orcanet.utilities.summarize_training

Module Contents

Classes

Summarizer

Summarize one or more trainings by giving their orcanet folder(s).

Functions

summarize(**kwargs)
get_parser()
main()

class orcanet.utilities.summarize_training.Summarizer(folders, metric='loss', smooth=None, labels=None, noplot=False, width=1.0, verbose=True)

Summarize one or more trainings by giving their orcanet folder(s).

• Plot the training and validation curves in a single plot and show them

• Print info about the best and worst epochs

Parameters

foldersstr or List, optional

Path to a orcanet folder, or to multiple folder as a list. [default: CWD].

metricstr or List, optional

The metric to plot [default: ‘loss’]. If its a list: Same length as folders. Plot a different metric for each folder.

smoothint, optional

Apply gaussian blur to the train curve with given sigma.

labelsList, optional

Labels for each folder.

noplotbool

Dont plot the train/val curves [default: False].

widthfloat

Scaling of the width of the curves and the marker size [default: 1].

summarize(show=True)

summarize_dirs()

Get the best and worst epochs of all given folders as a dict.

Returns

minimadict

Keys : Name of folder. Values : [Epoch, metric] of where the metric is lowest.

maximadict

As above, but for where the metric is highest.

orcanet.utilities.summarize_training.summarize(**kwargs)

orcanet.utilities.summarize_training.get_parser()

orcanet.utilities.summarize_training.main()