-
Notifications
You must be signed in to change notification settings - Fork 2k
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
When Recurrence meets Transformers to keras 3.0 #1984
Conversation
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.
Thanks for the PR! I'm not seeing any usage of TF in the example, so does it run on JAX and torch?
The patching operation is done using a
keras.layers.Conv2D
instance instead of a
traditionaltf.image.extract_patches
to allow for vectorization.
Please update this -- no reference to TF needed.
@fchollet, thanks for the review.
It now supports all backends.
Done. |
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.
Looks great -- please add the generated ipynb and md files!
@@ -925,6 +981,12 @@ index = random.randint(0, config["batch_size"]) | |||
orig_image = images[index] | |||
overlay_image = upsampled_heat_map[index, ..., 0] | |||
|
|||
if keras.backend.backend() == "torch": |
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.
Please note, if you use ops.convert_to_numpy()
you don't need any torch specific code. This is not a blocker so we can merge regardless.
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.
LGTM, thanks!
This PR changes the When Recurrence meets Transformers to keras 3.0, as requested in keras-team/keras-cv#2211
Please review the attached gist