[https://nvbugs/5458798][fix] Relaxed test threshold, added documentation #6997
Conversation
Signed-off-by: Eran Geva <[email protected]>
📝 WalkthroughWalkthroughUpdate to a unit test: adjust memory expectation constants (extra_consumption_mb 2500→2700), change post-pass free-memory range computation to equal pre-pass, rename validation variable to memory_reduction and message text, and change default Changes
Estimated code review effort🎯 2 (Simple) | ⏱️ ~8 minutes Possibly related issues
Possibly related PRs
Suggested labels
Suggested reviewers
Tip 🔌 Remote MCP (Model Context Protocol) integration is now available!Pro plan users can now connect to remote MCP servers from the Integrations page. Connect with popular remote MCPs such as Notion and Linear to add more context to your reviews and chats. ✨ 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 |
![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 (3)
tests/unittest/_torch/auto_deploy/unit/singlegpu/test_ad_trtllm_bench.py (3)
282-286: Promote extra_consumption_mb to a named constant for clarity and consistencyThe relaxed threshold to 2700 MB looks fine. To avoid drift between code and documentation and to make the intent explicit, consider using an UPPER_SNAKE_CASE constant and referencing it in the range calculation.
Apply this diff:
- extra_consumption_mb = 2700 - expected_free_mem_range = ( - total_mem_mb - estimated_model_size_mb - extra_consumption_mb, + EXTRA_CONSUMPTION_MB = 2700 + expected_free_mem_range = ( + total_mem_mb - estimated_model_size_mb - EXTRA_CONSUMPTION_MB, total_mem_mb - estimated_model_size_mb, )
609-628: Docstring formulas: fix units (MB→bytes) and align metric names/signs with codeGreat added documentation. A few nits to prevent confusion:
- Use the same metric names as the code (free_mem_pre_mb/free_mem_post_mb).
- Step 3 should be pre - post, not post - pre.
- Include MB→bytes conversion in the new_cache_size formula to match the parser/validator.
Apply this diff:
- 1. free_mem_pre_fw_pass is in: - [Total mem - expected_model_size - extra_consumption, Total mem - expected_model_size] - 2. free_mem_post_fw_pass is in: - [Total mem - expected_model_size - extra_consumption - 1000, Total mem - expected_model_size - 500] - 3. free_mem_post_fw_pass - free_mem_pre_fw_pass < 5000 - 4. expected_new_cache = free_mem_post * free_mem_ratio + current_cache_size + 1. free_mem_pre_mb is in: + [Total mem - expected_model_size - extra_consumption_mb, Total mem - expected_model_size] + 2. free_mem_post_mb is in: + [Total mem - expected_model_size - extra_consumption_mb - 1000, Total mem - expected_model_size - 500] + 3. 0 < free_mem_pre_mb - free_mem_post_mb < 5000 + 4. expected_new_cache = free_mem_post_mb * 1024 * 1024 * free_mem_ratio + current_cache_size cache_size_diff = abs(new_cache_size - expected_new_cache) / expected_new_cache assert cache_size_diff <= 0.01 - extra_consumption_mb = 2700 - this is unexplained memory consumption to be investigated. + extra_consumption_mb = 2700 — unexplained memory consumption (see https://github.com/NVIDIA/TensorRT-LLM/issues/6335).
1-6: Missing NVIDIA copyright headerPer repo guidelines, Python sources should include the NVIDIA copyright header (current year). This file is missing it.
Do you want me to add the standard header used in this repo to the file prolog?
📜 Review details
Configuration used: .coderabbit.yaml
Review profile: CHILL
Plan: Pro
💡 Knowledge Base configuration:
- MCP integration is disabled by default for public repositories
- Jira integration is disabled by default for public repositories
- Linear integration is disabled by default for public repositories
You can enable these sources in your CodeRabbit configuration.
📒 Files selected for processing (1)
tests/unittest/_torch/auto_deploy/unit/singlegpu/test_ad_trtllm_bench.py(2 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/unittest/_torch/auto_deploy/unit/singlegpu/test_ad_trtllm_bench.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/unittest/_torch/auto_deploy/unit/singlegpu/test_ad_trtllm_bench.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/unittest/_torch/auto_deploy/unit/singlegpu/test_ad_trtllm_bench.py
MrGeva
changed the title
Signed-off-by: Eran Geva <[email protected]>
There was a problem hiding this comment.
