from timeit import default_timer as timer
from collections import defaultdict
from BBalpha_dropout import *
if len(sys.argv) != 3:
print("Call this program like this:\n"
" ./mnist-mlp-train.py alpha run\n"
" e.g. ./mnist-mlp-train.py 0.5 1")
exit()
# extract command line arguments
alpha = float(sys.argv[1])
run = sys.argv[2]
# get dataset
train, validation, _ = load_mnist(flatten=True)
# constants
nb_train = train[0].shape[0]
nb_val = validation[0].shape[0]
input_dim = train[0].shape[1]
nb_classes = train[1].shape[1]
batch_size = 128
nb_layers = 2
nb_units = 100
p = 0.5
wd = 1e-6
K_mc = 10
import sys
import os
if len(sys.argv) != 3:
print("Call this program like this:\n"
" ./mnist-cnn-train.py alpha run\n"
" e.g. ./mnist-cnn-train.py 0.5 1"
)
exit()
# extract command line arguments
alpha = float(sys.argv[1])
run = sys.argv[2]
# get dataset
train, validation, _ = load_mnist(flatten=False, channels_first=True)
# constants
assert train[0].shape[2] == train[0].shape[3], 'Input image not square'
input_size = train[0].shape[2]
in_channels = train[0].shape[1]
nb_train = train[0].shape[0]
nb_val = validation[0].shape[0]
input_dim = train[0].shape[1]
nb_classes = train[1].shape[1]
batch_size = 128
val_batch_size = nb_val
# nb_layers = 2
nb_units = 100
p = 0.5
if len(sys.argv) != 3:
print("Call this program like this:\n"
" ./cnn-train.py dataset run\n"
" e.g. ./cnn-train.py mnist 1\n"
"Dataset is either ['mnist', 'cifar10', 'svhn']"
)
exit()
# extract command line arguments
dataset = sys.argv[1]
run = sys.argv[2]
# get dataset
if dataset == 'mnist':
train, validation, _ = load_dataset.load_mnist(flatten=False,
channels_first=False)
elif dataset == 'cifar10':
train, validation, _ = load_dataset.load_cifar10(channels_first=False)
elif dataset == 'svhn':
train, validation, _ = load_dataset.load_svhn(channels_first=False)
else:
print("Unrecognized dataset, use 'mnist', 'cifar10', or 'svhn'")
print("Exiting...")
exit()
# otherwise TF grabs all available gpu memory
if not hasattr(K, "tf"):
raise RuntimeError("This code requires keras to be configured"
" to use the TensorFlow backend.")
config = tf.ConfigProto()