direct.config package

Submodules

direct.config.defaults module

class direct.config.defaults.CheckpointerConfig(checkpoint_steps: int = 500)

Bases: BaseConfig

checkpoint_steps: int = 500

class direct.config.defaults.DefaultConfig(model: direct.config.defaults.ModelConfig = '???', additional_models: Optional[Any] = None, physics: direct.config.defaults.PhysicsConfig = <factory>, training: direct.config.defaults.TrainingConfig = <factory>, validation: direct.config.defaults.ValidationConfig = <factory>, inference: Optional[direct.config.defaults.InferenceConfig] = None, logging: direct.config.defaults.LoggingConfig = <factory>)

Bases: BaseConfig

additional_models: Optional[Any] = None

inference: Optional[InferenceConfig] = None

logging: LoggingConfig

model: ModelConfig = '???'

physics: PhysicsConfig

training: TrainingConfig

validation: ValidationConfig

class direct.config.defaults.FunctionConfig(function: str = '???', multiplier: float = 1.0)

Bases: BaseConfig

function: str = '???'

multiplier: float = 1.0

class direct.config.defaults.InferenceConfig(dataset: direct.data.datasets_config.DatasetConfig = <factory>, batch_size: int = 1, crop: Optional[str] = None)

Bases: BaseConfig

batch_size: int = 1

crop: Optional[str] = None

dataset: DatasetConfig

class direct.config.defaults.LoggingConfig(log_as_image: Optional[List[str]] = None, tensorboard: direct.config.defaults.TensorboardConfig = <factory>)

Bases: BaseConfig

log_as_image: Optional[List[str]] = None

tensorboard: TensorboardConfig

class direct.config.defaults.LossConfig(crop: Optional[str] = None, losses: List[Any] = <factory>)

Bases: BaseConfig

crop: Optional[str] = None

losses: List[Any]

class direct.config.defaults.ModelConfig(model_name: str = '???', engine_name: str | None = None)

Bases: BaseConfig

engine_name: Optional[str] = None

model_name: str = '???'

class direct.config.defaults.PhysicsConfig(forward_operator: str = 'fft2', backward_operator: str = 'ifft2', use_noise_matrix: bool = False, noise_matrix_scaling: float | None = 1.0)

Bases: BaseConfig

backward_operator: str = 'ifft2'

forward_operator: str = 'fft2'

noise_matrix_scaling: Optional[float] = 1.0

use_noise_matrix: bool = False

class direct.config.defaults.TensorboardConfig(num_images: int = 8)

Bases: BaseConfig

num_images: int = 8

class direct.config.defaults.TrainingConfig(datasets: List[Any] = <factory>, model_checkpoint: Optional[str] = None, optimizer: str = 'Adam', lr: float = 0.0005, weight_decay: float = 1e-06, batch_size: int = 2, lr_step_size: int = 5000, lr_gamma: float = 0.5, lr_warmup_iter: int = 500, swa_start_iter: Optional[int] = None, num_iterations: int = 50000, validation_steps: int = 1000, gradient_steps: int = 1, gradient_clipping: float = 0.0, gradient_debug: bool = False, loss: direct.config.defaults.LossConfig = <factory>, checkpointer: direct.config.defaults.CheckpointerConfig = <factory>, metrics: List[str] = <factory>, regularizers: List[str] = <factory>)

Bases: BaseConfig

batch_size: int = 2

checkpointer: CheckpointerConfig

datasets: List[Any]

gradient_clipping: float = 0.0

gradient_debug: bool = False

gradient_steps: int = 1

loss: LossConfig

lr: float = 0.0005

lr_gamma: float = 0.5

lr_step_size: int = 5000

lr_warmup_iter: int = 500

metrics: List[str]

model_checkpoint: Optional[str] = None

num_iterations: int = 50000

optimizer: str = 'Adam'

regularizers: List[str]

swa_start_iter: Optional[int] = None

validation_steps: int = 1000

weight_decay: float = 1e-06

class direct.config.defaults.ValidationConfig(datasets: List[Any] = <factory>, batch_size: int = 8, metrics: List[str] = <factory>, regularizers: List[str] = <factory>, crop: Optional[str] = 'training')

Bases: BaseConfig

batch_size: int = 8

crop: Optional[str] = 'training'

datasets: List[Any]

metrics: List[str]

regularizers: List[str]

Module contents

class direct.config.BaseConfig

Bases: object