At some point, I managed to configure verbosity, so that on a Jupyter Notebook it showed a pandas DataFrame-looking widget that updated with the tuning progress. Currently, I am getting a tremendous amount of logged stuff, like the one shown below, that does not seem to be easily configurable with the verbose parameter from the tune.run
method.
How can I achieve this again?
Result for cnn_ray_loss_wrapper_d2f06376:
AUPRC: 0.09326016157865524
AUROC: 0.5208970308303833
accuracy: 0.9120290279388428
balanced_accuracy: 0.5001329779624939
date: 2021-01-13_14-01-42
done: false
experiment_id: 3edbc04b763a40008e50589c6ce5e936
f1_score: 0.0022402687463909388
fall_out: 0.000865636917296797
hostname: fc60b1c56261
iterations_since_restore: 1
loss: 0.30360624670958813
mcc: 0.002518440131098032
miss_rate: 0.998868465423584
node_ip: 192.168.127.128
pid: 39725