elif args.channels == 1:
normalize = transforms.Normalize(mean=[0.5], std=[0.5])
# create model
avg_pool_size = (args.avg_pooling_height, args.avg_pooling_width)
model = densenet_multi(num_init_features=64,
growth_rate=32,
block_config=(6, 12, 24, 16),
num_classes=args.num_classes,
channels=args.channels,
avg_pooling_size=avg_pool_size)
train_transforms = transforms.Compose([transforms.ToTensor(), normalize])
val_transforms = transforms.Compose([transforms.ToTensor(), normalize])
# create optimizer
optimizer = optim.Adam(model.parameters(), lr=args.lr, betas=(0.5, 0.999))
# start main loop
main(args,
model,
pil_loader,
pil_loader,
normalize,
train_dataset=MultiTaskDataset,
train_model=train_model,
validate_model=validate_model,
train_transforms=train_transforms,
val_transforms=val_transforms,
optimizer=optimizer)
# create model
avg_pool_size = (args.avg_pooling_height, args.avg_pooling_width)
if args.pretrained:
if args.channels != 3:
print('(pretrained) Must fix channels == 3')
exit()
if args.avg_pooling_width != 7 and args.avg_pooling_height != 7:
print('(pretrained) Must fix avg_pooling_size == 7')
exit()
# fix num_init_features=64, growth_rate=32, block_config=(6, 12, 24, 16)
model = densenet121_pretrained(num_classes=args.num_classes)
else:
model = DenseNet(num_init_features=32, growth_rate=16, block_config=(6, 12, 24, 16),
num_classes=args.num_classes,
channels=args.channels, avg_pooling_size=avg_pool_size)
train_transforms = transforms.Compose([transforms.RandomCrop((args.image_height, args.image_width)),
transforms.ToTensor(),
normalize,
])
# create optimizer
optimizer = torch.optim.Adam(model.parameters(), lr=args.lr)
# start main loop
main(args, model, train_image_loader, valid_image_loader, normalize, optimizer,
train_dataset=TrainDataset, valid_dataset=ValDataset,
train_model=train_model, validate_model=validate_model,
train_transforms=train_transforms)
args.image_height = 224
args.num_classes = 1000
args.channels = 3
def pil_loader(path):
with open(path, 'rb') as f:
with Image.open(f) as img:
img = img.convert('RGB') if args.channels == 3 else img.convert(
'L')
img = img.resize(
(args.resize_image_width, args.resize_image_height))
return img
if __name__ == '__main__':
normalize = transforms.Normalize(mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225])
# create model
model = models.__dict__[args.arch](num_classes=args.num_classes,
pretrained=True)
# create optimizer
optimizer = torch.optim.Adam(model.parameters(),
lr=args.lr,
weight_decay=args.weight_decay)
# start main loop
main(args, model, pil_loader, pil_loader, normalize, optimizer)
cat_dims, emb_dims = get_dims(find_csv(args.data, 'train'))
# create model
model = FeedForwardNN(emb_dims,
no_of_cont=5,
lin_layer_sizes=[50, 100],
output_size=2,
emb_dropout=0.04,
lin_layer_dropouts=[0.2, 0.2])
# lin_layer_dropouts=[0.001, 0.01])
# create optimizer
optimizer = torch.optim.SGD(model.parameters(), lr=args.lr)
transforms = transforms.Compose(transforms.ToTensor())
dumy = lambda x: x
# start main loop
main(args,
model,
dummyImageLoader,
dummyImageLoader,
None,
optimizer,
train_dataset=DummyDataset,
valid_dataset=DummyDataset,
train_model=train_model,
validate_model=validate_model,
train_transforms=dumy,
val_transforms=dumy)
if __name__ == '__main__':
if args.channels == 3:
normalize = transforms.Normalize(mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225])
elif args.channels == 1:
normalize = transforms.Normalize(mean=[0.5], std=[0.5])
# create model
avg_pool_size = (args.avg_pooling_height, args.avg_pooling_width)
if args.simple_resnet:
model = resnet18(num_classes=args.num_classes,
channels=args.channels,
avg_pooling_size=avg_pool_size)
else:
model = densenet121(num_classes=args.num_classes,
channels=args.channels,
avg_pooling_size=avg_pool_size)
train_transforms = transforms.Compose([transforms.ToTensor(), normalize])
val_transforms = transforms.Compose([transforms.ToTensor(), normalize])
# start main loop
main(args,
model,
pil_loader,
pil_loader,
normalize,
validate_model=validate_model,
train_transforms=train_transforms,
val_transforms=val_transforms)