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
TF-TRT SpaceToDepth and DepthToSpace op converters for dynamic shape mode. #47590
TF-TRT SpaceToDepth and DepthToSpace op converters for dynamic shape mode. #47590
Conversation
cdc9d9e
to
4749454
Compare
4749454
to
026b878
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.
Thanks @bixia1 for the review, I have addressed the issues.
026b878
to
2b283d8
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.
Thanks @bixia1 for the review, I have addressed the issues!
Prefer using return values over output parameters: they improve readability, and often provide the same or better performance |
This PR updates the converter for the SpaceToDepth and DepthToSpace ops to work with explicit batch and dynamic shape mode.
While the explicit batch converter needs only a slight modification to insert the batch dim, the dynamic shape converter has to redefine the same target shape calculation using TRT shape tensor arithmetic.
The unit test are updated to run all three trt modes. The parameters are not modified, just refactored to allow execution in all trt modes.
Tagging @bixia1 for review and @DEKHTIARJonathan for visibility.
Tracker: #45481