get_best_model_version

behavenet.fitting.utils.get_best_model_version(expt_dir, measure='val_loss', best_def='min', n_best=1)

Get best model version from a test tube experiment.

Parameters:

• expt_dir (str) – test tube experiment directory containing version_%i subdirectories

• measure (str, optional) – heading in csv file that is used to determine which model is best

• best_def (str, optional) – how measure should be parsed; ‘min’ | ‘max’

• n_best (int, optional) – top n_best models are returned

Returns:

list of best models, with best first

Return type:

list