Cherry pick feat/llama4 to main

[nv-yilinf]

PR title

Cherry pick feat/llama4 to main

Description

This PR is based on @chenfeiz0326 's check-pick PR with new merge conflicts resolved. I made a new PR because I don't have push permission to his branch

Perf

===========================================================
= PERFORMANCE OVERVIEW
===========================================================
Request Throughput (req/sec):                     0.7471
Total Output Throughput (tokens/sec):             897.0175
Total Token Throughput (tokens/sec):              1679.6140
Total Latency (ms):                               42833.0554
Average request latency (ms):                     1338.4794
Per User Output Throughput [w/ ctx] (tps/user):   900.4043
Per GPU Output Throughput (tps/gpu):              112.1272
Average time-to-first-token [TTFT] (ms):          52.7782
Average time-per-output-token [TPOT] (ms):        1.0700
Per User Output Speed (tps/user):                 936.9597

-- Acceptance Rate Details --------------------------------

[AR] MINIMUM: 2.63
[AR] MAXIMUM: 3.18
[AR] AVERAGE: 2.99
[AR] P50    : 3.04
[AR] P90    : 3.15
[AR] P95    : 3.17
[AR] P99    : 3.18
===========================================================

Accuracy verification

MMLU weighted average accuracy: 85.91 (14042)

GPQA:

{'score:std': np.float64(0.4969039949999533), 'score:stderr': 0.03540294377095367, 'score': np.float64(0.5555555555555556), 'task_name': 'gpqa_diamond'}

Output tokens:

[0] Prompt: 'Hello, my name is', Generated text: ' {{ name }} and I am {{ age }} years old.\n    </body>\n</html>\n"""\n\n# Define the data to be used in the template\ndata = {\n    "name": "John Doe",\n    "age": 30\n}\n\n# Render the template with the data\nrendered_template'
[1] Prompt: 'The president of the United States is', Generated text: ' the head of state and head of government of the United States, and is also the commander-in-chief of the armed forces. The president is responsible for executing the laws and policies of the government, and is also responsible for representing the United States on the international stage. The president is elected by the people'
[2] Prompt: 'The capital of France is', Generated text: ' Paris, and it is known for its rich history, art, and culture. The city is home to many famous landmarks, including the Eiffel Tower, the Louvre Museum, and Notre-Dame Cathedral. Paris is also known for its fashion, cuisine, and romantic atmosphere, making it a popular destination'
[3] Prompt: 'The future of AI is', Generated text: ' a topic of much debate and speculation. As we continue to develop and refine AI technologies, what can we expect the future of AI to hold, and how will it impact various aspects of our lives and society as a whole?\n\n1. **Advancements in AI Capabilities**: The future of AI is expected to'

Test Coverage

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.

run [--disable-fail-fast --skip-test --stage-list "A10-1, xxx" --gpu-type "A30, H100_PCIe" --add-multi-gpu-test --only-multi-gpu-test --disable-multi-gpu-test --post-merge --extra-stage "H100_PCIe-[Post-Merge]-1, xxx"]

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

--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 "A10-1, xxx" (OPTIONAL) : Only run the specified test stages. Examples: "A10-1, 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.

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

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.

[nv-yilinf]

/bot run --stage-list A100X-PyTorch-1

[nv-yilinf]

/bot run --stage-list A100X-PyTorch-1

[tensorrt-cicd]

PR_Github #6800 [ run ] triggered by Bot

[nv-yilinf]

/bot kill

[hlu1]

/bot run --disable-fail-fast --add-multi-gpu-test

[tensorrt-cicd]

PR_Github #6801 [ run ] triggered by Bot

[tensorrt-cicd]

PR_Github #6800 [ run ] completed with state ABORTED

[tensorrt-cicd]

PR_Github #6802 [ kill ] triggered by Bot

[tensorrt-cicd]

PR_Github #6801 [ run ] completed with state ABORTED

[tensorrt-cicd]

PR_Github #6802 [ kill ] completed with state SUCCESS
Successfully killed previous jobs for commit f02b8a2

[nv-yilinf]

/bot run --disable-fail-fast --add-multi-gpu-test

[nv-yilinf]

/bot run --disable-fail-fast --add-multi-gpu-test

[nv-yilinf]

/bot run --disable-fail-fast --add-multi-gpu-test

[tensorrt-cicd]

PR_Github #6806 [ run ] triggered by Bot

[tensorrt-cicd]

PR_Github #6806 [ run ] completed with state SUCCESS
/LLM/main/L0_MergeRequest_PR pipeline #4960 completed with status: 'FAILURE'

[nv-yilinf]

/bot run --add-multi-gpu-test

[nv-yilinf]

/bot run --stage-list A10-TensorRT-1--add-multi-gpu-test

[tensorrt-cicd]

PR_Github #6821 [ run ] triggered by Bot

[tensorrt-cicd]

PR_Github #6822 [ run ] triggered by Bot

[tensorrt-cicd]

PR_Github #6916 [ run ] triggered by Bot

[nvpohanh]

https://prod.blsm.nvidia.com/sw-tensorrt-top-1/blue/organizations/jenkins/LLM%2Fmain%2FL0_Test-x86_64/detail/L0_Test-x86_64/12086/pipeline/242

4 out of 5 multi-gpu stages passed. and the failing one is caused by infra issue

[nvpohanh]

/bot run --only-multi-gpu-test --disable-fail-fast

[tensorrt-cicd]

PR_Github #6916 [ run ] completed with state SUCCESS
/LLM/main/L0_MergeRequest_PR pipeline #5021 (Partly Tested) completed with status: 'FAILURE'

[nv-yilinf]

/bot run --only-multi-gpu-test --disable-fail-fast

[tensorrt-cicd]

PR_Github #6936 [ run ] triggered by Bot

[tensorrt-cicd]

PR_Github #6936 [ run ] completed with state SUCCESS
/LLM/main/L0_MergeRequest_PR pipeline #5036 (Partly Tested) completed with status: 'FAILURE'

[nv-yilinf]

/bot run --only-multi-gpu-test --disable-fail-fast

[nv-yilinf]

/bot run --only-multi-gpu-test --disable-fail-fast

[tensorrt-cicd]

PR_Github #6950 [ run ] triggered by Bot

[tensorrt-cicd]

PR_Github #6950 [ run ] completed with state SUCCESS
/LLM/main/L0_MergeRequest_PR pipeline #5046 (Partly Tested) completed with status: 'SUCCESS'
Pipeline passed with automatic retried tests. Check the rerun report for details.

[hlu1]

/bot skip --comment "Combined CI passes"

[tensorrt-cicd]

PR_Github #6957 [ skip ] triggered by Bot

[tensorrt-cicd]

PR_Github #6957 [ skip ] completed with state SUCCESS
Skipping testing for commit e344f8d

[hlu1]

/bot run --post-merge

[hlu1]

/bot run --post-merge --disable-fail-fast

[tensorrt-cicd]

PR_Github #6959 [ ] completed with state FAILURE
Not allowed on merged PR

[tensorrt-cicd]

PR_Github #6960 [ ] completed with state FAILURE
Not allowed on merged PR

[SimengLiu-nv]

Created #4780 as a dup to trigger post_merge pipelines.