Skip to content
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.

Sign up for GitHub

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

Jump to bottom

ConvTranspose (auto_pad) #839

Closed
Tracked by #828
zjgarvey opened this issue Sep 18, 2024 · 2 comments
Closed
Tracked by #828

ConvTranspose (auto_pad) #839

zjgarvey opened this issue Sep 18, 2024 · 2 comments
Assignees
@guacamoleo

Comments

@zjgarvey
Copy link
Collaborator

zjgarvey commented Sep 18, 2024 •
edited
Loading

Assigned to Luka Tsabadze
@zjgarvey zjgarvey mentioned this issue Sep 18, 2024
Open
25 tasks
@guacamoleo guacamoleo self-assigned this Oct 9, 2024
@guacamoleo guacamoleo mentioned this issue Oct 15, 2024
Merged
zjgarvey pushed a commit to llvm/torch-mlir that referenced this issue Oct 18, 2024
@guacamoleo
Adds onnx ConvTranspose support for autopadding. (#3797)
02327af 
Adds onnx ConvTranspose support for autopadding
(nod-ai/SHARK-ModelDev#839).

- Adds support for attribute auto_pad="SAME_UPPER" or "SAME_LOWER" which
will automatically calculate padding of input based on output shape.
- Adds support, during auto-padding, for output_shape=[H,W] which
overrides the default output shape of input_shape[i]*stride[i] (for
spatial dimensions only).
- Adds lit test for auto-padding.
- Tests are added by nod-ai/SHARK-TestSuite#370


NOTE: ConvTranspose still doesn't support asymmetric padding, therefore
multiple original onnx tests still won't pass.
@guacamoleo
Copy link

Support (and testing) has been added for ConvTranspose autopadding. The remaining unsupported component is asymmetric padding; however asymmetric padding may be more pervasive than just ConvTranspose, so someone else should create an issue for it.

@guacamoleo
Copy link

Closing turbine camp issue.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees

@guacamoleo guacamoleo

Labels
None yet
Projects
None yet
Milestone
No milestone
Development

No branches or pull requests
2 participants
@guacamoleo @zjgarvey