How to access the intermediate experiment result?

Ray Libraries (Data, Train, Tune, Serve) Ray Tune
1

How can access the intermediate results of the analysis step. So, currently I save for each trial its best corresponding model using the ModelCheckpoint. The drawback of this strategy is the big number of models that will be saved. How can for example access the result of the exucted trial at time t for example, and compare the val_loss(for example) and if the new model has a small val_loss then overwrite the old by this new one.

Thanks

2

You can try tune.run(keep_checkpoints_num=1, checkpoint_score_attr="val_loss" or something related.

3

is not working still i get all the models