I just wasted ~12 hours because ./rs train ran for a single epoch and then right at the end it asserted in def validate on:
assert num_samples > 0, "dataset contains training images and labels"
This is of course not ideal at all, it would be best for things like that to be checked before any training work is done.
I just wasted ~12 hours because ./rs train ran for a single epoch and then right at the end it asserted in
def validate
on:assert num_samples > 0, "dataset contains training images and labels"
This is of course not ideal at all, it would be best for things like that to be checked before any training work is done.