trw.train.callback_activation_statistics

Module Contents

Classes

CallbackActivationStatistics

Calculate activation statistics of each layer of a neural network.

Functions

model_summary(model, batch, logger)

Attributes

logger

trw.train.callback_activation_statistics.logger

trw.train.callback_activation_statistics.model_summary(model, batch, logger)

class trw.train.callback_activation_statistics.CallbackActivationStatistics(dataset_name=None, split_name='train', logger_fn=utilities.log_and_print)

Bases: trw.train.callback.Callback

Calculate activation statistics of each layer of a neural network.

This can be useful to detect connectivity issues within the network, overflow and underflow which may impede the training of the network.

first_time(self, datasets)

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