Conversation
…after addressing all identified issues and implementing the corresponding fixes. This cleanup streamlines the repository and eliminates outdated documentation.
- Introduced `cuda-bindings` version 13.1.1 with multiple wheel distributions for various platforms. - Added `cuda-core` version 0.6.0, supporting different numpy versions based on Python compatibility. - Updated benchmarks to handle modal imports more gracefully, allowing local execution without modal. - Refactored benchmark scripts to improve structure and error handling, including warmup for JIT compilation. - Enhanced documentation for running benchmarks locally and on Modal.
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 6b49a8d02b
ℹ️ About Codex in GitHub
Your team has set up Codex to 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 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| from sklearn.datasets import fetch_california_housing | ||
| from sklearn.model_selection import train_test_split | ||
|
|
||
| data = fetch_california_housing() |
There was a problem hiding this comment.
The new default benchmark path calls fetch_california_housing(), which downloads data when it is not already cached; this means benchmarks/run.py can fail on first run in offline or restricted CI environments even though the script advertises “zero external dependencies” and runs regression by default under --task all. In those environments the whole benchmark suite aborts before any timing results are produced, so this should use a bundled dataset or a local fallback path when download is unavailable.
Useful? React with 👍 / 👎.
1 participant
No description provided.