Let's not cast them all (#18471)

Commit

3 years ago

Let's not cast them all (#18471) * add correct dtypes when checking for params dtype * forward contrib credits * Update src/transformers/modeling_utils.py Co-authored-by: Thomas Wang <24695242+thomasw21@users.noreply.github.com> * more comments - added more comments on why we cast only floating point parameters * Update src/transformers/modeling_utils.py Co-authored-by: Sylvain Gugger <35901082+sgugger@users.noreply.github.com> Co-authored-by: sgugger <35901082+sgugger@users.noreply.github.com> Co-authored-by: Thomas Wang <24695242+thomasw21@users.noreply.github.com>

References

#18471 - Let's not cast them all

#19449 - [WIP] Fix weights initialization of several vision models

#27720 - Add common processor tests

#29969 - [SigLIP] Add fast tokenizer

#32831 - [Docs] Update resources

#33111 - [Backbone] Remove out_features everywhere

#33174 - [Zero-shot image classification pipeline] Remove tokenizer_kwargs

#39821 - Support MetaCLIP 2

#58 - Add EoMT DINOv3 model

#59 - Fix attention mask handling in EoMT-DINOv3 converter

#41212 - Add EoMT with DINOv3 backbone