KEMBAR78
[Fix][Chore][Qwen3] fix bug of using fp4 on sm120 by byshiue · Pull Request #6065 · NVIDIA/TensorRT-LLM · GitHub
Skip to content

[Fix][Chore][Qwen3] fix bug of using fp4 on sm120 #6065

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Merged
merged 4 commits into from
Jul 20, 2025
Merged

[Fix][Chore][Qwen3] fix bug of using fp4 on sm120 #6065

merged 4 commits into from
Jul 20, 2025
+5 −5

Conversation

@byshiue
Copy link
Collaborator

@byshiue byshiue commented Jul 15, 2025 •
edited by coderabbitai bot
Loading

Summary by CodeRabbit

  • Bug Fixes

    • Improved compatibility checks to ensure NVFP4 output is not supported on specific hardware (SM 120), preventing potential issues on unsupported devices.

  • Documentation

    • Updated comments to clarify hardware support details for NVFP4 output.
    • Corrected model name in the Qwen3 support matrix to reflect the accurate identifier.

  • Tests

    • Adjusted GPU memory reservation settings in key-value cache tests for improved resource management.
    • Removed skips on specific tests previously waived due to a resolved issue; updated bug references accordingly.

@byshiue
Copy link
Collaborator Author

byshiue commented Jul 16, 2025

/bot run

@byshiue byshiue enabled auto-merge (squash) July 16, 2025 15:15
@tensorrt-cicd
Copy link
Collaborator

PR_Github #12092 [ run ] triggered by Bot

@tensorrt-cicd
Copy link
Collaborator

PR_Github #12092 [ run ] completed with state FAILURE
/LLM/main/L0_MergeRequest_PR pipeline #8981 completed with status: 'FAILURE'

@byshiue
Copy link
Collaborator Author

byshiue commented Jul 16, 2025

/bot run

@tensorrt-cicd
Copy link
Collaborator

PR_Github #12120 [ run ] triggered by Bot

@tensorrt-cicd
Copy link
Collaborator

PR_Github #12120 [ run ] completed with state FAILURE
/LLM/main/L0_MergeRequest_PR pipeline #9005 completed with status: 'FAILURE'

@coderabbitai
Copy link
Contributor

coderabbitai bot commented Jul 17, 2025 •
edited
Loading

Walkthrough

A function determining NVFP4 output support in attention operations was updated to explicitly exclude SM version 120, in addition to versions below 100. The code comment was revised to reflect that SM 120 does not support NVFP4 output. The Qwen3 model name was corrected in documentation. A test parameter adjusting GPU memory fraction was changed, and two skip entries for specific tests were removed from the waiver list. No changes were made to function signatures or public interfaces.

Changes

File Change Summary
cpp/tensorrt_llm/thop/attentionOp.cpp Updated condition to exclude SM 120 from NVFP4 support; revised comment for clarification.
examples/models/core/qwen/README.md Corrected Qwen3 model name in support matrix from "Qwen3-235B-A3B" to "Qwen3-235B-A22B".
tests/integration/defs/accuracy/test_llm_api_pytorch.py Changed free_gpu_memory_fraction in KvCacheConfig from 0.6 to 0.4 in test_nvfp4 test method.
tests/integration/test_lists/waives.txt Removed two skip entries for TestQwen3_235B_A22B::test_nvfp4 tests with [latency_moe_cutlass] and [latency_moe_trtllm] parameters.

Suggested reviewers

  • yiqingy0
  • litaotju

Poem

In the land of SMs, a change takes its flight,
SM 120 now fades from NVFP4's light.
The Qwen3 name was fixed with care,
Tests freed from skips, memory set fair.
The code hops ahead, as all rabbits know,
With tweaks that help the models grow! 🐇✨

📜 Recent review details

Configuration used: .coderabbit.yaml
Review profile: CHILL
Plan: Pro

📥 Commits

Reviewing files that changed from the base of the PR and between ec855c4 and 0fb6e59.

📒 Files selected for processing (1)
  • tests/integration/test_lists/waives.txt (1 hunks)
🚧 Files skipped from review as they are similar to previous changes (1)
  • tests/integration/test_lists/waives.txt
🪧 Tips

Chat

There are 3 ways to chat with CodeRabbit:

  • Review comments: Directly reply to a review comment made by CodeRabbit. Example:
    • I pushed a fix in commit <commit_id>, please review it.
    • Explain this complex logic.
    • Open a follow-up GitHub issue for this discussion.
  • Files and specific lines of code (under the "Files changed" tab): Tag @coderabbitai in a new review comment at the desired location with your query. Examples:
    • @coderabbitai explain this code block.
    • @coderabbitai modularize this function.
  • PR comments: Tag @coderabbitai in a new PR comment to ask questions about the PR branch. For the best results, please provide a very specific query, as very limited context is provided in this mode. Examples:
    • @coderabbitai gather interesting stats about this repository and render them as a table. Additionally, render a pie chart showing the language distribution in the codebase.
    • @coderabbitai read src/utils.ts and explain its main purpose.
    • @coderabbitai read the files in the src/scheduler package and generate a class diagram using mermaid and a README in the markdown format.
    • @coderabbitai help me debug CodeRabbit configuration file.

Support

Need help? Create a ticket on our support page for assistance with any issues or questions.

Note: Be mindful of the bot's finite context window. It's strongly recommended to break down tasks such as reading entire modules into smaller chunks. For a focused discussion, use review comments to chat about specific files and their changes, instead of using the PR comments.

CodeRabbit Commands (Invoked using PR comments)

  • @coderabbitai pause to pause the reviews on a PR.
  • @coderabbitai resume to resume the paused reviews.
  • @coderabbitai review to trigger an incremental review. This is useful when automatic reviews are disabled for the repository.
  • @coderabbitai full review to do a full review from scratch and review all the files again.
  • @coderabbitai summary to regenerate the summary of the PR.
  • @coderabbitai generate docstrings to generate docstrings for this PR.
  • @coderabbitai generate sequence diagram to generate a sequence diagram of the changes in this PR.
  • @coderabbitai resolve resolve all the CodeRabbit review comments.
  • @coderabbitai configuration to show the current CodeRabbit configuration for the repository.
  • @coderabbitai help to get help.

Other keywords and placeholders

  • Add @coderabbitai ignore anywhere in the PR description to prevent this PR from being reviewed.
  • Add @coderabbitai summary to generate the high-level summary at a specific location in the PR description.
  • Add @coderabbitai anywhere in the PR title to generate the title automatically.

Documentation and Community

  • Visit our Documentation for detailed information on how to use CodeRabbit.
  • Join our Discord Community to get help, request features, and share feedback.
  • Follow us on X/Twitter for updates and announcements.

@byshiue
Copy link
Collaborator Author

byshiue commented Jul 17, 2025

/bot run

@tensorrt-cicd
Copy link
Collaborator

PR_Github #12231 [ run ] triggered by Bot

@tensorrt-cicd
Copy link
Collaborator

PR_Github #12231 [ run ] completed with state FAILURE
/LLM/main/L0_MergeRequest_PR pipeline #9086 completed with status: 'FAILURE'

@byshiue
Copy link
Collaborator Author

byshiue commented Jul 17, 2025

/bot run

@tensorrt-cicd
Copy link
Collaborator

PR_Github #12245 [ run ] triggered by Bot

@tensorrt-cicd
Copy link
Collaborator

PR_Github #12244 [ ] completed with state ABORTED

@tensorrt-cicd
Copy link
Collaborator

PR_Github #12245 [ run ] completed with state FAILURE
/LLM/main/L0_MergeRequest_PR pipeline #9094 completed with status: 'FAILURE'

