[Tune] output of ray.tuner.fit().get_best_result

[enori]

Description

results_grids = ray.tuner.fit()
best_results = results_grids.get_best_result(
    metric="mean_accuracy",
    mode="max"
)

print("Best config: ", best_results)

After performing hyperparameter tuning with the above code, the output results were as follows:

Best config:  Result(
  metrics={'mean_accuracy': 0.965625, 'lr': 0.12270249077413087},
  path='ray-results/pbt_test/train_convnet_03fb1_00001_1_lr=0.1227,momentum=0.8192_2024-12-17_02-45-22',
  filesystem='s3',
  checkpoint=Checkpoint(filesystem=s3, path=ray-results/pbt_test/train_convnet_03fb1_00001_1_lr=0.1227,momentum=0.8192_2024-12-17_02-45-22/checkpoint_000007)
)

I was confused because lr (learning_rate) was included in the metrics.
Additionally, I expected the hyperparameters to be displayed in the output, but they were not. Instead, they were stored in the file ray-results/pbt_test/train_convnet_03fb1_00001_1_lr=0.1227,momentum=0.8192_2024-12-17_02-45-22/params.json, which required accessing the file.

Therefore, I was wondering if the output could be changed as follows:

Best config:  Result(
  metrics={'mean_accuracy': 0.965625, 'loss': xxxxxxx, 'mean_absolute_error': xxxxxxx},
  params={'checkpoint_interval': 5, 'lr': 0.12270249077413087, 'momentum': 0.8192463882731192},
  path='ray-results/pbt_test/train_convnet_03fb1_00001_1_lr=0.1227,momentum=0.8192_2024-12-17_02-45-22',
  filesystem='s3',
  checkpoint=Checkpoint(filesystem=s3, path=ray-results/pbt_test/train_convnet_03fb1_00001_1_lr=0.1227,momentum=0.8192_2024-12-17_02-45-22/checkpoint_000007)
)

Thank you always for your amazing supports and Ray!