dmlcloud.TrainingHistory

class dmlcloud.TrainingHistory

Stores the training history of a model.

Metrics can either be ArrayLike objects or any pickleable object.

Usage:

history = TrainingHistory() history.append_metric(‘loss’, 0.5) history.append_metric(‘accuracy’, 0.99) history.next_step()

for metric in history:

print(f’{metric}’: history[metric])

__init__()

Methods

__init__()
append_metric(name, value)	Adds a value for a metric at the current step.
append_metrics(**metrics)	Adds multiple metrics at the current step.
current()	Returns the current, but not yet saved, value for each metric.
items()
keys()
last()	Returns the last value for each metric.
max()	Returns a namedtuple (value, step) containing the maximum value and the corresponding step for each metric across all steps.
min()	Returns a namedtuple (value, step) containing the minimum value and the corresponding step for each metric across all steps.
next_step()	Advances the step counter.
values()