Saving intermediate training results in case of a long training

When I am training my network for a large number of epochs and each epoch takes a significant time, in case I already getting quite good results after few epochs is there a way to save the intermediate trained network model?

If it does not happen automatically what happens if I just stop training in the middle and save whatever I have at this moment, would it work?
Can I continue to train my network after this interruption again?

1 Like

I think you are looking for this https://docs.fast.ai/callbacks.html#SaveModelCallback :slight_smile:

3 Likes