OptimHistory.loss_df()

OptimHistory.loss_df()

Converts training and monitoring losses into a pandas.DataFrame.

Return type:

DataFrame

Returns:

pandas.DataFrame – Data frame with "epoch", "loss_train", and "loss_monitor" columns.

Examples

>>> import jax.numpy as jnp
>>> from liesel.optim import OptimHistory
>>> history = OptimHistory.from_epochs(epochs=2, position=None)
>>> history.loss_train = history.loss_train.at[0].set(1.5)
>>> history.loss_monitor = history.loss_monitor.at[0].set(2.5)
>>> history.loss_df().iloc[0].to_dict()
{'epoch': 0.0, 'loss_train': 1.5, 'loss_monitor': 2.5}