def main():
"""Create the model and start the evaluation process."""
args = get_arguments()
update_config(config, args)
print(args)
os.environ["CUDA_VISIBLE_DEVICES"] = args.gpu
gpus = [int(i) for i in args.gpu.split(',')]
h, w = map(int, args.input_size.split(','))
input_size = (h, w)
model = get_cls_net(config=config,
num_classes=args.num_classes,
is_train=False)
normalize = transforms.Normalize(mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225])
transform = transforms.Compose([
transforms.ToTensor(),
normalize,
])
print('-------Load Data', args.data_dir)
if 'vehicle_parsing_dataset' in args.data_dir:
parsing_dataset = VPDataSet(args.data_dir,
args.dataset,
crop_size=input_size,
transform=transform,
list_path=args.list_path)
elif 'LIP' in args.data_dir:
parsing_dataset = LIPDataSet(args.data_dir,
args.dataset,
crop_size=input_size,
transform=transform)
num_samples = len(parsing_dataset)
valloader = data.DataLoader(parsing_dataset,
batch_size=args.batch_size * len(gpus),
shuffle=False,
pin_memory=True)
print('-------Load Weight', args.restore_from)
restore_from = args.restore_from
state_dict = model.state_dict().copy()
state_dict_old = torch.load(restore_from)
state_dict_old = state_dict_old['state_dict']
for key, nkey in zip(state_dict_old.keys(), state_dict.keys()):
if key != nkey:
# remove the 'module.' in the 'key'
state_dict[key[7:]] = deepcopy(state_dict_old[key])
else:
state_dict[key] = deepcopy(state_dict_old[key])
model.load_state_dict(state_dict)
model = DataParallelModel(model)
model.eval()
model.cuda()
print('-------Start Testing...')
parsing_preds, scales, centers, time_list = valid(model, valloader,
input_size, num_samples,
len(gpus))
print('-------Saving Results', args.save_dir)
write_results(parsing_preds,
scales,
centers,
args.data_dir,
args.dataset,
args.save_dir,
input_size=input_size)
print('total time is ', sum(time_list))
print('avg time is ', sum(time_list) / len(time_list))
def main():
"""Create the model and start the training."""
print(args)
if not os.path.exists(args.snapshot_dir):
os.makedirs(args.snapshot_dir)
writer = SummaryWriter(args.snapshot_dir)
gpus = [int(i) for i in args.gpu.split(',')]
if not args.gpu == 'None':
os.environ["CUDA_VISIBLE_DEVICES"] = args.gpu
h, w = map(int, args.input_size.split(','))
input_size = [h, w]
# cudnn related setting
cudnn.enabled = True
cudnn.benchmark = True
torch.backends.cudnn.deterministic = False
torch.backends.cudnn.enabled = True
deeplab = get_cls_net(config=config, is_train=True)
# model = DataParallelModel(deeplab)
print('-------Load Weight', args.restore_from)
saved_state_dict = torch.load(args.restore_from)
if args.start_epoch > 0:
model = DataParallelModel(deeplab)
model.load_state_dict(saved_state_dict['state_dict'])
else:
new_params = deeplab.state_dict().copy()
state_dict_pretrain = saved_state_dict
for state_name in state_dict_pretrain:
if state_name in new_params:
new_params[state_name] = state_dict_pretrain[state_name]
print('LOAD', state_name)
else:
print('NOT LOAD', state_name)
deeplab.load_state_dict(new_params)
model = DataParallelModel(deeplab)
print('-------Load Weight', args.restore_from)
model.cuda()
criterion = CriterionAll(loss_type=args.loss)
criterion = DataParallelCriterion(criterion)
criterion.cuda()
normalize = transforms.Normalize(mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225])
transform = transforms.Compose([
transforms.ToTensor(),
normalize,
])
print("-------Loading data...")
if 'vehicle_parsing_dataset' in args.data_dir:
parsing_dataset = VPDataSet(args.data_dir,
args.dataset,
crop_size=input_size,
transform=transform)
elif 'LIP' in args.data_dir:
parsing_dataset = LIPDataSet(args.data_dir,
args.dataset,
crop_size=input_size,
transform=transform)
print("Data dir : ", args.data_dir)
print("Dataset : ", args.dataset)
trainloader = data.DataLoader(parsing_dataset,
batch_size=args.batch_size * len(gpus),
shuffle=True,
num_workers=8,
pin_memory=True)
'''
list_map = []
for part in deeplab.path_list:
list_map = list_map + list(map(id, part.parameters()))
base_params = filter(lambda p: id(p) not in list_map,
deeplab.parameters())
params_list = []
params_list.append({'params': base_params, 'lr':args.learning_rate*0.1})
for part in deeplab.path_list:
params_list.append({'params': part.parameters()})
print ('len(params_list)',len(params_list))
'''
optimizer = optim.SGD(model.parameters(),
lr=args.learning_rate,
momentum=args.momentum,
weight_decay=args.weight_decay)
if args.start_epoch > 0:
optimizer.load_state_dict(saved_state_dict['optimizer'])
print('========Load Optimizer', args.restore_from)
total_iters = args.epochs * len(trainloader)
for epoch in range(args.start_epoch, args.epochs):
model.train()
for i_iter, batch in enumerate(trainloader):
i_iter += len(trainloader) * epoch
lr = adjust_learning_rate(optimizer, i_iter, total_iters)
images, labels, _ = batch
labels = labels.long().cuda(non_blocking=True)
preds = model(images)
loss = criterion(preds, labels)
optimizer.zero_grad()
loss.backward()
optimizer.step()
if i_iter % 100 == 0:
writer.add_scalar('learning_rate', lr, i_iter)
writer.add_scalar('loss', loss.data.cpu().numpy(), i_iter)
print(
f'epoch = {epoch}, iter = {i_iter}/{total_iters}, lr={lr:.6f}, loss = {loss.data.cpu().numpy():.6f}'
)
if (epoch + 1) % args.save_step == 0 or epoch == args.epochs:
time.sleep(10)
print("-------Saving checkpoint...")
save_checkpoint(model, epoch, optimizer)
time.sleep(10)
save_checkpoint(model, epoch, optimizer)
end = timeit.default_timer()
print(end - start, 'seconds')