-
Notifications
You must be signed in to change notification settings - Fork 7.2k
Affine CV-CUDA Backend #9294
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
base: main
Are you sure you want to change the base?
Affine CV-CUDA Backend #9294
Conversation
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/vision/9294
Note: Links to docs will display an error until the docs builds have been completed. ❗ 1 Active SEVsThere are 1 currently active SEVs. If your PR is affected, please view them below: ❌ 5 New FailuresAs of commit 77e800b with merge base aa35ca1 ( NEW FAILURES - The following jobs have failed:
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
d2f72db to
1860d73
Compare
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I left some comments for the first round review. Feel free to let me know your thoughts.
| if make_input is make_image_cvcuda: | ||
| # CV-CUDA nearest interpolation does not follow same algorithm as PIL/torch | ||
| assert mae < 255 if interpolation is transforms.InterpolationMode.NEAREST else 1, f"mae: {mae}" | ||
| else: | ||
| assert mae < 2 if interpolation is transforms.InterpolationMode.NEAREST else 8, f"mae: {mae}" |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
enum comparison is preferred to use "==".
Also see comment here
The threshold number might need more investigations. @NicolasHug
|
|
||
| interp = _get_cvcuda_interp(interpolation) | ||
|
|
||
| xform = np.array([[matrix[0], matrix[1], matrix[2]], [matrix[3], matrix[4], matrix[5]]], dtype=np.float32) |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
We can rewrite this to np.array(matrix, dtype=np.float32).reshape(2, 3) for simplicity.
Summary
Affine CV-CUDA backend
Testing
python3 -m pytest test/test_transforms_v2.py::TestAffine