nntoolbox.callbacks.nan module

class nntoolbox.callbacks.nan.NaNWarner

Bases: nntoolbox.callbacks.callbacks.Callback

on_batch_end(logs: Dict[str, Any])

class nntoolbox.callbacks.nan.SkipNaN

Bases: nntoolbox.callbacks.callbacks.Callback

Skip when loss or output is nan (UNTESTED)

after_losses(losses: Dict[str, torch.Tensor], train: bool) → bool

after_outputs(outputs: Dict[str, torch.Tensor], train: bool) → bool