[https://nvbugs/5448525][fix] Mistral Small 3.1 accuracy tests #6909

2ez4bz · 2025-08-14T18:27:10Z

Summary by CodeRabbit

Tests
- Updated expected accuracy reference for the CNN/DailyMail benchmark under FP8 quantization for the Mistral Small 24B variant.
- Enhanced the auto-dtype test by supplying a kv-cache configuration with a defined GPU memory fraction to better reflect runtime behavior.
Chores
- Synchronized test data to reflect the latest model performance under FP8, ensuring consistency across integration checks.

Description

This commit lowers the GPU memory allocated for KV cache in accuracy tests, and adjusts a threshold for Mistral Small 3.1 24B for FP8.

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 the stage-list parameter 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.md
and the scripts/test_to_stage_mapping.py helper.

kill

kill

Kill all running builds associated with pull request.

skip

skip --comment COMMENT

Skip 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-pipeline

Reuse 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.

This commit lowers the GPU memory allocated for KV cache in accuracy tests, and adjusts a threshold for Mistral Small 3.1 24B for FP8. Signed-off-by: William Zhang <133824995+2ez4bz@users.noreply.github.com>

coderabbitai · 2025-08-14T18:27:16Z

📝 Walkthrough

Walkthrough

Updates a CNN/DailyMail accuracy reference value for the Mistral-Small-3.1-24B-Instruct model under FP8 KV-cache quantization and modifies a PyTorch integration test to instantiate LLM with a KvCacheConfig including free_gpu_memory_fraction=0.75.

Changes

Cohort / File(s)	Summary
Accuracy reference update `tests/integration/defs/accuracy/references/cnn_dailymail.yaml`	Adjusts FP8 (kv_cache_quant_algo: FP8) accuracy for mistralai/Mistral-Small-3.1-24B-Instruct-2503 from 29.0 to 27.0; no other entries changed.
PyTorch integration test config `tests/integration/defs/accuracy/test_llm_api_pytorch.py`	In TestMistralSmall24B.test_auto_dtype, constructs KvCacheConfig with free_gpu_memory_fraction=0.75 and passes it to LLM via kv_cache_config; assertions unchanged.

Sequence Diagram(s)

(omitted)

Estimated code review effort

🎯 2 (Simple) | ⏱️ ~8 minutes

Possibly related PRs

[test] Add accuracy regression test for Mistral3.1 #6322: Touches the same model’s accuracy reference and adds the related test; directly connected to the updated YAML and test.
tests: add TestNemotronH cuda graph tests #6390: Introduces passing KvCacheConfig into LLM in the same test suite; similar code changes.
[https://nvbugs/5436461][infra] Adjust free_gpu_memory_fraction of test_eagle3 to prevent OOM on CI #6631: Modifies KvCacheConfig construction to include free_gpu_memory_fraction in related tests.

Suggested labels

Community want to contribute

Suggested reviewers

LarryXFly
litaotju
yiqingy0

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 Docstrings

🧪 Generate unit tests

Create PR with unit tests
Post copyable unit tests in a comment

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.

❤️ Share

🪧 Tips

Chat

There are 3 ways to chat with CodeRabbit:

Review comments: Directly reply to a review comment made by CodeRabbit. Example:
- I pushed a fix in commit <commit_id>, please review it.
- Open a follow-up GitHub issue for this discussion.
Files and specific lines of code (under the "Files changed" tab): Tag @coderabbitai in a new review comment at the desired location with your query.
PR comments: Tag @coderabbitai in a new PR comment to ask questions about the PR branch. For the best results, please provide a very specific query, as very limited context is provided in this mode. Examples:
- @coderabbitai gather interesting stats about this repository and render them as a table. Additionally, render a pie chart showing the language distribution in the codebase.
- @coderabbitai read the files in the src/scheduler package and generate a class diagram using mermaid and a README in the markdown format.

Support

Need help? Create a ticket on our support page for assistance with any issues or questions.

