Skip to content

Conversation

@PerkzZheng
Copy link
Collaborator

@PerkzZheng PerkzZheng commented Sep 8, 2025

@coderabbitai summary

Description

This MR adds the fp8-kv-cache support for gpt-oss models. It also enables the fp8 context fmha on blackwell when needed (fp8 kv cache + chunked prefill/kv cache reuse).

The gsm8k scores:

  • gpt-oss-20b
    • bf16 kv cache: 88.2487
    • fp8 kv cache: 88.0970
  • gpt-oss-120b:
    • bf16 kv cache: 90.2199
    • fp8 kv cache: 89.3859

trtllm-gen commit: 93e21886ec1e32e0dff657dbe628a74e8d62b151

This MR also fixes a regression of trtllm-gen kernels due to loading scales from global memory.

Test Coverage

PR Checklist

Please review the following before submitting your PR:

  • PR description clearly explains what and why. If using CodeRabbit's summary, please make sure it makes sense.

  • PR Follows TRT-LLM CODING GUIDELINES to the best of your knowledge.

  • Test cases are provided for new code paths (see test instructions)

  • Any new dependencies have been scanned for license and vulnerabilities

  • CODEOWNERS updated if ownership changes

  • Documentation updated as needed

  • The reviewers assigned automatically/manually are appropriate for the PR.

  • Please check this after reviewing the above items as appropriate for this PR.

GitHub Bot Help

/bot [-h] ['run', 'kill', 'skip', 'reuse-pipeline'] ...

Provide a user friendly way for developers to interact with a Jenkins server.

Run /bot [-h|--help] to print this help message.

See details below for each supported subcommand.

Details

run [--reuse-test (optional)pipeline-id --disable-fail-fast --skip-test --stage-list "A10-PyTorch-1, xxx" --gpu-type "A30, H100_PCIe" --test-backend "pytorch, cpp" --add-multi-gpu-test --only-multi-gpu-test --disable-multi-gpu-test --post-merge --extra-stage "H100_PCIe-TensorRT-Post-Merge-1, xxx" --detailed-log --debug(experimental)]

Launch build/test pipelines. All previously running jobs will be killed.

--reuse-test (optional)pipeline-id (OPTIONAL) : Allow the new pipeline to reuse build artifacts and skip successful test stages from a specified pipeline or the last pipeline if no pipeline-id is indicated. If the Git commit ID has changed, this option will be always ignored. The DEFAULT behavior of the bot is to reuse build artifacts and successful test results from the last pipeline.

--disable-reuse-test (OPTIONAL) : Explicitly prevent the pipeline from reusing build artifacts and skipping successful test stages from a previous pipeline. Ensure that all builds and tests are run regardless of previous successes.

--disable-fail-fast (OPTIONAL) : Disable fail fast on build/tests/infra failures.

--skip-test (OPTIONAL) : Skip all test stages, but still run build stages, package stages and sanity check stages. Note: Does NOT update GitHub check status.

--stage-list "A10-PyTorch-1, xxx" (OPTIONAL) : Only run the specified test stages. Examples: "A10-PyTorch-1, xxx". Note: Does NOT update GitHub check status.

--gpu-type "A30, H100_PCIe" (OPTIONAL) : Only run the test stages on the specified GPU types. Examples: "A30, H100_PCIe". Note: Does NOT update GitHub check status.

--test-backend "pytorch, cpp" (OPTIONAL) : Skip test stages which don't match the specified backends. Only support [pytorch, cpp, tensorrt, triton]. Examples: "pytorch, cpp" (does not run test stages with tensorrt or triton backend). Note: Does NOT update GitHub pipeline status.

--only-multi-gpu-test (OPTIONAL) : Only run the multi-GPU tests. Note: Does NOT update GitHub check status.

--disable-multi-gpu-test (OPTIONAL) : Disable the multi-GPU tests. Note: Does NOT update GitHub check status.

--add-multi-gpu-test (OPTIONAL) : Force run the multi-GPU tests in addition to running L0 pre-merge pipeline.

--post-merge (OPTIONAL) : Run the L0 post-merge pipeline instead of the ordinary L0 pre-merge pipeline.

--extra-stage "H100_PCIe-TensorRT-Post-Merge-1, xxx" (OPTIONAL) : Run the ordinary L0 pre-merge pipeline and specified test stages. Examples: --extra-stage "H100_PCIe-TensorRT-Post-Merge-1, xxx".

--detailed-log (OPTIONAL) : Enable flushing out all logs to the Jenkins console. This will significantly increase the log volume and may slow down the job.

--debug (OPTIONAL) : Experimental feature. Enable access to the CI container for debugging purpose. Note: Specify exactly one stage in the stage-list parameter to access the appropriate container environment. Note: Does NOT update GitHub check status.

For guidance on mapping tests to stage names, see docs/source/reference/ci-overview.md
and the scripts/test_to_stage_mapping.py helper.

kill

kill

Kill all running builds associated with pull request.

skip

skip --comment COMMENT

Skip testing for latest commit on pull request. --comment "Reason for skipping build/test" is required. IMPORTANT NOTE: This is dangerous since lack of user care and validation can cause top of tree to break.

reuse-pipeline

reuse-pipeline

Reuse a previous pipeline to validate current commit. This action will also kill all currently running builds associated with the pull request. IMPORTANT NOTE: This is dangerous since lack of user care and validation can cause top of tree to break.

