Is there a "learner.destroy" in v2?

fastai users fastai dev
#1

I know this has been mentioned before … just don’t know if there is yet an answer :slight_smile:

Anyways, would be a definite plus to have so that we/I don’t have to restart our notebooks when we get those lovely CUDA OOM errors, or are re-using Learner objects in tests.

#2

As of yet no (I’ve been watching the commits), so just the del learn torch.cuda.empty_cache() combo (don’t know if it’s in Jeremy’s todo list though)

1 Like