Complete List of Available Command Line Options (Old)

Charles Siegel edited this page Mar 28, 2017 · 1 revision

The following table lists command-line parameters, the default value, an additional example value, and their brief description:

 Parameter	Default	Alt. Example	Description

--conv_layers	None	32 32	numbers of features in convolutional layers with 5x5 window and stride of 2

--full_layers	30	200 100 50	numbers of features in fully connected layers

--train_batch	128	1024	sample batch size

--epochs	30	300	number of epochs to run

--time	-1	60.0	stop running after given number of seconds; when present, --epochs is ignored

--learning_rate	0.1	1.1	learning rate

--data	MNIST	CIFAR10	data set, one of MNIST, CIFAR10, CIFAR100, CSV or PNETCDF

--inter_threads	0	5	sets inter_threads used by TensorFlow, when 0 is set dynamically by TensorFlow

--intra_threads	0	5	sets intra_threads used by TensorFlow, when 0 is set dynamically by TensorFlow

--threads	0	5	sets both inter_threads and intra_threads, when 0 is set dynamically by TensorFlow

--filename	None	file.csv	filename to use with --data CSV and --data PNETCDF

--filename	None	file.nc	filename to use with --data PNETCDF

--valid_pct	0.1	0.05	fraction of samples to reserve for validation set (used by all data sets)

--test_pct	0.1	0.05	fraction of samples to reserve for testing set (used only by --data CSV)

--error_batch	False	True	if set, breaks error and testing into batches. Does not take an argument.

--top	1	5	sets how many of the top rated outputs will be checked against the label for correctness