def train(M, FLAGS, model_name=None):
print("train called")
#nn = importlib.import_module(".{}".format(FLAGS.nn), package='networks')
data = get_data(FLAGS.src, FLAGS)
print(f"The number of train data is {data.train_num}")
print(f"The number of test data is {data.test_num}")
print(f"The number of validation data is {data.val_num}")
images, labels = data.train.next_batch(128)
print(f"The shape of the batch images is {images.shape}")
print(f"The shape of the batch labels is {labels.shape}")
print(np.max(images))
print(np.min(images))
# This code shows data shape and plots it.
for i in range(1):
first_data = images[i]
print(first_data.shape)
#trans_data = np.transpose(first_data, (1, 2, 0)).astype(np.float32)
# trans_data = np.tile(np.transpose(first_data, (1, 2, 0)).astype(np.float32), (1, 1, 3))
#print(trans_data.shape)
#first_data = np.tile(first_data, (1, 1, 3))
plt.imshow(first_data)
plt.show()
print(f"The label of the image is {np.argmax(labels[i])}")
config.label_type = 'vec'
config.label_order = 'freq_first'
# cuda
use_cuda = torch.cuda.is_available() and len(opt.gpus) > 0
if use_cuda:
torch.cuda.set_device(opt.gpus[0])
torch.cuda.manual_seed(opt.seed)
print(use_cuda)
# data
print('loading data...\n')
start_time = time.time()
item_train, label_train, \
item_test, label_test = dataset.get_data(config)
'''
trainset = dataset.Dataset(item_train, label_train, config.label_set_size, input_type = config.module_type)
validset = dataset.Dataset(item_test, label_test, config.label_set_size, input_type = config.module_type)
testset = dataset.Dataset(item_test, label_test, config.label_set_size, input_type = config.module_type)
'''
trainloader = dataloader.wgan_data_loader(item_train,
label_train,
config.batch_size,
config,
shuffle=True,
balance=True)
validloader = dataloader.wgan_data_loader(item_test,
label_test,
config.batch_size,
config,
# [0, 1, 0, 1, 1]
"""
result = tf.zeros(depth)
for scalar in input:
result += tf.one_hot(scalar, depth)
return result
""" Main stub """
print(colors.BOLD, 'Loading dataset...', colors.ENDC)
data_set, labels = data.get_data()
print(colors.BOLD, 'Transforming data...', colors.ENDC, end='')
data_set = list(map(partial(transform_input_one_hot, depth=__dim__), data_set))
data_set = tf.stack(data_set)
data_set = tf.map_fn(lambda t: t / tf.reduce_max(t), data_set)
train = data_set[:800]
validation = data_set[800:]
tf.InteractiveSession()
train_np = train.eval()
validation_np = validation.eval()