[None][doc] Add qwen3-next doc into deployment guide and test case into L0. #8288
Conversation
nv-guomingz
force-pushed
the
user/guomingz/cherry-pick-8195
branch
from
a2757b0 to
c23c27b
Compare
|
/bot run |
📝 WalkthroughWalkthroughAdds documentation for deploying Qwen3-Next on TensorRT-LLM and updates the deployment guide index. Updates supported models docs. Adds accuracy reference entries for Qwen3-Next-80B-A3B-Thinking. Introduces a new PyTorch accuracy test class and updates DGX H100/B200 test lists, including new disaggregated serving tests and some e2e test removals. Changes
Estimated code review effort🎯 3 (Moderate) | ⏱️ ~20 minutes Pre-merge checks and finishing touches❌ Failed checks (2 warnings)
✅ Passed checks (1 passed)
✨ Finishing touches
🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
📝 WalkthroughWalkthroughAdds a new TensorRT-LLM quick-start guide for Qwen3-Next, updates supported models docs, introduces accuracy reference entries for GSM8K and MMLU, adds an integration test class for Qwen3-Next-80B-A3B-Thinking with auto-dtype and CUDA Graph/KV cache configs, and adjusts test list YAMLs to include the new test and reorganize entries. Changes
Estimated code review effort🎯 3 (Moderate) | ⏱️ ~25 minutes Pre-merge checks and finishing touches❌ Failed checks (2 warnings, 1 inconclusive)
✨ Finishing touches
🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 2
🧹 Nitpick comments (1)
docs/source/deployment-guide/quick-start-recipe-for-qwen3-next-on-trtllm.md (1)
86-89: Clarify the relationship between command-line and YAML parameters.The documentation for
--kv_cache_free_gpu_memory_fractiondoesn't clarify how it relates tokv_cache_config.free_gpu_memory_fractionin the YAML file (line 50). If both are set, which takes precedence?Consider adding a note about parameter precedence or consolidating all KV cache settings in the YAML file to avoid confusion.
📜 Review details
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro
📒 Files selected for processing (7)
docs/source/deployment-guide/quick-start-recipe-for-qwen3-next-on-trtllm.md(1 hunks)docs/source/models/supported-models.md(2 hunks)tests/integration/defs/accuracy/references/gsm8k.yaml(1 hunks)tests/integration/defs/accuracy/references/mmlu.yaml(1 hunks)tests/integration/defs/accuracy/test_llm_api_pytorch.py(1 hunks)tests/integration/test_lists/test-db/l0_dgx_b200.yml(1 hunks)tests/integration/test_lists/test-db/l0_dgx_h100.yml(3 hunks)
🧰 Additional context used
📓 Path-based instructions (3)
**/*.{h,hpp,hh,hxx,cpp,cxx,cc,cu,cuh,py}
📄 CodeRabbit inference engine (CODING_GUIDELINES.md)
Use only spaces, no tabs; indent with 4 spaces.
Files:
tests/integration/defs/accuracy/test_llm_api_pytorch.py
**/*.py
📄 CodeRabbit inference engine (CODING_GUIDELINES.md)
**/*.py: Python code must target Python 3.8+.
Indent Python code with 4 spaces; do not use tabs.
Maintain module namespace when importing; prefer 'from package.subpackage import foo' then 'foo.SomeClass()' instead of importing the class directly.
Python filenames should be snake_case (e.g., some_file.py).
Python classes use PascalCase names.
Functions and methods use snake_case names.
Local variables use snake_case; prefix 'k' for variables that start with a number (e.g., k_99th_percentile).
Global variables use upper SNAKE_CASE prefixed with 'G' (e.g., G_MY_GLOBAL).
Constants use upper SNAKE_CASE (e.g., MY_CONSTANT).
Avoid shadowing variables from an outer scope.
Initialize all externally visible members of a class in the constructor.
Prefer docstrings for interfaces that may be used outside a file; comments for in-function or file-local interfaces.
