I have trained a document understanding custom model. i have retrained it multiple times and the latest minor version is 4 and the major version is 11. I know that during retraining we should always select the minor version as 0 (as mentioned in the docs) but should I follow the same advice during evaluation run and a Full run?
@shrey.shah Since it is always recomended to retrain on the lower version (0) so the evaluation and full run also schedule on the same version. So that the model keeps on improving every time we run the pipeline
@ushu Ok. Also how do I evaluate the model? After training my model should I import a few unseen docs by marking “Make this an evaluation set” and then run an evaluation pipeline on the evaluated docs only?