CodeRabbit Commands (Invoked using PR/Issue comments)

Type @coderabbitai help to get the list of available commands.

Other keywords and placeholders

Add @coderabbitai ignore anywhere in the PR description to prevent this PR from being reviewed.
Add @coderabbitai summary to generate the high-level summary at a specific location in the PR description.
Add @coderabbitai or @coderabbitai title anywhere in the PR title to generate the title automatically.

Status, Documentation and Community

Visit our Status Page to check the current availability of CodeRabbit.
Visit our Documentation for detailed information on how to use CodeRabbit.
Join our Discord Community to get help, request features, and share feedback.
Follow us on X/Twitter for updates and announcements.

coderabbitai

Actionable comments posted: 0

🧹 Nitpick comments (1)

tests/integration/defs/accuracy/test_llm_api_pytorch.py (1)
714-716: Ensure the FP8 path actually uses FP8 KV (not just auto) to match the updated reference

When the FP8 prequantized model path is used, dtype="auto" may or may not select FP8 for the KV cache depending on backend defaults. To remove ambiguity and align with the updated reference (which is tagged kv_cache_quant_algo: FP8), set KV dtype explicitly when expected_quant_algo is FP8.

Apply this diff to conditionally enforce FP8 KV in the FP8 case:
     def test_auto_dtype(self, model_path, expected_quant_algo):
-        kv_cache_config = KvCacheConfig(free_gpu_memory_fraction=0.75)
+        kv_cache_config = KvCacheConfig(free_gpu_memory_fraction=0.75)
+        if expected_quant_algo == QuantAlgo.FP8:
+            kv_cache_config.dtype = "fp8"
         with LLM(model_path, kv_cache_config=kv_cache_config) as llm:
Optionally, assert the KV quantization algo in the FP8 branch to catch regressions:
# Insert right after the QuantAlgo assertion
if expected_quant_algo == QuantAlgo.FP8:
    assert llm.args.quant_config.kv_cache_quant_algo == QuantAlgo.FP8

📜 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 settings in your CodeRabbit configuration.

📥 Commits

Reviewing files that changed from the base of the PR and between 7ebb770 and 9289718.

📒 Files selected for processing (2)

tests/integration/defs/accuracy/references/cnn_dailymail.yaml (1 hunks)
tests/integration/defs/accuracy/test_llm_api_pytorch.py (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/defs/accuracy/test_llm_api_pytorch.py

🧬 Code Graph Analysis (1)

tests/integration/defs/accuracy/test_llm_api_pytorch.py (2)

tensorrt_llm/llmapi/llm_args.py (1)

KvCacheConfig (923-1002)

tensorrt_llm/llmapi/llm.py (1)

LLM (1111-1127)

🔇 Additional comments (2)

tests/integration/defs/accuracy/references/cnn_dailymail.yaml (1)

206-209: Reference adjusted for FP8 KV-cache: sanity-check against latest baseline runs

Lowering cnn_dailymail accuracy to 27.0 for mistralai/Mistral-Small-3.1-24B-Instruct-2503 under FP8 KV aligns with the PR intent to reduce KV cache memory. Please confirm that recent CI accuracy runs (with the new kv cache fraction) consistently land near this value to avoid future flakes.

tests/integration/defs/accuracy/test_llm_api_pytorch.py (1)

714-716: Good call: explicitly cap KV cache memory to 75% to reduce OOM risk

This is consistent with similar large-model tests and the PR’s goal. No functional issues spotted.

2ez4bz · 2025-08-14T23:01:57Z

/bot run

tensorrt-cicd · 2025-08-14T23:07:02Z

PR_Github #15348 [ run ] triggered by Bot

tensorrt-cicd · 2025-08-15T02:37:58Z

PR_Github #15348 [ run ] completed with state SUCCESS
/LLM/release-1.0/L0_MergeRequest_PR pipeline #126 completed with status: 'FAILURE'

2ez4bz · 2025-08-15T04:04:49Z

/bot run

tensorrt-cicd · 2025-08-15T04:10:01Z

PR_Github #15392 [ run ] triggered by Bot

tensorrt-cicd · 2025-08-15T07:28:36Z

PR_Github #15392 [ run ] completed with state SUCCESS
/LLM/release-1.0/L0_MergeRequest_PR pipeline #134 completed with status: 'SUCCESS'

2ez4bz · 2025-08-15T16:58:28Z

/bot run --extra-stage "H100_PCIe-PyTorch-Post-Merge-1"

tensorrt-cicd · 2025-08-15T17:03:56Z

PR_Github #15462 [ run ] triggered by Bot

tests/integration/defs/accuracy/test_llm_api_pytorch.py

brb-nv

LGTM.

tensorrt-cicd · 2025-08-15T20:50:35Z

PR_Github #15462 [ run ] completed with state SUCCESS
/LLM/release-1.0/L0_MergeRequest_PR pipeline #147 completed with status: 'SUCCESS'

The error should have been fixed by NVIDIA#6909. Signed-off-by: William Zhang <133824995+2ez4bz@users.noreply.github.com>

The error was fixed by #6909. Signed-off-by: William Zhang <133824995+2ez4bz@users.noreply.github.com>

…A#6909) This commit lowers the GPU memory allocated for KV cache in accuracy tests, and adjusts a threshold for Mistral Small 3.1 24B for FP8. Signed-off-by: William Zhang <133824995+2ez4bz@users.noreply.github.com>

…A#6909) This commit lowers the GPU memory allocated for KV cache in accuracy tests, and adjusts a threshold for Mistral Small 3.1 24B for FP8. Signed-off-by: William Zhang <133824995+2ez4bz@users.noreply.github.com> Signed-off-by: Wangshanshan <30051912+dominicshanshan@users.noreply.github.com>

This commit lowers the GPU memory allocated for KV cache in accuracy tests, and adjusts a threshold for Mistral Small 3.1 24B for FP8. Signed-off-by: William Zhang <133824995+2ez4bz@users.noreply.github.com> Signed-off-by: Wangshanshan <30051912+dominicshanshan@users.noreply.github.com>

[https://nvbugs/5448525][fix] Mistral Small 3.1 accuracy tests

9289718

This commit lowers the GPU memory allocated for KV cache in accuracy tests, and adjusts a threshold for Mistral Small 3.1 24B for FP8. Signed-off-by: William Zhang <133824995+2ez4bz@users.noreply.github.com>

2ez4bz requested a review from a team as a code owner August 14, 2025 18:27

coderabbitai bot reviewed Aug 14, 2025

View reviewed changes

brb-nv reviewed Aug 15, 2025

View reviewed changes

tests/integration/defs/accuracy/test_llm_api_pytorch.py Show resolved Hide resolved

brb-nv approved these changes Aug 15, 2025

View reviewed changes

2ez4bz enabled auto-merge (squash) August 15, 2025 18:50

chzblych approved these changes Aug 18, 2025

View reviewed changes

2ez4bz merged commit c16aff5 into NVIDIA:release/1.0 Aug 18, 2025
5 checks passed

coderabbitai bot mentioned this pull request Aug 18, 2025

[https://nvbugs/5440241][fix] Fix 70B GSM8K Accuracy drop #6967

Merged

2ez4bz added a commit to 2ez4bz/TensorRT-LLM that referenced this pull request Aug 18, 2025

[https://nvbugs/5462007][ci] Unwaive Mistral Small 3.1 FP8 test

a2421b1

The error should have been fixed by NVIDIA#6909. Signed-off-by: William Zhang <133824995+2ez4bz@users.noreply.github.com>

2ez4bz added a commit to 2ez4bz/TensorRT-LLM that referenced this pull request Aug 18, 2025

[https://nvbugs/5462007][ci] Unwaive Mistral Small 3.1 FP8 test

73541d3

The error should have been fixed by NVIDIA#6909. Signed-off-by: William Zhang <133824995+2ez4bz@users.noreply.github.com>

2ez4bz added a commit to 2ez4bz/TensorRT-LLM that referenced this pull request Aug 18, 2025

[https://nvbugs/5462007][ci] Unwaive Mistral Small 3.1 FP8 test

482d9e2

The error should have been fixed by NVIDIA#6909. Signed-off-by: William Zhang <133824995+2ez4bz@users.noreply.github.com>

2ez4bz mentioned this pull request Aug 18, 2025

[https://nvbugs/5462007][ci] Unwaive Mistral Small 3.1 FP8 test #7008

Merged

coderabbitai bot mentioned this pull request Aug 18, 2025

[https://nvbugs/5454875][ci] Unwaive Mistral Small 3.1 test #7011

Merged

2ez4bz added a commit that referenced this pull request Aug 18, 2025

[https://nvbugs/5462007][ci] Unwaive Mistral Small 3.1 FP8 test (#7008)

790a105

The error was fixed by #6909. Signed-off-by: William Zhang <133824995+2ez4bz@users.noreply.github.com>

coderabbitai bot mentioned this pull request Aug 20, 2025

[https://nvbugs/5440241][fix] Fix 70B GSM8K Accuracy drop #7075

Merged

yuanjingx87 pushed a commit that referenced this pull request Aug 20, 2025

[https://nvbugs/5462007][ci] Unwaive Mistral Small 3.1 FP8 test (#7008)

76dd647

The error was fixed by #6909. Signed-off-by: William Zhang <133824995+2ez4bz@users.noreply.github.com>

This was referenced Aug 20, 2025

[None][fix] Correct KV cache percentage report out. #7102

Merged

[None] [fix] improve kvcache allocation in PyTorch runtime #5933

Merged

This was referenced Aug 24, 2025

[None][chore] remove CLI support for mamba cache dtype setting #7119

Merged

[https://nvbugs/5458798][fix] AD perf test outliers handling, tightened threshold, re-enabled in CI, fixed mem threshold #7189

Merged

coderabbitai bot mentioned this pull request Aug 27, 2025

[TRTLLM-6646][test] NIM migration to TRT-LLM LLMAPI : Add QWQ-32b torch test #7284

Merged

1 task

[https://nvbugs/5448525][fix] Mistral Small 3.1 accuracy tests #6909

[https://nvbugs/5448525][fix] Mistral Small 3.1 accuracy tests #6909

Uh oh!

Conversation

2ez4bz commented Aug 14, 2025 • edited by coderabbitai bot Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary by CodeRabbit

Description

Test Coverage

GitHub Bot Help

kill

skip

reuse-pipeline

Uh oh!

coderabbitai bot commented Aug 14, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Walkthrough

Changes

Sequence Diagram(s)

Estimated code review effort

Possibly related PRs

Suggested labels

Suggested reviewers

Chat

Support

CodeRabbit Commands (Invoked using PR/Issue comments)

Other keywords and placeholders

Status, Documentation and Community

Uh oh!

coderabbitai bot left a comment

Choose a reason for hiding this comment

Uh oh!

2ez4bz commented Aug 14, 2025

Uh oh!

tensorrt-cicd commented Aug 14, 2025

Uh oh!

tensorrt-cicd commented Aug 15, 2025

Uh oh!

2ez4bz commented Aug 15, 2025

Uh oh!

tensorrt-cicd commented Aug 15, 2025

Uh oh!

tensorrt-cicd commented Aug 15, 2025

Uh oh!

2ez4bz commented Aug 15, 2025

Uh oh!

tensorrt-cicd commented Aug 15, 2025

Uh oh!

Uh oh!

brb-nv left a comment

Choose a reason for hiding this comment

Uh oh!

tensorrt-cicd commented Aug 15, 2025

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

2ez4bz commented Aug 14, 2025 •

edited by coderabbitai bot

Loading

coderabbitai bot commented Aug 14, 2025 •

edited

Loading