[Usage]: Found Tesla P40 which is too old to be supported by the triton GPU compiler, which is used as the backend.

[amzfc]

Your current environment

docker images vllm-openai:v0.7.2

How would you like to use vllm

vllm server --model", "/models/Qwen2.5-7B-Instruct", "--served-model-name", "qwen2.5-7b-instruct", "--gpu-memory-utilization", "0.90", "--dtype" ,"half" ,"--tensor-parallel-size", "2"

when i run vllm , the error is raised.

RuntimeError: Found Tesla P40 which is too old to be supported by the triton GPU compiler, which is used as the backend. Triton only supports devices of CUDA Capability >= 7.0, but your device is of CUDA capability 6.1

How to solve it？

Before submitting a new issue...

• Make sure you already searched for relevant issues, and asked the chatbot living at the bottom right corner of the documentation page, which can answer lots of frequently asked questions.

[hmellor]

See https://docs.vllm.ai/en/latest/features/compatibility_matrix.html#feature-x-hardware for supported hardware

[NW15D]

it can be use with modified triton , only via manual install see it https://github.com/sasha0552/pascal-pkgs-ci