Skip to content

Conversation

@NicolasHug
Copy link
Member

As per @datumbox 's offline discussion, pretrained=True is equivalent to the V1 weights. This PR adds this information in the docs on all V1 weights.

image

datumbox
datumbox previously approved these changes May 11, 2022
Copy link
Contributor

@datumbox datumbox left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM, great idea.

@datumbox datumbox dismissed their stale review May 11, 2022 15:37

corner-cases

@datumbox
Copy link
Contributor

As discussed offline with @NicolasHug, there a few corner-cases such as IMAGENET1K_SWAG_E2E_V1 and IMAGENET1K_SWAG_LINEAR_V1 for which this approach won't work. I agree this feature is worth considering but not 100% sure how we can offer it. The easiest, none hacky way I can think of is to keep a mapping hidden somehwere until the deprecation takes place.

@NicolasHug
Copy link
Member Author

The easiest, none hacky way I can think of is to keep a mapping hidden somehwere until the deprecation takes place.

I agree. Not sure when this will be available, so I'll close this PR for now. But I think it would be valuable info to document, so let's just keep it in the back of our mind

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

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants