[https://nvbugs/5549111][fix] Fix 2-model overlap scheduler accuracy on very long prompts #8076
Conversation
|
/bot run --disable-fail-fast |
📝 WalkthroughWalkthroughUpdates speculative decoding control flow to disable it when no draft batch is available, resetting flags and clearing draft tokens. Adds a new long-prompt Eagle3 speculative decoding test for Llama-3.1-8B with optional CUDA graph; the same test appears twice in the file. Changes
Sequence Diagram(s)sequenceDiagram
autonumber
participant C as Client
participant PE as PyExecutor
participant ME as ModelEngine
participant DR as Draft Runner
C->>PE: generate(requests)
PE->>DR: produce_draft_batch()
alt Draft batch available
DR-->>PE: draft_batch
PE->>ME: enable_spec_decode = true
PE->>ME: run_with_drafts(draft_batch)
else No draft batch
DR-->>PE: none
note over PE: has_previous_draft_tokens = false
rect rgba(230,240,255,0.6)
PE->>PE: use_spec_decode = false
PE->>ME: enable_spec_decode = false
PE->>PE: clear scheduled requests' py_draft_tokens = []
end
PE->>ME: run_without_drafts()
end
ME-->>PE: tokens
PE-->>C: output
Estimated code review effort🎯 3 (Moderate) | ⏱️ ~25 minutes Pre-merge checks and finishing touches❌ Failed checks (1 warning)
✅ Passed checks (2 passed)
✨ Finishing touches
🧪 Generate unit tests
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. 🧪 Early access (Sonnet 4.5): enabledWe are currently testing the Sonnet 4.5 model, which is expected to improve code review quality. However, this model may lead to increased noise levels in the review comments. Please disable the early access features if the noise level causes any inconvenience. Note:
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: 1
📜 Review details
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro
📒 Files selected for processing (2)
tensorrt_llm/_torch/pyexecutor/py_executor.py(1 hunks)tests/unittest/_torch/speculative/test_eagle3.py(1 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:
tensorrt_llm/_torch/pyexecutor/py_executor.pytests/unittest/_torch/speculative/test_eagle3.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:
tensorrt_llm/_torch/pyexecutor/py_executor.pytests/unittest/_torch/speculative/test_eagle3.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:
tensorrt_llm/_torch/pyexecutor/py_executor.pytests/unittest/_torch/speculative/test_eagle3.py
🧠 Learnings (1)
📚 Learning: 2025-08-18T08:42:02.640Z
Learnt from: samuellees
PR: NVIDIA/TensorRT-LLM#6974
File: tensorrt_llm/serve/scripts/benchmark_dataset.py:558-566
Timestamp: 2025-08-18T08:42:02.640Z
Learning: In TensorRT-LLM's RandomDataset (tensorrt_llm/serve/scripts/benchmark_dataset.py), when using --random-token-ids option, sequence length accuracy is prioritized over semantic correctness for benchmarking purposes. The encode/decode operations should use skip_special_tokens=True and add_special_tokens=False to ensure exact target token lengths.
Applied to files:
tensorrt_llm/_torch/pyexecutor/py_executor.py
🧬 Code graph analysis (2)
tensorrt_llm/_torch/pyexecutor/py_executor.py (2)
tensorrt_llm/_torch/pyexecutor/cuda_graph_runner.py (1)
enable_spec_decode(90-91)tensorrt_llm/_torch/pyexecutor/scheduler.py (1)
all_requests(38-39)
tests/unittest/_torch/speculative/test_eagle3.py (3)
tests/unittest/_torch/modeling/test_modeling_qwen2_5vl.py (1)
llm_models_root(30-36)tensorrt_llm/llmapi/llm_args.py (3)
EagleDecodingConfig(449-562)speculative_model_dir(1520-1521)CudaGraphConfig(109-166)tensorrt_llm/llmapi/llm.py (4)
LLM(1052-1068)prompt(82-83)generate(241-319)shutdown(737-744)
⏰ 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 (2)
tensorrt_llm/_torch/pyexecutor/py_executor.py (1)
2107-2112: LGTM! Clean corner-case fix for skipped speculation.The logic correctly disables speculative decoding when no draft batch is available, preventing ModelEngine from mishandling requests with stale draft tokens. The comment clearly explains the rationale, and the implementation properly resets both the executor and engine flags while clearing draft tokens from all scheduled requests.
tests/unittest/_torch/speculative/test_eagle3.py (1)
179-227: Well-structured test for the long-prompt corner case.The test correctly exercises the scenario where sequence length exceeds the draft model's max, forcing speculation to be skipped. The exact-match assertion is appropriate since dynamic disabling should produce identical output to non-speculative generation. Proper memory checks and cleanup are in place.
|
PR_Github #20298 [ run ] triggered by Bot |
mikeiovine
force-pushed
the
fix-overlap
branch
from
a831569 to
c316cc6
Compare
|
/bot run --disable-fail-fast |
mikeiovine
force-pushed
the
fix-overlap
branch
from
c316cc6 to
b9961f5
Compare
mikeiovine
requested review from
laikhtewari,
nv-guomingz,
poweiw and
yuanjingx87
mikeiovine
force-pushed
the
fix-overlap
branch
from
b9961f5 to
a4c30b2
Compare
mikeiovine
removed request for
a team,
laikhtewari,
dongxuy04,
nv-guomingz,
poweiw and
yuanjingx87
|
/bot run --disable-fail-fast |
|
PR_Github #20299 [ run ] triggered by Bot |
|
PR_Github #20298 [ run ] completed with state |
|
PR_Github #20300 [ run ] triggered by Bot |
|
PR_Github #20299 [ run ] completed with state |
|
PR_Github #20300 [ run ] completed with state |
|
/bot run --disable-fail-fast |
|
PR_Github #20386 [ run ] triggered by Bot |
|
PR_Github #20386 [ run ] completed with state |
mikeiovine
force-pushed
the
fix-overlap
branch
from
37c5fc8 to
468259b
Compare
|
/bot run --disable-fail-fast |
Signed-off-by: Mike Iovine <6158008+mikeiovine@users.noreply.github.com> Signed-off-by: Michael Iovine <miovine@nvidia.com> Signed-off-by: Mike Iovine <6158008+mikeiovine@users.noreply.github.com>
mikeiovine
force-pushed
the
fix-overlap
branch
from
468259b to
11966ce
Compare
|
/bot run --disable-fail-fast |
|
PR_Github #20537 [ run ] triggered by Bot |
|
PR_Github #20538 [ run ] triggered by Bot |
|
PR_Github #20537 [ run ] completed with state |
Description
Handle the corner case where speculation is skipped (e.g. because the sequence length went past the draft model's max seq len for every request in the batch). You need to remove all draft tokens and turn speculation off or the
ModelEnginewill not handle the requests correctly and garbage will be generated.Test Coverage
Added a new test.
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.
Summary by CodeRabbit
Bug Fixes
Tests