default='spiral',
help='Dataset: select between "spiral" and "flower".')
parser.add_argument('-d',
'--debug',
action='store_true',
help='Used for gradient checking.')
FLAGS, unparsed = parser.parse_known_args()
##======================================================================
## STEP 1: Loading data
#
# In this section, we load the training examples and their labels.
if FLAGS.input_data == 'spiral':
instances, labels, numClasses = utils.load_spiral_dataset()
elif FLAGS.input_data == 'flower':
instances, labels, numClasses = utils.load_flower_dataset()
else:
print('Wrong dataset specified. Select between "spiral" and "flower".')
sys.exit(1)
inputSize = instances.shape[0]
numExamples = instances.shape[1]
# For debugging purposes, you may wish to reduce the size of the input data
# in order to speed up gradient checking.
# Here, we create synthetic dataset using random data for testing
if FLAGS.debug:
inputSize = 8
parser.add_argument('-d',
'--debug',
action='store_true',
help='Used for gradient checking.')
FLAGS, unparsed = parser.parse_known_args()
torch.manual_seed(3)
##======================================================================
## STEP 1: Load data
#
# In this section, we load the training instances and their labels.
if FLAGS.input_data == 'spiral':
X, y, n_y = utils.load_spiral_dataset()
# Set hyper-parameters.
n_h = 100
decay = 0.001
learning_rate = 1
num_epochs = 10000
elif FLAGS.input_data == 'flower':
X, y, n_y = utils.load_flower_dataset()
# Set hyper-parameters.
n_h = 20
decay = 0
learning_rate = 0.05
num_epochs = 20000
else:
print('Wrong dataset specified. Select between "spiral" and "flower".')
sys.exit(1)