Well this is common thing to do for validation metrics. Training metrics are calculated before validation so it gives better visualisation.
I found this information there: https://twitter.com/aureliengeron/status/1110839480024338432?s=20
Hi ObnoxiousStork61
Is it possible to report ie. validation scalars but shifted by 1/2 iteration?
No 😞 these are integers
What's the reason for the shift?
I'm also curious 🙂
Interesting. Will probably only matter for very small experiments or experiments, where validation is run very infrequently.
Trying this gives me expected TypeError: TypeError: Expected iter of type (<class 'int'>,), got float
Epochs are still round numbers ...
Multiply by 2?! 😅
This will work too, ill try.
My experiment has validation every 40k iterations so thats the case ;p.