Skip to content

Re-enable test_grad_scaling_state_dict_xpu - #5445

Draft
pponikox wants to merge 1 commit into
intel:mainfrom
pponikox:dev/pponikox/PYTORCHDGQ-9939
Draft

pponikox wants to merge 1 commit into
intel:mainfrom
pponikox:dev/pponikox/PYTORCHDGQ-9939

Conversation

@pponikox

Copy link
Copy Markdown
Contributor

Summary:

This patch fixes #5375 by removing the stale skip so the test runs in CI again.
The XPU override of test_grad_scaling_state_dict in test_torch_xpu.py already handles the XPU scale-tensor check, and the test passes on XPU. Its entry in skip_list_common.py is now stale.

Validation (Intel BMG):

python test_torch_xpu.py TestTorchDeviceTypeXPU.test_grad_scaling_state_dict_xpu -v
  • before: OK (run directly; skip entry present)
  • after: still OK (skip entry removed)

@pponikox pponikox self-assigned this Sep 18, 2026
@pponikox pponikox added the disable_auto Disable auto label workflow for PR label Sep 18, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

disable_auto Disable auto label workflow for PR

Projects

None yet

Development

Successfully merging this pull request may close these issues.

[upstream_ut] test_grad_scaling_state_dict_xpu: _scale is not torch.FloatTensor on XPU

1 participant