With the evolving and increasing training and validation datasets it is becoming increasingly difficult to evaluate models consistently and with equal metrics, making evaluation subjective when looking at plots.
This tasks aims to alleviate this by fixing the training and validation datasets to representative data, and re-using the same datasets across all future models.
This is tied to the Define fixed validation and test metrics task which aims to do the same.
Outcome
Completed and defined in Dipole datasets.