vllm nightly cannot be installed when trying to run DeepSeek-OCR

[pplmx]

Reproduction

docker run \
    --name vllm-tester \
    -dit \
    --gpus all \
    --shm-size=32g \
    --ipc=host \
    --network=host \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --entrypoint bash \
    vllm/vllm-openai:v0.11.0

docker container exec -it vllm-tester bash


(DeepSeek-OCR) root@product-1-23:/var/tmp/DeepSeek-OCR# uv pip install -U vllm --pre --extra-index-url https://wheels.vllm.ai/nightly
  x Failed to resolve dependencies for `vllm` (v0.11.1rc3.dev49+g88d3141ec.cu129)
  `-> Package `triton-kernels` was included as a URL dependency. URL dependencies must be expressed as direct requirements or constraints. Consider adding `triton-kernels @ git+https://github.com/triton-lang/triton.git@v3.5.0#subdirectory=python/triton_kernels`
      to your dependencies or constraints file.

FYI

https://docs.vllm.ai/projects/recipes/en/latest/DeepSeek/DeepSeek-OCR.html#installing-vllm

[jgenc]

Try creating a constraints.txt file with the following content:

triton-kernels @ git+https://github.com/triton-lang/triton.git@v3.5.0#subdirectory=python/triton_kernels

Then, add both --constraint constraints.txt and --prerelease=allow to command you ran. This worked for me.

[pplmx]

Try creating a constraints.txt file with the following content:

triton-kernels @ git+https://github.com/triton-lang/triton.git@v3.5.0#subdirectory=python/triton_kernels
Then, add both --constraint constraints.txt and --prerelease=allow to command you ran. This worked for me.

@jgenc Thanks a lot, it works fine now. :)