scvi.train.AdversarialTrainingPlanConfig

class scvi.train.AdversarialTrainingPlanConfig(optimizer='Adam', optimizer_creator=None, lr=0.001, weight_decay=1e-06, n_steps_kl_warmup=None, n_epochs_kl_warmup=400, reduce_lr_on_plateau=False, lr_factor=0.6, lr_patience=30, lr_threshold=0.0, lr_scheduler_metric='elbo_validation', lr_min=0.0, adversarial_classifier=False, scale_adversarial_loss='auto', compile=False, compile_kwargs=None, loss_kwargs=<factory>)

Config for AdversarialTrainingPlan.

Attributes table

adversarial_classifier
compile
compile_kwargs
lr
lr_factor
lr_min
lr_patience
lr_scheduler_metric
lr_threshold
n_epochs_kl_warmup
n_steps_kl_warmup
optimizer
optimizer_creator
reduce_lr_on_plateau
scale_adversarial_loss
weight_decay
loss_kwargs

Methods table

to_kwargs()

Attributes

AdversarialTrainingPlanConfig.adversarial_classifier: bool | Any = False

AdversarialTrainingPlanConfig.compile: bool = False

AdversarialTrainingPlanConfig.compile_kwargs: dict | None = None

AdversarialTrainingPlanConfig.lr: float = 0.001

AdversarialTrainingPlanConfig.lr_factor: float = 0.6

AdversarialTrainingPlanConfig.lr_min: float = 0.0

AdversarialTrainingPlanConfig.lr_patience: int = 30

AdversarialTrainingPlanConfig.lr_scheduler_metric: Literal['elbo_validation', 'reconstruction_loss_validation', 'kl_local_validation'] = 'elbo_validation'

AdversarialTrainingPlanConfig.lr_threshold: float = 0.0

AdversarialTrainingPlanConfig.n_epochs_kl_warmup: int | None = 400

AdversarialTrainingPlanConfig.n_steps_kl_warmup: int | None = None

AdversarialTrainingPlanConfig.optimizer: Literal['Adam', 'AdamW', 'Custom'] = 'Adam'

AdversarialTrainingPlanConfig.optimizer_creator: Callable[[Iterable[Tensor]], Optimizer] | None = None

AdversarialTrainingPlanConfig.reduce_lr_on_plateau: bool = False

AdversarialTrainingPlanConfig.scale_adversarial_loss: Union[float, Literal['auto']] = 'auto'

AdversarialTrainingPlanConfig.weight_decay: float = 1e-06

AdversarialTrainingPlanConfig.loss_kwargs: dict[str, Any]

Methods

AdversarialTrainingPlanConfig.to_kwargs()

Return type:

dict[str, Any]