Actionable comments posted: 0
🔭 Outside diff range comments (1)
tests/unittest/_torch/auto_deploy/unit/singlegpu/test_ad_trtllm_bench.py (1)
1-6: Add the mandatory NVIDIA copyright header.Per the coding guidelines, prepend the NVIDIA copyright header to all Python sources.
Apply at the very top of the file:
+# Copyright (c) 2025, NVIDIA CORPORATION & AFFILIATES. All rights reserved. + import json import re import subprocess import tempfile from pathlib import Path
🧹 Nitpick comments (6)
tests/unittest/_torch/auto_deploy/unit/singlegpu/test_ad_trtllm_bench.py (6)
281-286: Increasing extra_consumption_mb to 2700 MB: add a floor to avoid negative ranges.The widened slack looks fine but can produce a negative lower bound on small GPUs. Clamp to 0 to avoid invalid ranges.
- expected_free_mem_range = ( - total_mem_mb - estimated_model_size_mb - extra_consumption_mb, - total_mem_mb - estimated_model_size_mb, - ) + lower = max(total_mem_mb - estimated_model_size_mb - extra_consumption_mb, 0) + expected_free_mem_range = ( + lower, + total_mem_mb - estimated_model_size_mb, + )Optional verification: please run on the smallest CI GPU type we support to ensure the lower bound is non-negative in practice.
292-294: Post-forward free memory range should allow a small additional drop to reduce flakiness.Making post == pre range tightens the check. Allowing a modest cushion (e.g., 0.5–1.0 GB) better reflects transient allocator behavior during forward passes.
- expected_free_mem_post_range = expected_free_mem_range + # Allow a small additional drop during forward pass to reduce flakiness + expected_free_mem_post_range = ( + max(expected_free_mem_range[0] - 1024, 0), # -1 GB + max(expected_free_mem_range[1] - 512, 0), # -0.5 GB + )
347-354: Permit small allocator jitter in “memory reduction” check.A strict > 0 can be flaky due to allocator noise and async telemetry. Allow a small negative jitter (e.g., -50 MB) while still flagging true regressions.
- if free_mem_pre and free_mem_post: - memory_reduction = free_mem_pre - free_mem_post - assert memory_reduction > 0, ( - f"Expected memory reduction during forward pass, got {memory_reduction}MB" - ) - print(f" ✅ Memory reduction during forward pass: {memory_reduction}MB") + if free_mem_pre and free_mem_post: + memory_reduction = free_mem_pre - free_mem_post + min_reduction_mb = -50 # allow small allocator jitter + assert memory_reduction > min_reduction_mb, ( + f"Expected memory reduction during forward pass (allowing jitter {min_reduction_mb}MB), " + f"got {memory_reduction}MB" + ) + print(f" ✅ Memory reduction during forward pass: {memory_reduction}MB")
605-624: Docstring: align terminology with code keys and clarify units.Use the exact metric names and note the MB→bytes conversion in new_cache_size to avoid confusion.
- """Test that compares autodeploy backend performance against pytorch backend - with given relative and absolute thresholds. - - It also checks the memory footprint of the autodeploy backend by parsing the - log output from the resize_kv_cache function and extracting the following metrics: - current_cache_size - the cache size before resize - free_mem_pre_mb - the free memory before forward pass - free_mem_post_mb - the free memory after forward pass - new_cache_size - the cache size after resize - - The following checks are performed: - 1. free_mem_pre_fw_pass and free_mem_post_fw_pass are in: - [Total mem - expected_model_size - extra_consumption, Total mem - expected_model_size] - 2. memory_reduction = free_mem_pre_fw_pass - free_mem_post_fw_pass > 0 - 3. expected_new_cache = free_mem_post * free_mem_ratio + current_cache_size - cache_size_diff = abs(new_cache_size - expected_new_cache) / expected_new_cache - assert cache_size_diff <= 0.01 - - extra_consumption_mb = 2700 - this is unexplained memory consumption to be investigated. - """ + """Test autodeploy vs. PyTorch backend with performance and memory-footprint validation. + + Parses resize_kv_cache logs and extracts: + - current_cache_size (bytes): cache size before resize + - free_mem_pre_mb (MB): free memory before forward pass + - free_mem_post_mb (MB): free memory after forward pass + - new_cache_size (bytes): cache size after resize + + Checks performed: + 1) free_mem_pre_mb and free_mem_post_mb are in: + [TotalMB - expected_model_size_mb - extra_consumption_mb, TotalMB - expected_model_size_mb] + 2) memory_reduction = free_mem_pre_mb - free_mem_post_mb > 0 (allowing small allocator jitter) + 3) new_cache_size matches: + expected_new_cache = free_mem_post_mb * 1024 * 1024 * free_mem_ratio + current_cache_size + with relative error <= 1% + + Note: extra_consumption_mb = 2700 is a temporary allowance pending investigation. + """
409-414: require_metrics=False path still asserts — make it non-fatal as intended.The docstring says “just warn,” but the else-branches assert False. Return a neutral value instead so callers can opt into non-fatal behavior.
- else: - print(f"ℹ️ {message}") - assert False, "KV cache metrics are missing" + else: + print(f"ℹ️ {message}") + return None, None @@ - else: - print(f"ℹ️ KV cache validation skipped - {message}") - assert False, "KV cache metrics are missing" + else: + print(f"ℹ️ KV cache validation skipped - {message}") + return None, NoneAlso applies to: 427-429
67-71: Use sys.executable instead of hardcoding "python"/"python3".This ensures the subprocess uses the same interpreter as pytest, avoids PATH issues, and keeps consistency.
Example change (outside selected ranges):
import sys # In run_benchmark() cmd = [ sys.executable, "-m", "tensorrt_llm.commands.bench", # ... ] # In prepare_dataset() command = [ sys.executable, str(dataset_tool), "--stdout", # ... ]Also applies to: 239-241
📜 Review details
Configuration used: .coderabbit.yaml
Review profile: CHILL
Plan: Pro
💡 Knowledge Base configuration:
- MCP integration is disabled by default for public repositories
- Jira integration is disabled by default for public repositories
- Linear integration is disabled by default for public repositories
You can enable these sources in your CodeRabbit configuration.
📒 Files selected for processing (1)
tests/unittest/_torch/auto_deploy/unit/singlegpu/test_ad_trtllm_bench.py(5 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/unittest/_torch/auto_deploy/unit/singlegpu/test_ad_trtllm_bench.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/unittest/_torch/auto_deploy/unit/singlegpu/test_ad_trtllm_bench.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/unittest/_torch/auto_deploy/unit/singlegpu/test_ad_trtllm_bench.py
⏰ 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)
tests/unittest/_torch/auto_deploy/unit/singlegpu/test_ad_trtllm_bench.py (2)
282-282: LGTM: Relaxing extra_consumption_mb to 2700 MB.Given recent allocator behavior, this bump looks reasonable and matches the docstring note.
448-458: Confirm increase of backend_relative_tolerance to 30%We raised the default from 20% → 30% in
trtllm_bench_unified_comparison(tests/unittest/_torch/auto_deploy/unit/singlegpu/test_ad_trtllm_bench.py:447–454). A quick search in this file found:
- Default signature now uses
backend_relative_tolerance=0.3- One explicit call at line 625 relies on the new default
- Docstring at lines 469–473 doesn’t mention the old value
Please verify that:
- A 30% tolerance is acceptable given our hardware variability
- Any documentation or tests outside this file that reference the previous 20% threshold are updated accordingly (e.g., global search for “0.2” or “20%” in your docs/tests)
|
/bot run |
|
PR_Github #15619 [ run ] triggered by Bot |
|
/bot run |
|
PR_Github #15641 [ run ] triggered by Bot |
|
PR_Github #15619 [ run ] completed with state |
|
PR_Github #15641 [ run ] completed with state |
MrGeva
changed the title
3 participants
Summary by CodeRabbit
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.
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 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.