Skip to content

Add RunPod Lyra-2 installation runbook#56

Open
durul wants to merge 1 commit into
nv-tlabs:mainfrom
durul:runpod-install-runbook
Open

Add RunPod Lyra-2 installation runbook#56
durul wants to merge 1 commit into
nv-tlabs:mainfrom
durul:runpod-install-runbook

Conversation

@durul

@durul durul commented May 3, 2026

Copy link
Copy Markdown

Summary

  • Replaces the minimal Lyra-2 install notes with a RunPod-verified installation runbook.
  • Documents the verified H200 NVL environment, conda/CUDA/PyTorch pins, extension build steps, checkpoint download options, smoke tests, restart recovery, and troubleshooting.

Validation

  • Reviewed the single-file diff for scope and command consistency.
  • Ran git diff origin/main...HEAD --check successfully.
  • Checked Markdown code fences are balanced.
  • Ran markdownlint-cli2 Lyra-2/INSTALL.md; the repo has no markdownlint config, so the default style run reports line-length/list-format warnings for the long runbook prose.
  • Did not rerun GPU smoke tests locally; the document records the RunPod-verified smoke test outputs.

Replace the minimal INSTALL.md with a comprehensive RunPod-verified Lyra-2 installation runbook. Documents a full end-to-end install and verified configuration for an H200 NVL pod (Ubuntu 24.04), including Miniforge/conda env creation, pinned CUDA/PyTorch versions, build environment variables, NVTX symlinks, building transformer_engine/flash-attn/vipe/depth_anything_3, checkpoint download strategies, smoke tests, persistence caveats, and a detailed troubleshooting summary. Intended as a teammate-facing runbook with exact commands, expected outputs, and recovery steps for pod stop/recreate events.
@durul durul marked this pull request as ready for review May 3, 2026 18:56
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