Skip to content

Conversation

@alisonshao
Copy link
Collaborator

Summary

  • Add basic functionality test for mistralai/Mistral-Large-3-675B-Instruct-2512 to PR CI
  • Test runs on per-commit-8-gpu-h200 suite
  • Includes GSM8K accuracy check and single batch speed test

Configuration

  • TP=8
  • --attention-backend trtllm_mla
  • --model-loader-extra-config '{"enable_multithread_load": true}'
  • --chat-template mistral
  • SGLANG_ENABLE_JIT_DEEPGEMM=0

Related PR

Test plan

  • Verify test runs successfully on 8-gpu-h200 runner
  • Check GSM8K accuracy meets threshold (>0.90)
  • Check speed meets threshold (>50 token/s)

Add basic functionality test for mistralai/Mistral-Large-3-675B-Instruct-2512:
- GSM8K accuracy check
- Single batch speed test
- Runs on per-commit-8-gpu-h200 suite

Configuration:
- TP=8
- trtllm_mla attention backend
- Multithread loading enabled
- Mistral chat template
- SGLANG_ENABLE_JIT_DEEPGEMM=0
@gemini-code-assist
Copy link
Contributor

Warning

You have reached your daily quota limit. Please wait up to 24 hours and I will start processing your requests again!

@alisonshao
Copy link
Collaborator Author

/tag-and-rerun-ci

- Move test from per-commit-8-gpu-h200 to per-commit-8-gpu-b200 suite
- Add new unit-test-backend-8-gpu-b200 job in pr-test.yml
- trtllm_mla backend requires Blackwell GPUs (SM100)
@Fridge003 Fridge003 merged commit 16e8463 into main Dec 5, 2025
294 of 321 checks passed
@Fridge003 Fridge003 deleted the feature/mistral-large3-pr-test branch December 5, 2025 18:38
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

Projects

None yet

Development

Successfully merging this pull request may close these issues.

4 participants