sup3r.models.utilities.TrainingConfig

class TrainingConfig(input_resolution: dict | None = None, n_epoch: int | None = None, weight_gen_advers: float = 0.001, train_gen: bool = True, train_disc: bool = True, disc_loss_bounds: tuple = (0.45, 0.6), checkpoint_int: int = 10, out_dir: str | None = None, early_stop_on: str | None = None, early_stop_threshold: float = 0.005, early_stop_n_epoch: int = 5, adaptive_update_bounds: tuple = (0.0, 1.0), adaptive_update_fraction: float = 0.0, multi_gpu: bool = False, log_tb: bool = False, export_tb: bool = False)

Bases: object

Shared training configuration for trainable sup3r models.

Methods

for_conditional([config])	Build a conditional-model training config with defaults.
for_gan([config])	Build a GAN training config with GAN-specific defaults.
from_train_kwargs([config])	Build a training config from an existing config and/or kwargs.

Attributes

CONDITIONAL_DEFAULTS
GAN_DEFAULTS
adaptive_update_bounds
adaptive_update_fraction
checkpoint_int
disc_loss_bounds
early_stop_n_epoch
early_stop_on
early_stop_threshold
export_tb
input_resolution
log_tb
multi_gpu
n_epoch
out_dir
train_disc
train_gen
weight_gen_advers

classmethod from_train_kwargs(config=None, **kwargs)

Build a training config from an existing config and/or kwargs.

classmethod for_gan(config=None, **kwargs)

Build a GAN training config with GAN-specific defaults.

classmethod for_conditional(config=None, **kwargs)

Build a conditional-model training config with defaults.