trw.train.callback_export_history

Module Contents

Classes

CallbackExportHistory

Summarize the training history of a model (i.e., as a function of iteration)

Functions

extract_from_history(history, dataset_name, output_name, value_name)	Extract a specified value for all the epochs and given the dataset, split, output and value_name
merge_history_values(history_values_list)	Merge several history values (e.g., multiple runs)
default_metrics()

Attributes

logger

trw.train.callback_export_history.logger

trw.train.callback_export_history.extract_from_history(history, dataset_name, output_name, value_name)

Extract a specified value for all the epochs and given the dataset, split, output and value_name :param history: the history :param dataset_name: the dataset :param output_name: the output name :param value_name: the values to extract :return: a dictionary of list (split_name, values)

trw.train.callback_export_history.merge_history_values(history_values_list)

Merge several history values (e.g., multiple runs) :param history_values_list: a list of history values :return: a dictionary of list of list of values (split name, list of history values)

trw.train.callback_export_history.default_metrics()

class trw.train.callback_export_history.CallbackExportHistory(export_dirname='history', metrics_to_report=default_metrics())

Bases: trw.train.callback.Callback

Summarize the training history of a model (i.e., as a function of iteration)

• One plot per dataset

• splits are plotted together

__call__(self, options, history, model, losses, outputs, datasets, datasets_infos, callbacks_per_batch, **kwargs)