winglian commited on
Commit
abe5380
1 Parent(s): 3babc68

reduce logging verbosity

Browse files

I think you forgot to transfer over the logging change too from https://huggingface.co/LnL-AI/dbrx-base-converted/discussions/1

Files changed (1) hide show
  1. modeling_dbrx.py +1 -1
modeling_dbrx.py CHANGED
@@ -412,7 +412,7 @@ class DbrxFlashAttention2(DbrxAttention):
412
  **kwargs: Any,
413
  ) -> Tuple[torch.Tensor, Optional[torch.Tensor],
414
  Optional[Tuple[torch.Tensor]]]:
415
- logger.info(
416
  'Implicitly setting `output_attentions` to False as it is not supported in Flash Attention.'
417
  )
418
  output_attentions = False
 
412
  **kwargs: Any,
413
  ) -> Tuple[torch.Tensor, Optional[torch.Tensor],
414
  Optional[Tuple[torch.Tensor]]]:
415
+ logger.debug(
416
  'Implicitly setting `output_attentions` to False as it is not supported in Flash Attention.'
417
  )
418
  output_attentions = False