rai_toolbox.mushin.MetricsCallback

class rai_toolbox.mushin.MetricsCallback(save_dir='.', filename='metrics.pt')

Saves validation and test metrics stored in trainer.callback_metrics.

Parameters:

save_dirstr, optional (default=”.”)

filenamestr, optional (default=”metrics.pt”)

The base filename used to store metrics. For FITTING the file is prepended with “fit_” and for TESTING the file is prepended with test_.

Notes

No metrics will be saved during FITTING if no validation metrics are calculated. This is a limitation of PyTorch Lightning. Future versions will save the training step metrics when no validation metrics are calculated.

Examples

>>> from pytorch_lightning import Trainer
>>> from rai_toolbox.mushin import MetricsCallback

>>> metrics_callback = MetricsCallback()
>>> trainer = Trainer(callbacks=[metrics_callback])

__init__(save_dir='.', filename='metrics.pt')

Methods

__init__([save_dir, filename])