Skip to content

[ci] Validate Attention #6508 with upstream Layer 2 - #6517

Merged
yuxuan-lou merged 1 commit into
mainfrom
fix/attention-layer2-81c2cfb0
Nov 18, 2026
Merged

yuxuan-lou merged 1 commit into
mainfrom
fix/attention-layer2-81c2cfb0

Conversation

@b1ack-j

@b1ack-j b1ack-j commented Nov 10, 2026

Copy link
Copy Markdown
Collaborator

Purpose

Run the existing trusted same-repository Layer 2 checks for #6508, whose fork-origin GPU jobs are skipped.

This is a validation-only duplicate. Keep #6508 as the review and merge target; do not merge this validation PR.

Source

Validation requested

Run the unchanged CI / Layer 2 / Trusted PR GPU workflow: batches 2 4a 5, with its existing FAST_MODE setting, pinned dependency image, shared GPU lock, and idle-device checks. Batch 5 uses two idle GPUs; the other batches use one.

This is the standard core GPU regression. It does not include the separate strict FA3 Attention batch 4f, and is not a replacement for the operator numerical acceptance evidence documented in #6508. Results will be available in this PR's Checks and the workflow artifacts.

@b1ack-j
b1ack-j requested a review from a team as a code owner November 10, 2026 12:41
@yuxuan-lou
yuxuan-lou merged commit 81c2cfb into main Nov 18, 2026
29 of 37 checks passed
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.

3 participants