[TPU] Update torch version to include paged attention kernel change #19706
Conversation
|
👋 Hi! Thank you for contributing to the vLLM project. 💬 Join our developer Slack at https://slack.vllm.ai to discuss your PR in #pr-reviews, coordinate on features in #feat- channels, or join special interest groups in #sig- channels. Just a reminder: PRs would not trigger full CI run by default. Instead, it would only run Once the PR is approved and ready to go, your PR reviewer(s) can run CI to test the changes comprehensively before merging. To run CI, PR reviewers can either: Add 🚀 |
![gemini-code-assist[bot]](https://avatars.githubusercontent.com/in/956858?s=60&v=4){kind=small}
There was a problem hiding this comment.
Summary of Changes
Hello @Chenyaaang, I'm Gemini Code Assist1! I'm currently reviewing this pull request and will post my feedback shortly. In the meantime, here's a summary to help you and other reviewers quickly get up to speed!
This pull request updates the core PyTorch and PyTorch/XLA dependencies within the requirements/tpu.txt file. The primary purpose of this version bump is to integrate a specific paged attention kernel change, ensuring that TPU-based operations leverage the latest optimizations and features available in the updated libraries.
Highlights
- Dependency Updates: I've updated the
torch,torchvision, andtorch_xladependencies inrequirements/tpu.txtto version2.8.0.dev20250615. This change is specifically intended to incorporate a paged attention kernel update for TPU environments.
Using Gemini Code Assist
The full guide for Gemini Code Assist can be found on our documentation page, here are some quick tips.
Invoking Gemini
You can request assistance from Gemini at any point in your pull request via creating an issue comment (i.e. comment on the pull request page) using either /gemini <command> or @gemini-code-assist <command>. Below is a summary of the supported commands.
| Feature | Command | Description |
|---|---|---|
| Code Review | /gemini review |
Performs a code review for the current pull request in its current state. |
| Pull Request Summary | /gemini summary |
Provides a summary of the current pull request in its current state. |
| Comment | @gemini-code-assist | Responds in comments when explicitly tagged, both in issue comments and review comments. |
| Help | /gemini help |
Displays a list of available commands. |
Customization
To customize Gemini Code Assist for GitHub experience, repository maintainers can create a configuration file and/or provide a custom code review style guide (such as PEP-8 for Python) by creating and adding files to a .gemini/ folder in the base of the repository. Detailed instructions can be found here.
Limitations & Feedback
Gemini Code Assist is currently in preview and may make mistakes. Please leave feedback on any instances where its feedback is incorrect or counter productive. You can react with 👍 and 👎 on @gemini-code-assist comments to provide feedback.
You can also get AI-powered code generation, chat, as well as code reviews directly in the IDE at no cost with the Gemini Code Assist IDE Extension.
Footnotes
-
Review the Privacy Notices, Generative AI Prohibited Use Policy, Terms of Service, and learn how to configure Gemini Code Assist in GitHub here. Gemini can make mistakes, so double check it and use code with caution. ↩
There was a problem hiding this comment.
Code Review
This pull request updates PyTorch, TorchVision, and TorchXLA to newer development versions in the TPU requirements file to include paged attention kernel change. Ensure the stability of these new versions and their compatibility with other dependencies.
requirements/tpu.txt
Outdated
| torch_xla[tpu, pallas] @ https://storage.googleapis.com/pytorch-xla-releases/wheels/tpuvm/torch_xla-2.8.0.dev20250615-cp39-cp39-linux_x86_64.whl ; python_version == "3.9" | ||
| torch_xla[tpu, pallas] @ https://storage.googleapis.com/pytorch-xla-releases/wheels/tpuvm/torch_xla-2.8.0.dev20250615-cp310-cp310-linux_x86_64.whl ; python_version == "3.10" | ||
| torch_xla[tpu, pallas] @ https://storage.googleapis.com/pytorch-xla-releases/wheels/tpuvm/torch_xla-2.8.0.dev20250615-cp311-cp311-linux_x86_64.whl ; python_version == "3.11" |
There was a problem hiding this comment.
Verify that the torch_xla version 2.8.0.dev20250615 is compatible with the underlying XLA compiler and libtpu versions in the base Docker image. Mismatches can lead to runtime errors.
requirements/tpu.txt
Outdated
| torch_xla[tpu, pallas] @ https://storage.googleapis.com/pytorch-xla-releases/wheels/tpuvm/torch_xla-2.8.0.dev20250605-cp39-cp39-linux_x86_64.whl ; python_version == "3.9" | ||
| torch_xla[tpu, pallas] @ https://storage.googleapis.com/pytorch-xla-releases/wheels/tpuvm/torch_xla-2.8.0.dev20250605-cp310-cp310-linux_x86_64.whl ; python_version == "3.10" | ||
| torch_xla[tpu, pallas] @ https://storage.googleapis.com/pytorch-xla-releases/wheels/tpuvm/torch_xla-2.8.0.dev20250605-cp311-cp311-linux_x86_64.whl ; python_version == "3.11" | ||
| torch==2.8.0.dev20250615 |
There was a problem hiding this comment.
Consider testing the stability of the torch version 2.8.0.dev20250615 and ensuring no regressions are introduced in the TPU environment, given that this is a development build.
requirements/tpu.txt
Outdated
| torch_xla[tpu, pallas] @ https://storage.googleapis.com/pytorch-xla-releases/wheels/tpuvm/torch_xla-2.8.0.dev20250605-cp310-cp310-linux_x86_64.whl ; python_version == "3.10" | ||
| torch_xla[tpu, pallas] @ https://storage.googleapis.com/pytorch-xla-releases/wheels/tpuvm/torch_xla-2.8.0.dev20250605-cp311-cp311-linux_x86_64.whl ; python_version == "3.11" | ||
| torch==2.8.0.dev20250615 | ||
| torchvision==0.23.0.dev20250615 |
There was a problem hiding this comment.
Consider testing the stability of the torchvision version 0.23.0.dev20250615 and ensuring no regressions are introduced in the TPU environment, given that this is a development build.
yaochengji
added
tpu
Signed-off-by: Chenyaaang <[email protected]>
Chenyaaang
force-pushed
the
update-requirement
branch
from
ec04145 to
d00eb38
Compare
…llm-project#19706) Signed-off-by: Chenyaaang <[email protected]>
…llm-project#19706) Signed-off-by: Chenyaaang <[email protected]> Signed-off-by: minpeter <[email protected]>
…llm-project#19706) Signed-off-by: Chenyaaang <[email protected]> Signed-off-by: Yang Wang <[email protected]>
…llm-project#19706) Signed-off-by: Chenyaaang <[email protected]>
…llm-project#19706) Signed-off-by: Chenyaaang <[email protected]>
…llm-project#19706) Signed-off-by: Chenyaaang <[email protected]>
…llm-project#19706) Signed-off-by: Chenyaaang <[email protected]> Signed-off-by: avigny <[email protected]>
No description provided.