Use Google-style docstrings for classes and functions (Sphinx-parsable).
Document attributes and variables inline so they render under the class/function docstring.
Avoid reflection when a simpler, explicit approach suffices (e.g., avoid dict(**locals()) patterns).
In try/except, catch the most specific exceptions possible.
For duck-typing try/except, keep the try body minimal and use else for the main logic.
Files:
tests/integration/defs/accuracy/test_llm_api_pytorch.py
**/*.{cpp,cxx,cc,h,hpp,hh,hxx,cu,cuh,py}
📄 CodeRabbit inference engine (CODING_GUIDELINES.md)
Prepend the NVIDIA Apache-2.0 copyright header with current year to the top of all source files (e.g., .cpp, .h, .cu, .py).
Files:
tests/integration/defs/accuracy/test_llm_api_pytorch.py
🧠 Learnings (2)
📚 Learning: 2025-07-28T17:06:08.621Z
Learnt from: moraxu
PR: NVIDIA/TensorRT-LLM#6303
File: tests/integration/test_lists/qa/examples_test_list.txt:494-494
Timestamp: 2025-07-28T17:06:08.621Z
Learning: In TensorRT-LLM testing, it's common to have both CLI flow tests (test_cli_flow.py) and PyTorch API tests (test_llm_api_pytorch.py) for the same model. These serve different purposes: CLI flow tests validate the traditional command-line workflow, while PyTorch API tests validate the newer LLM API backend. Both are legitimate and should coexist.
Applied to files:
tests/integration/test_lists/test-db/l0_dgx_h100.yml
📚 Learning: 2025-09-09T09:40:45.658Z
Learnt from: fredricz-20070104
PR: NVIDIA/TensorRT-LLM#7645
File: tests/integration/test_lists/qa/llm_function_core.txt:648-648
Timestamp: 2025-09-09T09:40:45.658Z
Learning: In TensorRT-LLM test lists, it's common and intentional for the same test to appear in multiple test list files when they serve different purposes (e.g., llm_function_core.txt for comprehensive core functionality testing and llm_function_core_sanity.txt for quick sanity checks). This duplication allows tests to be run in different testing contexts.
Applied to files:
tests/integration/test_lists/test-db/l0_dgx_h100.yml
🧬 Code graph analysis (1)
tests/integration/defs/accuracy/test_llm_api_pytorch.py (3)
tests/integration/defs/accuracy/accuracy_core.py (5)
LlmapiAccuracyTestHarness(844-855)MMLU(315-329)evaluate(184-245)evaluate(763-773)GSM8K(332-347)tests/integration/defs/conftest.py (2)
llm_models_root(79-93)get_device_count(1979-1981)tensorrt_llm/llmapi/llm_args.py (2)
KvCacheConfig(1106-1240)CudaGraphConfig(109-166)
⏰ Context from checks skipped due to timeout of 90000ms. You can increase the timeout in your CodeRabbit configuration to a maximum of 15 minutes (900000ms). (1)
- GitHub Check: Pre-commit Check
🔇 Additional comments (10)
docs/source/models/supported-models.md (2)
26-26: LGTM! Model entry follows the existing pattern.The new Qwen3NextForCausalLM entry is properly formatted and includes the correct HuggingFace model reference that matches the test and documentation additions elsewhere in the PR.
37-37: Verify feature support limitations are intentional.All features in the support matrix are marked as "No" for Qwen3NextForCausalLM, which indicates very limited functionality compared to other models like Qwen3MoeForCausalLM (line 36) which has extensive feature support. Please confirm this is intentional for the initial release.
Based on the deployment guide (docs/source/deployment-guide/quick-start-recipe-for-qwen3-next-on-trtllm.md lines 41-49), the model does use CUDA Graph and KV cache configurations, which suggests some features may be supported but not documented here yet.
tests/integration/defs/accuracy/references/gsm8k.yaml (1)
126-127: LGTM! GSM8K accuracy reference properly added.The new accuracy reference for Qwen3-Next-80B-A3B-Thinking is correctly formatted and will be used by the test in tests/integration/defs/accuracy/test_llm_api_pytorch.py (line 3555-3556).
tests/integration/defs/accuracy/references/mmlu.yaml (1)
232-233: LGTM! MMLU accuracy reference properly added.The new accuracy reference for Qwen3-Next-80B-A3B-Thinking is correctly formatted and will be used by the test in tests/integration/defs/accuracy/test_llm_api_pytorch.py (line 3553-3554).
tests/integration/test_lists/test-db/l0_dgx_b200.yml (1)
42-42: LGTM! Test entry correctly added to B200 test list.The new test entry for Qwen3NextThinking is properly placed in the 4-GPU pre-merge block and follows the established naming convention. It correctly references the test method added in tests/integration/defs/accuracy/test_llm_api_pytorch.py (lines 3537-3556).
tests/integration/test_lists/test-db/l0_dgx_h100.yml (2)
22-22: LGTM! Test organization improved with section headers.Adding "Disaggregated serving tests" headers improves the readability of the test list by clearly delineating different test categories.
Also applies to: 76-76
74-75: Verify if these test additions are related to the Qwen3-Next PR.These test_e2e.py additions appear unrelated to the Qwen3-Next model support being added in this PR. If they are incidental changes that were already in the branch, consider whether they should be in a separate PR for cleaner history.
tests/integration/defs/accuracy/test_llm_api_pytorch.py (3)
3528-3532: LGTM! Test class follows established patterns.The new test class is properly structured following the LlmapiAccuracyTestHarness pattern used by other model tests in this file. The MODEL_NAME and MODEL_PATH setup is consistent with the model reference added in other files.
3533-3556: Verify test configuration aligns with deployment guide.The test configuration looks correct, but please verify the following alignments with the deployment guide:
max_batch_size mismatch: Test uses 720 (line 3544), deployment guide uses 16 (docs/source/deployment-guide/quick-start-recipe-for-qwen3-next-on-trtllm.md line 63). Is this intentional for testing purposes?
Missing moe_config: The deployment guide specifies
moe_config: backend: TRTLLM(line 45), but the test doesn't explicitly set this. Will it default appropriately?stream_interval: The deployment guide sets
stream_interval: 20(line 46) but the test doesn't configure this parameter.The core configuration (KV cache with
enable_block_reuse=False, CUDA Graph with padding enabled, 4-way TP and EP) correctly matches the deployment guide's recommendations.
3538-3539: Device count validation is good practice.The explicit device count check prevents the test from running with incorrect parallelism configurations, which would lead to confusing failures.
docs/source/deployment-guide/quick-start-recipe-for-qwen3-next-on-trtllm.md
Outdated
Show resolved
Hide resolved
docs/source/deployment-guide/quick-start-recipe-for-qwen3-next-on-trtllm.md
Show resolved
Hide resolved
|
PR_Github #21079 [ run ] triggered by Bot |
|
PR_Github #21079 [ run ] completed with state |
|
/bot run |
|
PR_Github #21082 [ run ] triggered by Bot |
|
PR_Github #21082 [ run ] completed with state |
nv-guomingz
force-pushed
the
user/guomingz/cherry-pick-8195
branch
from
c23c27b to
b0339df
Compare
There was a problem hiding this comment.
Actionable comments posted: 0
♻️ Duplicate comments (2)
docs/source/deployment-guide/quick-start-recipe-for-qwen3-next-on-trtllm.md (2)
34-34: Align prose withenable_block_reuseThe text still tells readers to flip
kv_cache_reuse, but the actual server knob isenable_block_reuse. Please name the correct option so the instructions match the YAML and implementation.-Note that we should set kv_cache_reuse to false. +Note that we should set `enable_block_reuse` to false to disable KV cache reuse for this recipe.