@PerkzZheng
Copy link
Collaborator Author

/bot run

@tensorrt-cicd
Copy link
Collaborator

PR_Github #18033 [ run ] triggered by Bot

@tensorrt-cicd
Copy link
Collaborator

PR_Github #18033 [ run ] completed with state FAILURE

@PerkzZheng
Copy link
Collaborator Author

/bot run

@tensorrt-cicd
Copy link
Collaborator

PR_Github #18036 [ run ] triggered by Bot

@tensorrt-cicd
Copy link
Collaborator

PR_Github #18036 [ run ] completed with state SUCCESS
/LLM/main/L0_MergeRequest_PR pipeline #13515 completed with status: 'FAILURE'

@PerkzZheng PerkzZheng force-pushed the user/perkzz/fp8-gpt-oss branch from aa3e71b to 1268843 Compare September 9, 2025 02:40
@PerkzZheng
Copy link
Collaborator Author

/bot run

@PerkzZheng PerkzZheng requested review from hlu1 and yuxianq September 9, 2025 02:43
@tensorrt-cicd
Copy link
Collaborator

PR_Github #18131 [ run ] triggered by Bot

@tensorrt-cicd
Copy link
Collaborator

PR_Github #18131 [ run ] completed with state FAILURE

@PerkzZheng PerkzZheng added the TRTLLM-Gen update updates the trtllm-gen kernels (for blackwell). label Sep 9, 2025
@tensorrt-cicd
Copy link
Collaborator

PR_Github #18153 [ run ] triggered by Bot

@tensorrt-cicd
Copy link
Collaborator

PR_Github #18153 [ run ] completed with state FAILURE

@tensorrt-cicd
Copy link
Collaborator

PR_Github #18174 [ run ] triggered by Bot

@PerkzZheng
Copy link
Collaborator Author

/bot run

@tensorrt-cicd
Copy link
Collaborator

PR_Github #18185 [ run ] triggered by Bot

@tensorrt-cicd
Copy link
Collaborator

PR_Github #18174 [ run ] completed with state ABORTED

@PerkzZheng
Copy link
Collaborator Author

/bot run

@tensorrt-cicd
Copy link
Collaborator

PR_Github #18189 [ run ] triggered by Bot

@tensorrt-cicd
Copy link
Collaborator

PR_Github #18185 [ run ] completed with state ABORTED

Signed-off-by: Perkz Zheng <[email protected]>
@PerkzZheng PerkzZheng force-pushed the user/perkzz/fp8-gpt-oss branch from e255040 to c339a11 Compare September 9, 2025 09:24
@PerkzZheng
Copy link
Collaborator Author

/bot run

@tensorrt-cicd
Copy link
Collaborator

PR_Github #18206 [ run ] triggered by Bot

@tensorrt-cicd
Copy link
Collaborator

PR_Github #18206 [ run ] completed with state SUCCESS
/LLM/main/L0_MergeRequest_PR pipeline #13643 completed with status: 'FAILURE'

Signed-off-by: Perkz Zheng <[email protected]>
@PerkzZheng
Copy link
Collaborator Author

/bot run

@tensorrt-cicd
Copy link
Collaborator

PR_Github #18230 [ run ] triggered by Bot

@PerkzZheng
Copy link
Collaborator Author

@NVIDIA/trt-llm-torch-models-devs @NVIDIA/trt-llm-doc-owners can you help review my changes ? Thanks.

Copy link
Collaborator

@yechank-nvidia yechank-nvidia left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM.

Signed-off-by: Perkz Zheng <[email protected]>
@PerkzZheng
Copy link
Collaborator Author

/bot run

@tensorrt-cicd
Copy link
Collaborator

PR_Github #18304 [ run ] triggered by Bot

@PerkzZheng PerkzZheng enabled auto-merge (squash) September 10, 2025 06:02
@tensorrt-cicd
Copy link
Collaborator

PR_Github #18304 [ run ] completed with state FAILURE
/LLM/main/L0_MergeRequest_PR pipeline #13724 completed with status: 'FAILURE'

@PerkzZheng
Copy link
Collaborator Author

/bot run

@tensorrt-cicd
Copy link
Collaborator

PR_Github #18322 [ run ] triggered by Bot

@tensorrt-cicd
Copy link
Collaborator

PR_Github #18322 [ run ] completed with state FAILURE
/LLM/main/L0_MergeRequest_PR pipeline #13740 completed with status: 'FAILURE'

@PerkzZheng
Copy link
Collaborator Author

/bot run

@tensorrt-cicd
Copy link
Collaborator

PR_Github #18346 [ run ] triggered by Bot

@tensorrt-cicd
Copy link
Collaborator

PR_Github #18346 [ run ] completed with state SUCCESS
/LLM/main/L0_MergeRequest_PR pipeline #13760 completed with status: 'FAILURE'

@PerkzZheng
Copy link
Collaborator Author

/bot run

@tensorrt-cicd
Copy link
Collaborator

PR_Github #18510 [ run ] triggered by Bot

@tensorrt-cicd
Copy link
Collaborator

PR_Github #18510 [ run ] completed with state SUCCESS
/LLM/main/L0_MergeRequest_PR pipeline #13895 completed with status: 'SUCCESS'
Pipeline passed with automatic retried tests. Check the rerun report for details.

@PerkzZheng PerkzZheng merged commit 1b29c2e into NVIDIA:main Sep 14, 2025
5 checks passed
Wong4j pushed a commit to Wong4j/TensorRT-LLM that referenced this pull request Sep 20, 2025
MrGeva pushed a commit to nv-auto-deploy/TensorRT-LLM that referenced this pull request Sep 21, 2025
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

TRTLLM-Gen update updates the trtllm-gen kernels (for blackwell).

Projects

None yet

Development

Successfully merging this pull request may close these issues.

4 participants