device = setup.init_seed()
train_data = gmm_data.get_data()
###################################### visualize ###############################################################
# create loader with data to iterate, batch size is 100
data_loader = torch.utils.data.DataLoader(train_data,
batch_size=100,
shuffle=True)
# number of batches, 100000 / 100 = 1000
num_batches = len(data_loader)
# start writing networks
dis = model.DNet().float().to(device) # generate a funtion
gen = model.GNet().float().to(device) # generate a function
# hidden: torch.randn(size, 100)
################################################# loss function #############################################
criterion = nn.BCELoss().to(
device
) # binary cross entropy loss, nothing fancy but a dirty way to compute the log
lr = 0.01
################################################# wrap the training as functions ################################
# the problem is a min-max optimization: min_G max_D log(D(x)) + log(1 - D(G(z)))
# minimize -log(D(x)) - log (1 - D(G(z))), take one gradient descent step
TRAIN_FOLDER = '/home/alfarihfz/data/Pulmonary/train/'
BATCH_SIZE=64
EPOCH = 500
DEVICE = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
def image_loader(path):
return Image.open(path)
transform = transforms.Compose(
[
transforms.ToTensor(),
]
)
gnet = model.GNet(device=DEVICE)
hog = model.HOGLayer(device=DEVICE)
trainData = datasets.DatasetFolder(root=TRAIN_FOLDER,
loader=image_loader,
extensions=['jpeg', 'png'],
transform=transform)
dataLoader = DataLoader(trainData, batch_size=BATCH_SIZE, shuffle=True)
optimizer = torch.optim.Adam(gnet.parameters(), weight_decay=5e-4)
criterion = torch.nn.MSELoss()
for epoch in range(EPOCH):
batch_loss = 0.0
for i, data in enumerate(dataLoader):
img,_ = data
nargs='?',
default='',
help='Path to image folder. This is where the images from the run will be saved.'
)
args = parser.parse_args()
# check that model Keras version is same as local Keras version
f = h5py.File(args.pred_model, mode='r')
model_version = f.attrs.get('keras_version')
keras_version = str(keras_version).encode('utf8')
if model_version != keras_version:
print('You are using Keras version ', keras_version,
', but the model was built using ', model_version)
pred_model = model.GNet()._model
pred_model.load_weights(args.pred_model)
if args.image_folder != '':
print("Creating image folder at {}".format(args.image_folder))
if not os.path.exists(args.image_folder):
os.makedirs(args.image_folder)
else:
shutil.rmtree(args.image_folder)
os.makedirs(args.image_folder)
print("RECORDING THIS RUN ...")
else:
print("NOT RECORDING THIS RUN ...")
# wrap Flask application with engineio's middleware
app = socketio.Middleware(sio, app)