[trainer] add tf32-mode control (#14606)

Commit

4 years ago

[trainer] add tf32-mode control (#14606) * [trainer] add --tf32 support * it's pt>=.17 * it's pt>=.17 * flip the default to True * add experimental note * simplify logic * style * switch to 3-state logic * doc * Apply suggestions from code review Co-authored-by: Sylvain Gugger <35901082+sgugger@users.noreply.github.com> * re-style code Co-authored-by: Sylvain Gugger <35901082+sgugger@users.noreply.github.com>

References

#14606 - [trainer] add tf32-mode control

#15748 - Fix segformer reshape last stage

#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

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

#41212 - Add EoMT with DINOv3 backbone