You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I'm not sure if the title of this issue is right, so please change this to whatever you think is most appropriate... in short, I couldn't decipher the cause of the problem :smile
When running torch 2.3.0 against kornia test suite we got some new errors around dynamo (CI in kornia/kornia#2912). For the kornia.geometry.transform.homography_warp operator it's failing to run within batch size != 1.
It's working fine in the previous torch version, and now it's working on CUDA, but not on CPU. Eager mode is ok, and it's failing under the inductor backend.
* chore (CI): ensure support to pytorch 2.3.0
* chore: skip specific dynamo tests for torch 2.3.0
- Report in pytorch/pytorch#126617
* chore: skip specific dynamo tests for torch 2.3.0
- Report in pytorch/pytorch#126619
馃悰 Describe the bug
I'm not sure if the title of this issue is right, so please change this to whatever you think is most appropriate... in short, I couldn't decipher the cause of the problem :smile
When running torch 2.3.0 against kornia test suite we got some new errors around dynamo (CI in kornia/kornia#2912). For the
kornia.geometry.transform.homography_warp
operator it's failing to run within batch size != 1.It's working fine in the previous torch version, and now it's working on CUDA, but not on CPU. Eager mode is ok, and it's failing under the inductor backend.
The falling code snippet
Error logs
Traceback:
Minified repro
Versions
cc @ezyang @msaroufim @bdhirsh @anijain2305 @chauhang
The text was updated successfully, but these errors were encountered: