Sometimes it can be nice to interrupt a model's training process, but still have it terminate cleanly, i.e. finish the current epoch and still get the scores and generate the output reports. My initial thought is to have something like:
try:
<code>
except KeyboardInterupt as e:
<wrap things up code>
but this would kill the current training epoch and scores would be hard to reproduce anyways. How do we handle this more softly?
Sometimes it can be nice to interrupt a model's training process, but still have it terminate cleanly, i.e. finish the current epoch and still get the scores and generate the output reports. My initial thought is to have something like:
but this would kill the current training epoch and scores would be hard to reproduce anyways. How do we handle this more softly?