221-222: Remove attention-DP tuning adviceAttention data parallelism remains unsupported for Qwen3NextForCausalLM (see the supported-models matrix), so the throughput tip shouldn’t instruct users to sweep “with attention DP on.” Please reword or drop that clause.
-To achieve max through-put, with attention DP on, one needs to sweep up to `concurrency = max_batch_size * num_gpus`. +To achieve max throughput, sweep concurrency up to `concurrency = max_batch_size * num_gpus`; attention data parallelism is unsupported for this model today.
🧹 Nitpick comments (1)
docs/source/deployment-guide/quick-start-recipe-for-qwen3-next-on-trtllm.md (1)
24-30: Annotate fenced blocks with languagesAdding the language hints keeps markdownlint happy and improves highlighting—
shellfor the Docker commands andjsonfor the sample response.-``` +```shell cd TensorRT-LLM make -C docker release_build IMAGE_TAG=qwen3-next-local make -C docker release_run IMAGE_NAME=tensorrt_llm IMAGE_TAG=qwen3-next-local LOCAL_USER=1-``` +```json {"id":"chatcmpl-64ac201c77bf46a7a3a4eca7759b1fd8","object":"chat.completion","created":1759022940,"model":"Qwen/Qwen3-Next-80B-A3B-Thinking",...}Also applies to: 172-174
📜 Review details
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro
📒 Files selected for processing (8)
docs/source/deployment-guide/index.rst(1 hunks)docs/source/deployment-guide/quick-start-recipe-for-qwen3-next-on-trtllm.md(1 hunks)docs/source/models/supported-models.md(2 hunks)tests/integration/defs/accuracy/references/gsm8k.yaml(1 hunks)tests/integration/defs/accuracy/references/mmlu.yaml(1 hunks)tests/integration/defs/accuracy/test_llm_api_pytorch.py(1 hunks)tests/integration/test_lists/test-db/l0_dgx_b200.yml(1 hunks)tests/integration/test_lists/test-db/l0_dgx_h100.yml(3 hunks)
🧰 Additional context used
📓 Path-based instructions (3)
**/*.{h,hpp,hh,hxx,cpp,cxx,cc,cu,cuh,py}
📄 CodeRabbit inference engine (CODING_GUIDELINES.md)
Use only spaces, no tabs; indent with 4 spaces.
Files:
tests/integration/defs/accuracy/test_llm_api_pytorch.py
**/*.py
📄 CodeRabbit inference engine (CODING_GUIDELINES.md)
**/*.py: Python code must target Python 3.8+.
Indent Python code with 4 spaces; do not use tabs.
Maintain module namespace when importing; prefer 'from package.subpackage import foo' then 'foo.SomeClass()' instead of importing the class directly.
Python filenames should be snake_case (e.g., some_file.py).
Python classes use PascalCase names.
Functions and methods use snake_case names.
Local variables use snake_case; prefix 'k' for variables that start with a number (e.g., k_99th_percentile).
Global variables use upper SNAKE_CASE prefixed with 'G' (e.g., G_MY_GLOBAL).
Constants use upper SNAKE_CASE (e.g., MY_CONSTANT).
Avoid shadowing variables from an outer scope.
Initialize all externally visible members of a class in the constructor.
Prefer docstrings for interfaces that may be used outside a file; comments for in-function or file-local interfaces.
Use Google-style docstrings for classes and functions (Sphinx-parsable).
Document attributes and variables inline so they render under the class/function docstring.
Avoid reflection when a simpler, explicit approach suffices (e.g., avoid dict(**locals()) patterns).
In try/except, catch the most specific exceptions possible.
For duck-typing try/except, keep the try body minimal and use else for the main logic.
Files:
tests/integration/defs/accuracy/test_llm_api_pytorch.py
**/*.{cpp,cxx,cc,h,hpp,hh,hxx,cu,cuh,py}
📄 CodeRabbit inference engine (CODING_GUIDELINES.md)
Prepend the NVIDIA Apache-2.0 copyright header with current year to the top of all source files (e.g., .cpp, .h, .cu, .py).
