facebookresearch / fairseq2

FAIR Sequence Modeling Toolkit 2
https://facebookresearch.github.io/fairseq2/
MIT License
613 stars 59 forks source link

Warn grads in broadcast_module #594

Closed cbalioglu closed 2 weeks ago

cbalioglu commented 2 weeks ago

A small PR that warns when the provided module to broadcast_module() has gradients.