[WIP] Try new Triton package 20260608#392
Conversation
Signed-off-by: Lei Zhang <antiagainst@gmail.com>
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 6dc5712a68
ℹ️ About Codex in GitHub
Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".
| tokenspeed-proton==3.7.10.post20260608 | ||
| tokenspeed-triton==3.7.10.post20260608 |
There was a problem hiding this comment.
Keep staged pins out of release requirements
When this exact post20260608 pin is still only available from TestPyPI, the tokenspeed-kernel release build has no way to resolve it: .github/workflows/release-tokenspeed-kernel.yml builds wheels with only the PyTorch extra index, and setup.py's build path installs requirements/cuda.txt, which includes this file. The new CI install script preinstalls from TestPyPI only for test/ci_system/install_deps.sh, so release wheels and any source install outside that script fail before the kernel build starts. Either make the release/source build use the same index explicitly or avoid committing TestPyPI-only exact pins in the shared requirements.
Useful? React with 👍 / 👎.
|
will we bump the stable version @antiagainst |
No description provided.