def chemprop_hyperopt() -> None:
"""Runs hyperparameter optimization for chemprop model."""
grid_search(HyperoptArgs().parse_args())
def chemprop_hyperopt() -> None:
"""Runs hyperparameter optimization for a Chemprop model.
This is the entry point for the command line command :code:`chemprop_hyperopt`.
"""
hyperopt(args=HyperoptArgs().parse_args())
fmin(objective,
SPACE,
algo=tpe.suggest,
max_evals=args.num_iters,
rstate=np.random.RandomState(args.seed))
# Report best result
results = [
result for result in results if not np.isnan(result['mean_score'])
]
best_result = min(results,
key=lambda result:
(1
if args.minimize_score else -1) * result['mean_score'])
logger.info('best')
logger.info(best_result['hyperparams'])
logger.info(f'num params: {best_result["num_params"]:,}')
logger.info(
f'{best_result["mean_score"]} +/- {best_result["std_score"]} {args.metric}'
)
# Save best hyperparameter settings as JSON config file
makedirs(args.config_save_path, isfile=True)
with open(args.config_save_path, 'w') as f:
json.dump(best_result['hyperparams'], f, indent=4, sort_keys=True)
if __name__ == '__main__':
grid_search(HyperoptArgs().parse_args())