training

General training functions

training.callbacks

Custom callbacks used for training

NeptuneModelLogger Objects

class NeptuneModelLogger(Callback)

Saves out the last and best models after each validation epoch.

If the files don't exists, does nothing.

Example:: from pl_bolts.callbacks import NeptuneModelLogger trainer = Trainer(callbacks=[NeptuneModelLogger()])

__init__

def __init__() -> None

Base initialization, nothing specific needed here

on_validation_epoch_end

def on_validation_epoch_end(trainer: Trainer,
                            pl_module: LightningModule) -> None

Save the best and last model checkpoints to Neptune after each validation

Arguments:

• trainer - PyTorchLightning trainer
• pl_module - LightningModule that is being trained

Returns:

None

on_fit_end

def on_fit_end(trainer: Trainer, pl_module: LightningModule) -> None

Save out the best and last model checkpoints at the end of trainer.fit to Neptune

Arguments:

• trainer - PyTorchLightning Trainer
• pl_module - LightningModule being used for training

Returns:

None