src.superphot_plus.config

Module Contents

Classes

SuperphotConfig

Holds information about the specific training

class SuperphotConfig

Holds information about the specific training configuration of a model. The default values are sampled by ray tune for parameter optimization.

create_dirs: bool | None = True

relative_dirs: bool | None = True

data_dir: str | None = '.'

fits_dir: str | None = 'fits'

input_csvs: list | None

models_dir: str | None = 'models'

figs_dir: str | None = 'figs'

metrics_dir: str | None = 'metrics'

fit_plots_dir: str | None = 'fits'

cm_dir: str | None = 'confusion_matrices'

wrongly_classified_dir: str | None = 'wrongly_classified'

log_fn: str | None = 'results.log'

probs_dir: str | None = 'probabilities'

probs_fn: str | None = 'probs_%d.csv'

prefix: str | None = 'best-model'

target_label: str | None

prob_threshhold: float | None = 0.5

input_dim: int | None

output_dim: int | None

normalization_means: List[float] | None

normalization_stddevs: List[float] | None

neurons_per_layer: int | None

num_hidden_layers: int | None

goal_per_class: int | None = 4500

num_folds: int | None

num_epochs: int | None

batch_size: int | None

learning_rate: float | None

best_val_loss: float | None

device

__post_init__()

Ensure subdirectory structure exists.

set_non_tunable_params(input_dim, output_dim, norm_means, norm_stddevs)

Adds information about the params that are not tunable.

set_best_val_loss(best_val_loss)

Sets the best validation loss from training.

write_to_file(file: str)

Save configuration data to a YAML file.

classmethod from_file(file: str) → typing_extensions.Self

Load configuration data from a YAML file.