[None][fix] fix Llama3 eagle3 test case OOM #6832
Conversation
📝 WalkthroughWalkthroughReplace standalone Changes
Sequence Diagram(s)sequenceDiagram
participant PyTest as PyTest
participant LLM as LLM(API)
participant CUDA as CUDA Graphs
participant Eval as CNN/DailyMail Eval
PyTest->>LLM: Init(model=...-fp8, tensor_parallel_size=8, max_batch_size=16, speculative_config=Eagle, kv_cache_config=...)
LLM->>CUDA: cuda_graph_config(max_batch_size=1)
PyTest->>LLM: Run generate/eval
LLM-->>Eval: Outputs for evaluation
Eval-->>PyTest: Accuracy compared to FP8 reference
Estimated code review effort🎯 2 (Simple) | ⏱️ ~8 minutes Possibly related PRs
Suggested labels
Suggested reviewers
✨ 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. 🪧 TipsChatThere are 3 ways to chat with CodeRabbit:
SupportNeed help? Create a ticket on our support page for assistance with any issues or questions. CodeRabbit Commands (Invoked using PR/Issue comments)Type Other keywords and placeholders
Status, Documentation and Community
|
|
/bot run |
|
PR_Github #14988 [ run ] triggered by Bot |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 0
🧹 Nitpick comments (2)
tests/integration/defs/accuracy/test_llm_api_pytorch.py (2)
396-404: Add an explicit FP8 assertion to catch misconfigurations early.Helps fail-fast if the model path doesn’t resolve to FP8 quantization.
Apply within this block:
with LLM(model_path, max_batch_size=16, tensor_parallel_size=8, speculative_config=spec_config, kv_cache_config=kv_cache_config, **pytorch_config) as llm: + assert llm.args.quant_config.quant_algo == QuantAlgo.FP8 task = CnnDailymail(self.MODEL_NAME) task.evaluate(llm)
392-404: If OOM persists in CI, consider tightening memory knobs.If this still trips OOM on certain runners, consider one or more:
- Reduce free_gpu_memory_fraction to 0.5.
- Lower max_batch_size to 8.
- Explicitly bound max_seq_len (e.g., 8192) to cap warmup allocations.
Happy to update the test with a guarded config tweak for CI-only paths if desired.
- kv_cache_config = KvCacheConfig(free_gpu_memory_fraction=0.6) + kv_cache_config = KvCacheConfig(free_gpu_memory_fraction=0.5) @@ -with LLM(model_path, - max_batch_size=16, +with LLM(model_path, + max_batch_size=8, + max_seq_len=8192, tensor_parallel_size=8, speculative_config=spec_config, kv_cache_config=kv_cache_config, **pytorch_config) as llm:
📜 Review details
Configuration used: .coderabbit.yaml
Review profile: CHILL
Plan: Pro
📒 Files selected for processing (5)
tests/integration/defs/accuracy/references/cnn_dailymail.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/qa/llm_function_full.txt(1 hunks)tests/integration/test_lists/qa/llm_function_sanity.txt(1 hunks)
🧰 Additional context used
📓 Path-based instructions (2)
**/*.py
📄 CodeRabbit Inference Engine (CODING_GUIDELINES.md)
**/*.py: Python code must target Python 3.8+
Python indentation: 4 spaces, no tabs
Maintain module namespace in imports (from package.subpackage import foo; then use foo.SomeClass())
Python file names use snake_case
Python class names use PascalCase
Python functions/methods and local variables use snake_case; variables starting with a number get k_ prefix (e.g., k_99th_percentile)
Global variables use G_ prefixed UPPER_SNAKE_CASE (e.g., G_MY_GLOBAL)
Constants use UPPER_SNAKE_CASE in Python
Avoid shadowing variables from outer scopes in Python
Initialize all externally visible members of a Python class in init
Prefer docstrings for interfaces used outside a file; comments for local code
Use Google-style docstrings for classes and functions (Sphinx-parsable)
Document attributes/variables inline with short docstrings
Avoid reflection when simple alternatives exist (e.g., prefer explicit parameters over dict(**locals()))
In try/except, catch the narrowest exceptions possible
For duck-typing with try/except, keep try body minimal and put logic in else
Files:
tests/integration/defs/accuracy/test_llm_api_pytorch.py
**/*.{cpp,cxx,cc,cu,h,hpp,hxx,hh,cuh,py}
📄 CodeRabbit Inference Engine (CODING_GUIDELINES.md)
Prepend NVIDIA copyright header (current year) to all source files
Files:
tests/integration/defs/accuracy/test_llm_api_pytorch.py
🧠 Learnings (1)
📚 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/qa/llm_function_full.txttests/integration/test_lists/qa/llm_function_sanity.txt
🧬 Code Graph Analysis (1)
tests/integration/defs/accuracy/test_llm_api_pytorch.py (3)
tests/integration/defs/conftest.py (2)
parametrize_with_ids(1786-1811)llm_models_root(77-83)tensorrt_llm/llmapi/llm_args.py (4)
KvCacheConfig(923-1002)EagleDecodingConfig(406-432)speculative_model_dir(1321-1322)CudaGraphConfig(106-163)tensorrt_llm/llmapi/llm.py (1)
LLM(1111-1127)
⏰ 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 (5)
tests/integration/test_lists/qa/llm_function_full.txt (1)
453-454: All stale references removed andtest_fp8_eagle3_tp8confirmed
- No occurrences of the old name
test_eagle3_tp8[…]remain.- New test definition found in
tests/integration/defs/accuracy/test_llm_api_pytorch.pyat line 389.- QA listing correctly shows
test_fp8_eagle3_tp8[eagle3_one_model=True]and[…False].tests/integration/defs/accuracy/references/cnn_dailymail.yaml (1)
175-176: Accuracy reference schema adjustment is correct.Placing spec_dec_algo: Eagle under a quant_algo: FP8 bullet is consistent with other entries in this file. No issues.
tests/integration/defs/accuracy/references/mmlu.yaml (1)
62-63: MMLU reference updated to FP8+Eagle combo is coherent.The structure mirrors similar patterns (e.g., Llama-4-Maverick), and accuracy unchanged as expected. Looks good.
tests/integration/test_lists/qa/llm_function_sanity.txt (1)
69-70: Sanity and Full QA Lists Updated ConsistentlyThe renamed test
test_fp8_eagle3_tp8is now correctly referenced—with botheagle3_one_model=TrueandFalseparameterizations—inllm_function_sanity.txt(lines 69–70) andllm_function_full.txt(lines 453–454). The oldtest_eagle3_tp8entries have been fully removed.tests/integration/defs/accuracy/test_llm_api_pytorch.py (1)
386-394: Good gating and scope to mitigate OOM risk.
- skip_pre_hopper + world_size(8) is appropriate for 70B + EAGLE3.
- Moving to prequantized FP8 model and dropping MMLU keeps memory down for this test’s purpose.
No issues here.
|
PR_Github #14988 [ run ] completed with state |
Signed-off-by: Ivy Zhang <[email protected]>
crazydemo
force-pushed
the
llama3_eagle3
branch
from
02411fa to
f05f837
Compare
|
/bot run --disable-fail-fast |
|
PR_Github #15052 [ run ] triggered by Bot |
There was a problem hiding this comment.
Actionable comments posted: 0
🧹 Nitpick comments (3)
tests/integration/defs/accuracy/test_llm_api_pytorch.py (3)
390-390: Switch to FP8 model path is appropriate; assert quant algo for safetyTo harden the test against model mix-ups, add an assertion that the loaded model is FP8.
Apply:
with LLM(model_path, max_batch_size=16, tensor_parallel_size=8, speculative_config=spec_config, kv_cache_config=kv_cache_config, **pytorch_config) as llm: + assert llm.args.quant_config.quant_algo == QuantAlgo.FP8 task = CnnDailymail(self.MODEL_NAME) task.evaluate(llm)
396-398: CUDA graph capture: consider enabling padding to reduce shape-churnEnabling padding typically stabilizes captures and reduces graph proliferation without affecting batch size=1 cases.
Apply:
- pytorch_config = dict( - disable_overlap_scheduler=True, - cuda_graph_config=CudaGraphConfig(max_batch_size=1)) + pytorch_config = dict( + disable_overlap_scheduler=True, + cuda_graph_config=CudaGraphConfig(max_batch_size=1, enable_padding=True))
400-400: OOM guardrails: consider capping max_seq_len and/or lowering max_batch_sizeIf OOM persists in CI for this 70B TP8 path, two options have been effective elsewhere in this file:
- Set max_seq_len=8192 to keep warmup allocations bounded.
- Reduce max_batch_size to 8.
Option A (preferred: cap warmup seq length):
with LLM(model_path, - max_batch_size=16, + max_batch_size=16, + max_seq_len=8192, tensor_parallel_size=8, speculative_config=spec_config, kv_cache_config=kv_cache_config, **pytorch_config) as llm:Option B (if memory is still tight):
- max_batch_size=16, + max_batch_size=8,Please re-run the failing pipeline stage for this test after applying either option to confirm the OOM is resolved.
📜 Review details
Configuration used: .coderabbit.yaml
Review profile: CHILL
Plan: Pro
📒 Files selected for processing (5)
tests/integration/defs/accuracy/references/cnn_dailymail.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/qa/llm_function_full.txt(1 hunks)tests/integration/test_lists/qa/llm_function_sanity.txt(1 hunks)
🚧 Files skipped from review as they are similar to previous changes (4)
- tests/integration/test_lists/qa/llm_function_full.txt
- tests/integration/defs/accuracy/references/cnn_dailymail.yaml
- tests/integration/test_lists/qa/llm_function_sanity.txt
- tests/integration/defs/accuracy/references/mmlu.yaml
🧰 Additional context used
📓 Path-based instructions (2)
**/*.py
📄 CodeRabbit Inference Engine (CODING_GUIDELINES.md)
**/*.py: Python code must target Python 3.8+
Python indentation: 4 spaces, no tabs
Maintain module namespace in imports (from package.subpackage import foo; then use foo.SomeClass())
Python file names use snake_case
Python class names use PascalCase
Python functions/methods and local variables use snake_case; variables starting with a number get k_ prefix (e.g., k_99th_percentile)
Global variables use G_ prefixed UPPER_SNAKE_CASE (e.g., G_MY_GLOBAL)
Constants use UPPER_SNAKE_CASE in Python
Avoid shadowing variables from outer scopes in Python
Initialize all externally visible members of a Python class in init
Prefer docstrings for interfaces used outside a file; comments for local code
Use Google-style docstrings for classes and functions (Sphinx-parsable)
Document attributes/variables inline with short docstrings
Avoid reflection when simple alternatives exist (e.g., prefer explicit parameters over dict(**locals()))
In try/except, catch the narrowest exceptions possible
For duck-typing with try/except, keep try body minimal and put logic in else
Files:
tests/integration/defs/accuracy/test_llm_api_pytorch.py
**/*.{cpp,cxx,cc,cu,h,hpp,hxx,hh,cuh,py}
📄 CodeRabbit Inference Engine (CODING_GUIDELINES.md)
Prepend NVIDIA copyright header (current year) to all source files
Files:
tests/integration/defs/accuracy/test_llm_api_pytorch.py
🧬 Code Graph Analysis (1)
tests/integration/defs/accuracy/test_llm_api_pytorch.py (3)
tests/integration/defs/conftest.py (2)
parametrize_with_ids(1786-1811)llm_models_root(77-83)tensorrt_llm/llmapi/llm_args.py (4)
KvCacheConfig(923-1002)EagleDecodingConfig(406-432)speculative_model_dir(1321-1322)CudaGraphConfig(106-163)tensorrt_llm/llmapi/llm.py (1)
LLM(1111-1127)
⏰ 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 (1)
tests/integration/defs/accuracy/test_llm_api_pytorch.py (1)
386-389: All test name references have been updatedVerification confirms no occurrences of the old name
test_eagle3_tp8and correct references totest_fp8_eagle3_tp8in the QA lists:
- tests/integration/test_lists/qa/llm_function_sanity.txt (lines 69–70)
- tests/integration/test_lists/qa/llm_function_full.txt (lines 453–454)
No further action required.
|
PR_Github #15052 [ run ] completed with state |
Signed-off-by: Ivy Zhang <[email protected]>
Signed-off-by: Ivy Zhang <[email protected]> Signed-off-by: Wangshanshan <[email protected]>
Signed-off-by: Ivy Zhang <[email protected]> Signed-off-by: Wangshanshan <[email protected]>
Signed-off-by: Ivy Zhang <[email protected]> Signed-off-by: Wangshanshan <[email protected]>
Signed-off-by: Ivy Zhang <[email protected]> Signed-off-by: Wangshanshan <[email protected]>
Signed-off-by: Ivy Zhang <[email protected]> Signed-off-by: Wangshanshan <[email protected]>
Signed-off-by: Ivy Zhang <[email protected]> Signed-off-by: Wangshanshan <[email protected]>
Signed-off-by: Ivy Zhang <[email protected]> Signed-off-by: Wangshanshan <[email protected]>
Signed-off-by: Ivy Zhang <[email protected]> Signed-off-by: Wangshanshan <[email protected]>
Signed-off-by: Ivy Zhang <[email protected]> Signed-off-by: Wangshanshan <[email protected]>
Signed-off-by: Ivy Zhang <[email protected]> Signed-off-by: Wangshanshan <[email protected]>
Signed-off-by: Ivy Zhang <[email protected]> Signed-off-by: Wangshanshan <[email protected]>
Signed-off-by: Ivy Zhang <[email protected]> Signed-off-by: Wangshanshan <[email protected]>
Signed-off-by: Ivy Zhang <[email protected]> Signed-off-by: Wangshanshan <[email protected]>
Signed-off-by: Ivy Zhang <[email protected]> Signed-off-by: Wangshanshan <[email protected]>
Signed-off-by: Ivy Zhang <[email protected]> Signed-off-by: Wangshanshan <[email protected]>
Signed-off-by: Ivy Zhang <[email protected]> Signed-off-by: Wangshanshan <[email protected]>
Signed-off-by: Ivy Zhang <[email protected]> Signed-off-by: Wangshanshan <[email protected]>
Signed-off-by: Ivy Zhang <[email protected]> Signed-off-by: Wangshanshan <[email protected]>
Signed-off-by: Ivy Zhang <[email protected]> Signed-off-by: Wangshanshan <[email protected]>
Signed-off-by: Ivy Zhang <[email protected]> Signed-off-by: Wangshanshan <[email protected]>
Signed-off-by: Ivy Zhang <[email protected]> Signed-off-by: Wangshanshan <[email protected]>
Signed-off-by: Ivy Zhang <[email protected]> Signed-off-by: Wangshanshan <[email protected]>
Signed-off-by: Ivy Zhang <[email protected]> Signed-off-by: Wangshanshan <[email protected]>
Signed-off-by: Ivy Zhang <[email protected]> Signed-off-by: Wangshanshan <[email protected]>
Summary by CodeRabbit
Tests
Chores
Description
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 [--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.