accumulate_grad_batches seems not working #12166
Replies: 2 comments 3 replies
-
@amejri Seems like you're using an old version of PyTorch Lightning (<1.5.0 because the trainer flag |
Beta Was this translation helpful? Give feedback.
-
hey @amejri !
can you share more on this? how are you evaluating this conclusion? |
Beta Was this translation helpful? Give feedback.
-
Hi guys,
I'm trying to train a model with accumulating several batches for gradient. For that, I specified it in the trainer as follow :
but it seems like nothing changes. and the training evolution is the same as without adding accumulate_grad_batches parameter.
Do you have any idea how to solve it ?
With kind regards,
Beta Was this translation helpful? Give feedback.
All reactions