Open bryant1410 opened 4 months ago
I think shuffling only during training makes it better for debugging on the validation data and more predictable.
I think shuffling only during training makes it better for debugging on the validation data and more predictable.