@byshiue
Copy link
Collaborator Author

byshiue commented Jul 18, 2025

/bot run

@byshiue
fix bug of using fp4 on sm120
ef6d38a 
Signed-off-by: bhsueh <11360707+byshiue@users.noreply.github.com>
@byshiue
Copy link
Collaborator Author

byshiue commented Jul 18, 2025

/bot run

@tensorrt-cicd
Copy link
Collaborator

PR_Github #12349 [ run ] triggered by Bot

@tensorrt-cicd
Copy link
Collaborator

PR_Github #12350 [ run ] triggered by Bot

@tensorrt-cicd
Copy link
Collaborator

PR_Github #12349 [ run ] completed with state ABORTED

@byshiue
fix typo of qwen3 doc
12ddbe5 
Signed-off-by: bhsueh <11360707+byshiue@users.noreply.github.com>
@tensorrt-cicd
Copy link
Collaborator

PR_Github #12350 [ run ] completed with state SUCCESS
/LLM/main/L0_MergeRequest_PR pipeline #9172 completed with status: 'FAILURE'

byshiue added 2 commits July 18, 2025 22:47
@byshiue
use smaller free_gpu_memory_fraction to prevent OOM and unwavie the q…
ec855c4 
…wen3_235b tests

Signed-off-by: bhsueh <11360707+byshiue@users.noreply.github.com>
@byshiue
add qwen3 trtllm moe backend tests into waive list
0fb6e59 
Signed-off-by: bhsueh <11360707+byshiue@users.noreply.github.com>
@byshiue
Copy link
Collaborator Author

byshiue commented Jul 18, 2025

/bot run --disable-fail-fast

@byshiue
Copy link
Collaborator Author

byshiue commented Jul 19, 2025

/bot run --disable-fail-fast

@tensorrt-cicd
Copy link
Collaborator

PR_Github #12367 [ run ] triggered by Bot

@tensorrt-cicd
Copy link
Collaborator

PR_Github #12367 [ run ] completed with state SUCCESS
/LLM/main/L0_MergeRequest_PR pipeline #9188 completed with status: 'SUCCESS'

@byshiue byshiue merged commit 2e14c8f into NVIDIA:main Jul 20, 2025
3 checks passed
reasonsolo pushed a commit to reasonsolo/TensorRT-LLM that referenced this pull request Jul 21, 2025
@byshiue @reasonsolo
[Fix][Chore][Qwen3] fix bug of using fp4 on sm120 (NVIDIA#6065)
076d814 
Signed-off-by: bhsueh <11360707+byshiue@users.noreply.github.com>
timlee0212 pushed a commit to timlee0212/TensorRT-LLM that referenced this pull request Jul 21, 2025
@byshiue @timlee0212
[Fix][Chore][Qwen3] fix bug of using fp4 on sm120 (NVIDIA#6065)
9396966 
Signed-off-by: bhsueh <11360707+byshiue@users.noreply.github.com>
NVShreyas pushed a commit to NVShreyas/TensorRT-LLM that referenced this pull request Jul 28, 2025
@byshiue @NVShreyas
[Fix][Chore][Qwen3] fix bug of using fp4 on sm120 (NVIDIA#6065)
b22a9ae 
Signed-off-by: bhsueh <11360707+byshiue@users.noreply.github.com>
Signed-off-by: Shreyas Misra <shreyasm@nvidia.com>
Ransiki pushed a commit to Ransiki/TensorRT-LLM that referenced this pull request Jul 29, 2025
@byshiue @Ransiki
[Fix][Chore][Qwen3] fix bug of using fp4 on sm120 (NVIDIA#6065)
03e7064 
Signed-off-by: bhsueh <11360707+byshiue@users.noreply.github.com>
Signed-off-by: Ransiki Zhang <ransikiz@nvidia.com>
This was referenced Aug 8, 2025
Merged
Merged
@coderabbitai coderabbitai bot mentioned this pull request Aug 18, 2025
Merged
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@kaiyux kaiyux kaiyux approved these changes

+1 more reviewer

@lucifer1004 lucifer1004 lucifer1004 approved these changes

Reviewers whose approvals may not affect merge requirements

Assignees

No one assigned

Labels

None yet

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

4 participants

@byshiue @tensorrt-cicd @lucifer1004 @kaiyux