if not k in new:
new[k] = torch.Tensor(1)
new[k][0] = -valid['.'.join(parts[:-1] +
['weight'])].min() / 2
return new
#dct = torch.load('resnet.pth')
#dct = fix_state(model.state_dict(), dct)
#model.load_state_dict(dct)
if use_cuda:
model.cuda()
criterion = F.cross_entropy
optimizer = optim.Adam(model.parameters(),
lr=args.lr,
weight_decay=args.weight_decay)
def train(epoch):
model.train()
running_loss = 0
running_total = 0
correct = 0
for i, (inputs, labels) in enumerate(trainloader):
if use_cuda:
inputs, labels = inputs.cuda(), labels.cuda()
# wrap them in Variable
inputs, labels = Variable(inputs), Variable(labels)
if __name__ == "__main__":
parser = argparse.ArgumentParser(description='pytorch SqueezeNet on CUDA')
parser.add_argument('--data_path', type=str, default="./data",
help='path where the dataset is saved')
parser.add_argument('--ckpt_path', type=str, default="./checkpoint",
help='path where the checkpoint to be saved')
parser.add_argument('--device_id', type=int, default=0, help='device id of GPU. (Default: 0)')
args = parser.parse_args()
device = torch.device('cuda:'+str(args.device_id))
version = '1.0'
network = SqueezeNet(cfg.num_classes, version)
network.to(device)
criterion = LabelSmoothingCrossEntropy(reduction="mean", epsilon=cfg.label_smoothing_eps)
optimizer = optim.RMSprop(network.parameters(),
lr=cfg.lr_init,
eps=cfg.rmsprop_epsilon,
momentum=cfg.rmsprop_momentum,
alpha=cfg.rmsprop_decay)
dataloader = create_dataset_pytorch(args.data_path, is_train=True)
step_per_epoch = len(dataloader)
scheduler = optim.lr_scheduler.StepLR(
optimizer,
gamma=cfg.lr_decay_rate,
step_size=cfg.lr_decay_epoch*step_per_epoch)
# scheduler = optim.lr_scheduler.ExponentialLR(
# optimizer,
# gamma=cfg.lr_decay_rate)
q_ckpt = Queue(maxsize=cfg.keep_checkpoint_max)