Can I use callbacks?

PL 0.13.1

Question expecting the answer no - at least for now :wink:

I am investigating generalisation and I would like to be able to do some predictions at the end of each epoch (to be written to file) - is there some way I could do that with a callback - or any other way?

Hey @JulianSMoore,
Unfortunately I think it will be difficult right now. You can of course customize the components (and should be able to make them write to a file) but they don’t have any knowledge of which iteration you are at.
The best you could try to do is to write all outputs from the last component to a file by keep appending them to the end of the file.

This will be a lot easier as soon as we open up customization in the training loop, but that’s still a bit away.

OK thanks - good to know where we are.