Skip to content

[TAO-0000][Feature] Cache PyTorch CUDA builds - #71

Draft
ramanathan831 wants to merge 4 commits into
NVIDIA-TAO:mainfrom
ramanathan831:dev/ram/cache-pytorch-builds
Draft

[TAO-0000][Feature] Cache PyTorch CUDA builds#71
ramanathan831 wants to merge 4 commits into
NVIDIA-TAO:mainfrom
ramanathan831:dev/ram/cache-pytorch-builds

Conversation

@ramanathan831

Copy link
Copy Markdown
Contributor

Summary

  • add reusable CUDA/PyTorch build-cache management
  • integrate opt-in --use-cache behavior with the current deploy script
  • avoid the obsolete hard-coded container digest from the old GitLab patch

Root cause

The original GitLab source branch was deleted, so its final MR diff was ported semantically onto current GitHub main.

Validation

  • bash -n release/docker/deploy.sh scripts/cuda_cache.sh
  • git diff --check

@ramanathan831
ramanathan831 marked this pull request as ready for review August 5, 2026 21:29
@ramanathan831
ramanathan831 requested a review from a team as a code owner August 5, 2026 21:29
@tao-pr-bot

tao-pr-bot Bot commented Aug 5, 2026

Copy link
Copy Markdown

👋 Thanks for the contribution!

For security reasons, CI runs on NVIDIA's runners only after a maintainer vets this PR. A TAO maintainer will review and trigger CI (/build) for you — no action needed on your side beyond addressing review feedback.

Important

Before merging: every file needs a license header and every commit must be DCO signed-off — see the 📖 Contribution guide for details.

@ramanathan831

Copy link
Copy Markdown
Contributor Author

/build

1 similar comment
@christinayyw

Copy link
Copy Markdown
Contributor

/build

Signed-off-by: Ramanathan Arunachalam <rarunachalam@nvidia.com>
@ramanathan831
ramanathan831 force-pushed the dev/ram/cache-pytorch-builds branch from 4ebb336 to 3f6a4f6 Compare August 10, 2026 03:58
@ramanathan831
ramanathan831 marked this pull request as draft August 10, 2026 04:24

@vpraveen-nv vpraveen-nv left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

not sure if this MR is a blocker, but thanks for putting this together.

Comment thread scripts/cuda_cache.sh Outdated
Comment thread scripts/cuda_cache.sh
Comment thread scripts/cuda_cache.sh
Comment thread release/docker/deploy.sh Outdated
Signed-off-by: Ramanathan Arunachalam <rarunachalam@nvidia.com>
Signed-off-by: Ramanathan Arunachalam <rarunachalam@nvidia.com>
@ramanathan831

Copy link
Copy Markdown
Contributor Author

/build

@vpraveen-nv vpraveen-nv left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

All comments addressed — header + toolchain identity folded into the cache key, writability guard before destructive ops, and full-rebuild fallback in deploy.sh. LGTM.

@ramanathan831

Copy link
Copy Markdown
Contributor Author

/build

2 similar comments
@hqtran-nv

Copy link
Copy Markdown
Contributor

/build

@hqtran-nv

Copy link
Copy Markdown
Contributor

/build

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.

4 participants