Model contrib guidelines #5315
Model contrib guidelines #5315
Conversation
💊 CI failures summary and remediationsAs of commit 8df96a9 (more details on the Dr. CI page): ✅ None of the CI failures appear to be your fault 💚
🚧 1 ongoing upstream failure:These were probably caused by upstream breakages that are not fixed yet.
This comment was automatically generated by Dr. CI (expand for details).Please report bugs/suggestions to the (internal) Dr. CI Users group. |
|
@xiaohu2015 @zhiqwang since you have been through this process recently your feedback would be very valuable :) |
There was a problem hiding this comment.
Hi @jdsgomes , This guidelines look very useful and informative to me. I think one thing needs to be supplemented is that the added model needs to provide at least torchscript export support (especially the torch.jit.script mode as torchvision is currently using).
CONTRIBUTING_MODELS.md
Outdated
| - Highlight the important test metrics and how they compare to the original paper | ||
| - Highlight any design choices that deviate from the original paper/implementation and rationale for these choices | ||
|
|
||
| ## New Model Architectures - Implementation Details |
There was a problem hiding this comment.
We've discussed with Joao the possibility of moving this section out of the guideline doc (perhaps to an issue or a wiki article?). This is because it's quite low-level and potentially subjective to regular change. I would like to hear the thoughts of the community on this.
There was a problem hiding this comment.
I vote for issue because it allows for further discussion.
There was a problem hiding this comment.
agree, moved to a new issue for now, if we decide to move it back later that's ok: #5319
There was a problem hiding this comment.
Looks good. Worth adding on the new issue an explicit note about JIT-scriptability as @zhiqwang said. Going forwards we also expect all models to have scriptable preprocessing transforms as part of the multi-weight API, though I would probably omit this until the API graduates from prototype.
Co-authored-by: Vasilis Vryniotis <[email protected]>
|
Hey @jdsgomes! You merged this PR, but no labels were added. The list of valid labels is available at https://github.com/pytorch/vision/blob/main/.github/process_commit.py |
Summary: * Adding multiweight support for shufflenetv2 prototype models * Revert "Adding multiweight support for shufflenetv2 prototype models" This reverts commit 31fadbe. * Adding multiweight support for shufflenetv2 prototype models * Revert "Adding multiweight support for shufflenetv2 prototype models" This reverts commit 4e3d900. * Remove module vs method name clash * add model contribution guidelines * update CONTRIBUTING_MODELS.md * Fix formatting and typo * fix in-document links * Update CONTRIBUTING.md * remove Implementation Details section Reviewed By: NicolasHug Differential Revision: D34140246 fbshipit-source-id: 0c96404ac238bb039b0e2d057c7db961a6ab0512 Co-authored-by: Vasilis Vryniotis <[email protected]> Co-authored-by: Vasilis Vryniotis <[email protected]>
New model contribution guidelines that aim to clarify the process for adding new model architectures or improved model weights.
Feedback welcome, as always, before it gets merged.
cc: @datumbox @xiaohu2015 @zhiqwang