Skip to content

Long time no sync#268

Closed
wz337 wants to merge 1 commit into
facebookresearch:mainfrom
wz337:export-D112398609
Closed

Long time no sync#268
wz337 wants to merge 1 commit into
facebookresearch:mainfrom
wz337:export-D112398609

Conversation

@wz337

@wz337 wz337 commented Jul 16, 2026

Copy link
Copy Markdown
Contributor

Summary:
Re-sync of distributed_shampoo + gpa dev source to OSS
(facebookresearch/optimizers), regenerated via oss_sync.sh on top of the CI
fixes in the parent diff. Supersedes the prior sync (D112246130), which was red.

The two failures on the prior sync are fixed by the parent diff and included here:

  • gpu-tests: train_model now relocates pre-generated data onto the model's device
    (fixes 114 "tensors on different device" errors).
  • pre-commit: recompilation-test license merged into its docstring (fixes ruff
    E402 x16 caused by the sync prepending a second license docstring).

Validated locally against the OSS venv on H100:

  • ruff check distributed_shampoo: all checks passed
  • gpu_tests via torchrun --nproc_per_node=2: Ran 258 tests, OK (was 114 errors)
  • weight_decay_test: Ran 80 tests, OK
  • mypy: Success, no issues in 95 source files

Differential Revision: D112398609

Summary:
Re-sync of distributed_shampoo + gpa dev source to OSS
(facebookresearch/optimizers), regenerated via oss_sync.sh on top of the CI
fixes in the parent diff. Supersedes the prior sync (D112246130), which was red.

The two failures on the prior sync are fixed by the parent diff and included here:
- gpu-tests: train_model now relocates pre-generated data onto the model's device
  (fixes 114 "tensors on different device" errors).
- pre-commit: recompilation-test license merged into its docstring (fixes ruff
  E402 x16 caused by the sync prepending a second license docstring).

Validated locally against the OSS venv on H100:
- ruff check distributed_shampoo: all checks passed
- gpu_tests via torchrun --nproc_per_node=2: Ran 258 tests, OK (was 114 errors)
- weight_decay_test: Ran 80 tests, OK
- mypy: Success, no issues in 95 source files

Differential Revision: D112398609
@meta-cla meta-cla Bot added the CLA Signed This label is managed by the Facebook bot. Authors need to sign the CLA before a PR can be reviewed. label Jul 16, 2026
@meta-codesync

meta-codesync Bot commented Jul 16, 2026

Copy link
Copy Markdown

@wz337 has exported this pull request. If you are a Meta employee, you can view the originating Diff in D112398609.

@meta-codesync meta-codesync Bot closed this in 5b6bd7e Jul 17, 2026
@meta-codesync meta-codesync Bot added the Merged label Jul 17, 2026
@meta-codesync

meta-codesync Bot commented Jul 17, 2026

Copy link
Copy Markdown

This pull request has been merged in 5b6bd7e.

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. Merged meta-exported

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant