Skip to content

Conversation

@vmoens
Copy link
Collaborator

@vmoens vmoens commented Jan 8, 2026

Stack from ghstack (oldest at bottom):


  • Skip data-dependent truncation in vec_generalized_advantage_estimate during compile
  • Use expand instead of full_like to avoid .item() calls with unbacked symbols
  • Use torch.full(..., device=device) to avoid DeviceCopy in cudagraph
  • Skip validation during compile to avoid CUDA syncs

[ghstack-poisoned]
@pytorch-bot
Copy link

pytorch-bot bot commented Jan 8, 2026

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/rl/3303

Note: Links to docs will display an error until the docs builds have been completed.

❌ 1 New Failure, 23 Pending

As of commit 1c724c3 with merge base df00d61 (image):

NEW FAILURE - The following job has failed:

This comment was automatically generated by Dr. CI and updates every 15 minutes.

vmoens added 2 commits January 8, 2026 08:42
[ghstack-poisoned]
[ghstack-poisoned]
vmoens added a commit that referenced this pull request Jan 8, 2026
- Skip data-dependent truncation in vec_generalized_advantage_estimate during compile
- Use expand instead of full_like to avoid .item() calls with unbacked symbols
- Use torch.full(..., device=device) to avoid DeviceCopy in cudagraph
- Skip validation during compile to avoid CUDA syncs

ghstack-source-id: 519e9b8
Pull-Request: #3303
@vmoens vmoens merged commit 1c724c3 into gh/vmoens/184/base Jan 8, 2026
80 of 98 checks passed
@vmoens vmoens deleted the gh/vmoens/184/head branch January 8, 2026 09:06
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed.

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants