[None][fix] Correct KV cache percentage report out. #7102
Conversation
📝 WalkthroughWalkthroughCentralizes KV-cache configuration handling in tensorrt_llm/bench/dataclasses/reporting.py:get_statistics_dict by reading Changes
Sequence Diagram(s)sequenceDiagram
autonumber
participant Caller as Caller
participant Reporting as reporting.get_statistics_dict
participant Cfg as kv_cache_config (kwargs)
participant PT as PyTorch backend
participant TRT as TRT/backend
participant Val as validate_and_set_kv_cache_quant
participant World as world_info
Caller->>Reporting: get_statistics_dict(kwargs)
Reporting->>Cfg: read kv_cache_config from kwargs
alt kv_cache_config valid or absent
Reporting->>Reporting: derive kv_cache_dtype and kv_cache_mem_percent
Reporting->>Reporting: convert mem fraction -> percentage string (e.g., "12.34%") or "None"
else invalid type
Reporting-->>Caller: raise ValueError
end
par Backends
Reporting->>PT: call validate_and_set_kv_cache_quant(kv_cache_dtype)
PT->>Val: validate_and_set_kv_cache_quant(kv_cache_dtype)
Note right of PT: PyTorch uses centralized kv_cache_dtype
and
Reporting->>TRT: use pretrained-config sourced kv_cache_dtype (behavior preserved)
end
Reporting->>World: set kv_cache_percentage = percentage string
Reporting-->>Caller: return statistics dict
Estimated code review effort🎯 3 (Moderate) | ⏱️ ~20 minutes Possibly related PRs
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: 1
📜 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)
tensorrt_llm/bench/dataclasses/reporting.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:
tensorrt_llm/bench/dataclasses/reporting.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:
tensorrt_llm/bench/dataclasses/reporting.py
🔇 Additional comments (1)
tensorrt_llm/bench/dataclasses/reporting.py (1)
343-343: LGTM: percentage now reflects the configured fractionUsing kv_cache_mem_percent * 100.0 fixes the earlier mismatch and aligns printed/reporting percentage with the LLM API’s free_gpu_memory_fraction.
|
PR_Github #15954 [ run ] triggered by Bot |
|
/bot run |
|
PR_Github #15955 [ run ] triggered by Bot |
|
PR_Github #15954 [ run ] completed with state |
|
PR_Github #15955 [ run ] completed with state |
FrankD412
force-pushed
the
fdinatale/trtllm-bench/kv_cache_reporting
branch
from
3edeb3a to
eb3dfb0
Compare
|
/bot run |
|
PR_Github #15972 [ run ] triggered by Bot |
|
@Superjomn -- out of curiosity, I encountered the case where the |
|
PR_Github #15972 [ run ] completed with state |
FrankD412
force-pushed
the
fdinatale/trtllm-bench/kv_cache_reporting
branch
from
b48e16b to
a5b8b9f
Compare
|
/bot run |
|
PR_Github #15991 [ run ] triggered by Bot |
|
PR_Github #15991 [ run ] completed with state |
|
/bot run |
1 similar comment
|
/bot run |
|
PR_Github #16074 [ run ] triggered by Bot |
FrankD412
force-pushed
the
fdinatale/trtllm-bench/kv_cache_reporting
branch
from
e53eba1 to
6c10ba2
Compare
|
/bot run |
|
PR_Github #16074 [ run ] completed with state |
|
PR_Github #16078 [ run ] triggered by Bot |
|
PR_Github #16078 [ run ] completed with state |
|
/bot run |
|
PR_Github #16084 [ run ] triggered by Bot |
|
PR_Github #16084 [ run ] completed with state |
Signed-off-by: Frank Di Natale <[email protected]>
Signed-off-by: Frank Di Natale <[email protected]>
Signed-off-by: Frank Di Natale <[email protected]>
Signed-off-by: Frank Di Natale <[email protected]>
Signed-off-by: Frank Di Natale <[email protected]>
Signed-off-by: Frank Di Natale <[email protected]>
FrankD412
force-pushed
the
fdinatale/trtllm-bench/kv_cache_reporting
branch
from
79404b0 to
46449e2
Compare
|
/bot run |
There was a problem hiding this comment.
Actionable comments posted: 0
♻️ Duplicate comments (1)
tensorrt_llm/bench/dataclasses/reporting.py (1)
276-291: Prevent crash when kv_cache_config is None or dict missing free_gpu_memory_fraction; coerce and clamp fraction to [0,1]As written:
- kv_cache_config=None raises ValueError (Line 286).
- dict inputs without "free_gpu_memory_fraction" propagate None and cause TypeError at Line 289 when multiplied.
- Non-numeric strings will also fail at Line 289.
Fix by treating None as “not provided” (use KvCacheConfig() defaults), defaulting missing dict fields, coercing to float when present, and clamping to [0,1]. Also generate the display string only after normalization.
Apply this diff:
- # Retrieve KV cache information. - kv_cache_config = self.kwargs.get("kv_cache_config", KvCacheConfig()) - if isinstance(kv_cache_config, KvCacheConfig): - kv_cache_dtype = kv_cache_config.dtype - kv_cache_mem_percent = kv_cache_config.free_gpu_memory_fraction - elif isinstance(kv_cache_config, dict): - kv_cache_dtype = kv_cache_config.get("dtype", "auto") - kv_cache_mem_percent = kv_cache_config.get( - "free_gpu_memory_fraction") - else: - raise ValueError( - f"Invalid kv_cache_config type: {type(kv_cache_config)}.") - - kv_cache_mem_percent = f"{kv_cache_mem_percent * 100.0:.2f}%" \ - if kv_cache_mem_percent is not None else "None" + # Retrieve KV cache information and normalize. + kv_cache_config = self.kwargs.get("kv_cache_config") + if kv_cache_config is None: + default_cfg = KvCacheConfig() + kv_cache_dtype = default_cfg.dtype + kv_cache_mem_fraction = default_cfg.free_gpu_memory_fraction + elif isinstance(kv_cache_config, KvCacheConfig): + kv_cache_dtype = kv_cache_config.dtype + kv_cache_mem_fraction = kv_cache_config.free_gpu_memory_fraction + elif isinstance(kv_cache_config, dict): + default_cfg = KvCacheConfig() + kv_cache_dtype = kv_cache_config.get("dtype", default_cfg.dtype) + kv_cache_mem_fraction = kv_cache_config.get( + "free_gpu_memory_fraction", default_cfg.free_gpu_memory_fraction) + else: + raise ValueError( + f"Invalid kv_cache_config type: {type(kv_cache_config)}.") + + # Build display string; coerce when present, allow unset (None). + if kv_cache_mem_fraction is None: + kv_cache_mem_percent = "None" + else: + try: + mem_frac = float(kv_cache_mem_fraction) + except (TypeError, ValueError) as exc: + raise ValueError( + "kv_cache_config.free_gpu_memory_fraction must be a number in [0, 1]." + ) from exc + if not (0.0 <= mem_frac <= 1.0): + self.logger.warning( + f"free_gpu_memory_fraction={mem_frac} is out of [0,1]; clamping.") + mem_frac = max(0.0, min(1.0, mem_frac)) + kv_cache_mem_percent = f"{mem_frac * 100.0:.2f}%"To validate callers that may pass None or dicts without the fraction (the scenarios that can currently break), run:
#!/bin/bash # 1) Find kv_cache_config constructions/usages across the repo. rg -nP --type=py -C3 '\bkv_cache_config\b' # 2) Spot dict literals likely passed as kv_cache_config and check for missing 'free_gpu_memory_fraction'. rg -nP --type=py -C2 '(kv_cache_config\s*=\s*\{|\bkv_cache_config\b[^,\)\n]*\{)' | sed -n '1,200p' # 3) Locate explicit None assignments to kv_cache_config. rg -nP --type=py -C2 'kv_cache_config\s*=\s*None' # 4) Audit for callers that set dtype but omit the fraction (common pitfall). rg -nP --type=py -C2 'kv_cache_config[^=\n]*\{[^}]*dtype[^}]*\}' | rg -nP --invert-match 'free_gpu_memory_fraction'
🧹 Nitpick comments (1)
tensorrt_llm/bench/dataclasses/reporting.py (1)
535-535: Nit: align the printed label with the key nameConsider making the label match the key and concept (“KV Cache Percentage”) for clarity.
- f"KV Memory Percentage: {world_info['kv_cache_percentage']}\n" + f"KV Cache Percentage: {world_info['kv_cache_percentage']}\n"
📜 Review details
Configuration used: Path: .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)
tensorrt_llm/bench/dataclasses/reporting.py(3 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:
tensorrt_llm/bench/dataclasses/reporting.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:
tensorrt_llm/bench/dataclasses/reporting.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)
tensorrt_llm/bench/dataclasses/reporting.py (2)
289-291: Follow-up on percent formattingOnce the normalization above is applied, this formatting is safe and consistent. No additional change needed here.
346-346: LGTM: world_info now reflects configured KV cache percentThis correctly wires the reported percentage to the normalized KV cache configuration rather than the previous settings path, aligning with the PR objective.
|
PR_Github #16144 [ run ] triggered by Bot |
|
PR_Github #16144 [ run ] completed with state |
Signed-off-by: Frank Di Natale <[email protected]>
Signed-off-by: Frank Di Natale <[email protected]>
3 participants
Summary by CodeRabbit
Bug Fixes
Refactor
Chores
There was a minor bug in the reporting of
trtllm-benchwhere if the KV cache percentage was set in the LLM API extra options, the percentage would not be reflected in the final report out. This PR fixes the issue by directly reporting from the KV cache configuration.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.