epeters3 / autotensor Goto Github PK

Once the package finds the best model it can, that model should be persisted for the user to be able to consume in the wild. Include an output statement about the persisted model's path. When used from other python code, the method can return the best model, or a path to it.

When run, currently the package's dependencies are printing deprecation warnings and other warnings. Also, once the package finishes it's greedy search of hyperparameters, it throws a json serialization error when trying to produce the model report.

Currently, the package can only do hyperparameter turning. We want to do a more full neural architecture search where different numbers and types of layers can be tried.

Currently, the CLI only supports arff files. It would be great to support CSV files since they're so common.

When run, currently the package's dependencies are printing deprecation warnings and other warnings. Also, once the package finishes it's greedy search of hyperparameters, it throws a json serialization error when trying to produce the model report.