Skip to content

Conversation

@zwt-1234
Copy link
Collaborator

No description provided.

@XprobeBot XprobeBot added this to the v1.x milestone Dec 22, 2025
@qinxuye qinxuye changed the title Update Dockerfile to 12.9 to use VLLM v0.11.2 version BLD: Update Dockerfile to 12.9 to use VLLM v0.11.2 version Dec 23, 2025
cuda-python
sgl-kernel>=0.0.3.post3,<=0.1.4
IPython
numpy==1.26.4
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Should numpy be such an old version?

Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Version cannot exceed 2, this version is compatible, otherwise an error message will be generated stating that the version is too high

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Can you paste the error stack?

RUN pip3 install --pre torch torchvision torchaudio --index-url https://download.pytorch.org/whl/nightly/cu128 \
--no-deps
--no-deps && \
pip install triton==3.4.0 && \
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Why pin triton and torchcodec

Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Before the cu129hu version environment, it was directly read from the byte stream and then saved to memory; But in the cu129 version environment, torchauudio relies on torchcodec, which requires storing a temporary file first and then reading it into the byte stream

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Then why pinned triton?

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants