Skip to content

fix: resolve blocking issues for Modules 3/4/5 on GPU (A100)#19

Open
PicoNVIDIA wants to merge 1 commit intomainfrom
gtc-bugfixes
Open

fix: resolve blocking issues for Modules 3/4/5 on GPU (A100)#19
PicoNVIDIA wants to merge 1 commit intomainfrom
gtc-bugfixes

Conversation

@PicoNVIDIA
Copy link
Collaborator

Clean slate users should be good to run this for 3,4,5

  • Add deepagents and scikit-learn to requirements.txt (missing deps for Modules 4/5)
  • Add causal-conv1d and mamba-ssm CUDA builds to postBuild.bash (Nemotron model needs these)
  • Patch mamba-ssm for transformers 5.x compatibility (GreedySearchDecoderOnlyOutput removed)
  • Add missing 'coverage' key to evaluation_framework fallback dict (KeyError on LLM parse failure)
  • Disable LangSmith tracing by default (no API key configured = wall of 401 errors)

- Add deepagents and scikit-learn to requirements.txt (missing deps for Modules 4/5)
- Add causal-conv1d and mamba-ssm CUDA builds to postBuild.bash (Nemotron model needs these)
- Patch mamba-ssm for transformers 5.x compatibility (GreedySearchDecoderOnlyOutput removed)
- Add missing 'coverage' key to evaluation_framework fallback dict (KeyError on LLM parse failure)
- Disable LangSmith tracing by default (no API key configured = wall of 401 errors)

Made-with: Cursor
@PicoNVIDIA PicoNVIDIA requested a review from nv-edwli March 11, 2026 18:46
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant