[bugifx] QWen-1M context support[2/3] using current cuda stream in the DCA's kernel for bugfix. #8611
Conversation
There was a problem hiding this comment.
Summary of Changes
Hello @sighingnow, 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!
I've implemented a bugfix that is part of a larger effort to support QWen-1M context. This change specifically addresses an issue by ensuring that CUDA kernel launches within the vertical_slash_index.cu file explicitly utilize the current CUDA stream. This modification is essential for correct synchronization and execution, particularly in scenarios involving multiple CUDA streams, thereby resolving potential concurrency-related bugs.
Highlights
- CUDA Stream Integration: I've updated the
convert_vertical_slash_indexes_kernelandconvert_vertical_slash_indexes_kernel_mergeheadkernel launches to explicitly retrieve and pass the current CUDA stream. This ensures that these kernels execute on the correct stream, which is crucial for maintaining proper synchronization and preventing potential race conditions or incorrect behavior in multi-stream environments. - Header Inclusion: To facilitate the explicit use of the current CUDA stream, I've added
#include <c10/cuda/CUDAStream.h>. This header provides access to theat::cuda::getCurrentCUDAStream()function, which is used to obtain the stream handle.
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 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 or fill out our survey 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 correctly addresses a potential bug by ensuring the custom CUDA kernels in sgl-kernel/csrc/attention/vertical_slash_index.cu are launched on the current PyTorch CUDA stream. This is achieved by fetching the stream with at::cuda::getCurrentCUDAStream() and passing it to the kernel launch. The changes are applied consistently to both convert_vertical_slash_indexes_64x64 and convert_vertical_slash_indexes_64x64_mergehead, which is a crucial practice for preventing synchronization issues when integrating custom kernels with PyTorch. The addition of the <c10/cuda/CUDAStream.h> header is also appropriate. The changes are sound and improve the robustness of the CUDA integration.
Signed-off-by: Tao He <linzhu.ht@alibaba-inc.com>
…e DCA's kernel for bugfix. (sgl-project#8611) Signed-off-by: Tao He <linzhu.ht@alibaba-inc.com> Co-authored-by: sa-buc <linzhu.ht@w32d09270.cloud.sqa.na131>
…e DCA's kernel for bugfix. (sgl-project#8611) Signed-off-by: Tao He <linzhu.ht@alibaba-inc.com> Co-authored-by: sa-buc <linzhu.ht@w32d09270.cloud.sqa.na131>
…e DCA's kernel for bugfix. (#8611) Signed-off-by: Tao He <linzhu.ht@alibaba-inc.com> Co-authored-by: sa-buc <linzhu.ht@w32d09270.cloud.sqa.na131>
…e DCA's kernel for bugfix. (#8611) Signed-off-by: Tao He <linzhu.ht@alibaba-inc.com> Co-authored-by: sa-buc <linzhu.ht@w32d09270.cloud.sqa.na131>
…e DCA's kernel for bugfix. (#8611) Signed-off-by: Tao He <linzhu.ht@alibaba-inc.com> Co-authored-by: sa-buc <linzhu.ht@w32d09270.cloud.sqa.na131>
…e DCA's kernel for bugfix. (sgl-project#8611) Signed-off-by: Tao He <linzhu.ht@alibaba-inc.com> Co-authored-by: sa-buc <linzhu.ht@w32d09270.cloud.sqa.na131>
…e DCA's kernel for bugfix. (sgl-project#8611) Signed-off-by: Tao He <linzhu.ht@alibaba-inc.com> Co-authored-by: sa-buc <linzhu.ht@w32d09270.cloud.sqa.na131>
part of #5949