20250308+rocm6.3 + pytorch-triton-rocm==3.2.0+git4b3bb1f8 ``` Traceback ``` ERROR 04-15 19:30:19 [core.py:390] EngineCore hit...
@ozziemoreno thanks, would you mind opening up a PR with your patch?
You're missing oneAPI Deep Neural Network Library (oneDNN) dependency, you can install it along with other required deps with: ```bash apt-get install -y libnuma-dev libdnnl-dev opencl-dev ```
@youkaichao > why do we need build isolation? vllm build needs to build against pytorch, and it does not make sense to install pytorch (can be complicated) in an isolated...
Closing this in favour of https://github.com/vllm-project/vllm/pull/13480
Some extra context for this: having an older `cmake` version would result in a cryptic error when building ```CMake Error at cmake/utils.cmake:10 (message): Unable to find python matching: /home/dtrifiro/work/vllm/.venv/bin/python. Call...
CI failures seem unrelated: ``` test_traces ... ValueError: OpenTelemetry packages must be installed before configuring 'otlp_traces_endpoint' ```
Anything holding this back?