main program

This module contains functions and classes for simplifying the training of ANN classifiers. It accepts the following arguments:

usage: main.py [-h] -m MODEL -d {MNIST,FashionMNIST,CIFAR10,CIFAR100}
               [-b BATCH_SIZE] [-e EPOCHS] [-o OPTIMIZER] [-l LEARNING_RATE]
               [-a N] [-s SUBSAMPLE] [-v {tb,mpl}] [-V]
               [--spectral-norm TOPIC [TOPIC ...]]
               [--variance TOPIC [TOPIC ...]] [--histogram TOPIC [TOPIC ...]]
               [--ratio TOPIC,TOPIC [TOPIC,TOPIC ...]]
               [--norm TOPIC [TOPIC ...]] [--test-accuracy] [--train-accuracy]
               [--svcca] [--depth N] [--hessian] [--exponential-decay GAMMA]
               [--log-dir LOG_DIR] [--seed SEED]

Named Arguments

-m, --model	Model class to train
-d, --dataset	Possible choices: MNIST, FashionMNIST, CIFAR10, CIFAR100 Dataset to train on
-b, --batch-size
	Default: 128
-e, --epochs	Default: 10
-o, --optimizer
	Optimizer to use Default: “Adam”
-l, --learning-rate
	Learning rate Default: 0.01
-a, --ratio-average
	Number of ratios to average for stability (currently unused) Default: 10
-s, --subsample
	Number of batches to ignore between updates Default: 1
-v, --visualisation
	Possible choices: tb, mpl Visualisation backend to use. Default: “tb”
-V, --verbose	Show training progress bar Default: False
--spectral-norm
	Use spectral norm subscriber(s)
--variance	Use variance norm subscriber(s)
--histogram	Use histogram subscriber(s)
--ratio	Use ratio subscriber(s)
--norm	Use 2-norm subscriber(s)
--test-accuracy
	Use test set accuracy subscriber Default: False
--train-accuracy
	Use train accuracy subscriber Default: False
--svcca	Use SVCCA subscriber Default: False
--depth	Depth to which to add modules Default: -1
--hessian	Use Hessian tracker (substantially increases training time) Default: False
--exponential-decay
	Decay parameter for exponential decay
--log-dir	TensorBoard logdir Default: “runs”
--seed	Seed to use. None means don’t seed

main._main(dataset_str, model_str, batch_size, epochs, optimizer, **kwargs)

Run the training procedure.

Parameters:

• dataset_str (str) – Name of the dataset to use
• model_str (str) – Unqualified name of the model class to use
• batch_size (int) –
• epochs (int) –
• optimizer (str) – Name of the optimizer to use

main.get_parser()

Obtain a configured argument parser. This function is necessary for the sphinx argparse extension.

Returns: Return type:argparse.ArgumentParser

main.main()