Files:
tests/integration/defs/accuracy/test_llm_api_pytorch.py
🧠 Learnings (2)
📚 Learning: 2025-07-28T17:06:08.621Z
Learnt from: moraxu
PR: NVIDIA/TensorRT-LLM#6303
File: tests/integration/test_lists/qa/examples_test_list.txt:494-494
Timestamp: 2025-07-28T17:06:08.621Z
Learning: In TensorRT-LLM testing, it's common to have both CLI flow tests (test_cli_flow.py) and PyTorch API tests (test_llm_api_pytorch.py) for the same model. These serve different purposes: CLI flow tests validate the traditional command-line workflow, while PyTorch API tests validate the newer LLM API backend. Both are legitimate and should coexist.
Applied to files:
tests/integration/test_lists/test-db/l0_dgx_h100.yml
📚 Learning: 2025-09-09T09:40:45.658Z
Learnt from: fredricz-20070104
PR: NVIDIA/TensorRT-LLM#7645
File: tests/integration/test_lists/qa/llm_function_core.txt:648-648
Timestamp: 2025-09-09T09:40:45.658Z
Learning: In TensorRT-LLM test lists, it's common and intentional for the same test to appear in multiple test list files when they serve different purposes (e.g., llm_function_core.txt for comprehensive core functionality testing and llm_function_core_sanity.txt for quick sanity checks). This duplication allows tests to be run in different testing contexts.
Applied to files:
tests/integration/test_lists/test-db/l0_dgx_h100.yml
🧬 Code graph analysis (1)
tests/integration/defs/accuracy/test_llm_api_pytorch.py (4)
tests/integration/defs/accuracy/accuracy_core.py (5)
LlmapiAccuracyTestHarness(844-855)MMLU(315-329)evaluate(184-245)evaluate(763-773)GSM8K(332-347)tests/integration/defs/conftest.py (2)
llm_models_root(79-93)get_device_count(1979-1981)tensorrt_llm/llmapi/llm_args.py (2)
KvCacheConfig(1106-1240)CudaGraphConfig(109-166)tensorrt_llm/llmapi/llm.py (1)
LLM(1084-1100)
🪛 markdownlint-cli2 (0.18.1)
docs/source/deployment-guide/quick-start-recipe-for-qwen3-next-on-trtllm.md
24-24: Fenced code blocks should have a language specified
(MD040, fenced-code-language)
172-172: Fenced code blocks should have a language specified
(MD040, fenced-code-language)
⏰ Context from checks skipped due to timeout of 90000ms. You can increase the timeout in your CodeRabbit configuration to a maximum of 15 minutes (900000ms). (1)
- GitHub Check: Pre-commit Check
nv-guomingz
force-pushed
the
user/guomingz/cherry-pick-8195
branch
from
b0339df to
f5e0dc5
Compare
Signed-off-by: Faradawn Yang <[email protected]> Signed-off-by: Robin Kobus <[email protected]> Signed-off-by: nv-guomingz <[email protected]>
nv-guomingz
force-pushed
the
user/guomingz/cherry-pick-8195
branch
from
f5e0dc5 to
54e73bd
Compare
|
/bot reuse-pipeline |
|
PR_Github #21100 [ reuse-pipeline ] triggered by Bot |
|
PR_Github #21100 [ reuse-pipeline ] completed with state |
docs/source/deployment-guide/quick-start-recipe-for-qwen3-next-on-trtllm.md
Show resolved
Hide resolved
nv-guomingz
changed the title
4 participants
Summary by CodeRabbit
Documentation
Tests
It's a cherry-pick PR of #8195
Description
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.
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 thestage-listparameter 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.mdand the
scripts/test_to_stage_mapping.pyhelper.kill
killKill all running builds associated with pull request.
skip
skip --comment COMMENTSkip 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-pipelineReuse 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.