def train_net(args):
torch.manual_seed(7)
np.random.seed(7)
checkpoint = args.checkpoint
start_epoch = 0
best_acc = 0
writer = SummaryWriter()
epochs_since_improvement = 0
# Initialize / load checkpoint
if checkpoint is None:
if args.network == 'r18':
model = resnet18(args)
elif args.network == 'r34':
model = resnet34(args)
elif args.network == 'r50':
model = resnet50(args)
elif args.network == 'r101':
model = resnet101(args)
elif args.network == 'r152':
model = resnet152(args)
elif args.network == 'mobile':
model = MobileNet(1.0)
elif args.network == 'mr18':
print("mr18")
model = myResnet18()
else:
model = resnet_face18(args.use_se)
model = nn.DataParallel(model)
metric_fc = ArcMarginModel(args)
metric_fc = nn.DataParallel(metric_fc)
if args.optimizer == 'sgd':
optimizer = torch.optim.SGD([{
'params': model.parameters()
}, {
'params': metric_fc.parameters()
}],
lr=args.lr,
momentum=args.mom,
weight_decay=args.weight_decay)
else:
optimizer = torch.optim.Adam([{
'params': model.parameters()
}, {
'params': metric_fc.parameters()
}],
lr=args.lr,
weight_decay=args.weight_decay)
else:
checkpoint = torch.load(checkpoint)
start_epoch = checkpoint['epoch'] + 1
epochs_since_improvement = checkpoint['epochs_since_improvement']
model = checkpoint['model']
metric_fc = checkpoint['metric_fc']
optimizer = checkpoint['optimizer']
logger = get_logger()
# Move to GPU, if available
model = model.to(device)
metric_fc = metric_fc.to(device)
# Loss function
if args.focal_loss:
criterion = FocalLoss(gamma=args.gamma).to(device)
else:
criterion = nn.CrossEntropyLoss().to(device)
# Custom dataloaders
train_dataset = ArcFaceDataset('train')
train_loader = torch.utils.data.DataLoader(train_dataset,
batch_size=args.batch_size,
shuffle=True)
scheduler = StepLR(optimizer, step_size=args.lr_step, gamma=0.1)
# Epochs
for epoch in range(start_epoch, args.end_epoch):
scheduler.step()
if args.full_log:
lfw_acc, threshold = lfw_test(model)
writer.add_scalar('LFW_Accuracy', lfw_acc, epoch)
full_log(epoch)
start = datetime.now()
# One epoch's training
train_loss, train_top5_accs = train(train_loader=train_loader,
model=model,
metric_fc=metric_fc,
criterion=criterion,
optimizer=optimizer,
epoch=epoch,
logger=logger,
writer=writer)
writer.add_scalar('Train_Loss', train_loss, epoch)
writer.add_scalar('Train_Top5_Accuracy', train_top5_accs, epoch)
end = datetime.now()
delta = end - start
print('{} seconds'.format(delta.seconds))
# One epoch's validation
lfw_acc, threshold = lfw_test(model)
writer.add_scalar('LFW Accuracy', lfw_acc, epoch)
# Check if there was an improvement
is_best = lfw_acc > best_acc
best_acc = max(lfw_acc, best_acc)
if not is_best:
epochs_since_improvement += 1
print("\nEpochs since last improvement: %d\n" %
(epochs_since_improvement, ))
else:
epochs_since_improvement = 0
# Save checkpoint
save_checkpoint(epoch, epochs_since_improvement, model, metric_fc,
optimizer, best_acc, is_best)
def main():
parser = argparse.ArgumentParser()
parser.add_argument("--mode", type=str, default="train")
parser.add_argument("--model", type=str, default="mobilenet_v2")
parser.add_argument("--dataset", type=str, default="cifar10")
parser.add_argument("--dataroot", type=str, default="/tmp/data")
parser.add_argument("--batch_size", type=int, default=128)
parser.add_argument("--n_epochs", type=int, default=100)
parser.add_argument("--lr", type=float, default=1e-3)
parser.add_argument("--n_gpus", type=int, default=1)
parser.add_argument("--checkpoint", type=str, default="/tmp/chkpt.pth.tar")
parser.add_argument("--save_every", type=int, default=10)
parser.add_argument("--pretrained", type=str, default=None)
args = parser.parse_args()
print(args)
if torch.cuda.is_available():
print("cuda is available, use cuda")
device = torch.device("cuda")
else:
print("cuda is not available, use cpu")
device = torch.device("cpu")
print("download dataset: {}".format(args.dataset))
train_loader, test_loader, n_classes = get_loaders(
dataset=args.dataset, root=args.dataroot, batch_size=args.batch_size)
print("build model: {}".format(args.model))
if args.model == "mobilenet":
from models import MobileNet
model = MobileNet(n_classes=n_classes)
elif args.model == "mobilenet_v2":
from models import MobileNet_v2
model = MobileNet_v2(n_classes=n_classes)
elif args.model == "shufflenet":
from models import ShuffleNet
model = ShuffleNet(n_classes=n_classes)
elif args.model == "shufflenet_v2":
from models import ShuffleNet_v2
model = ShuffleNet_v2(n_classes=n_classes)
elif args.model == "squeezenet":
from models import SqueezeNet
model = SqueezeNet(n_classes=n_classes)
else:
raise NotImplementedError
model = model.to(device)
if args.pretrained:
model.load_state_dict(torch.load(args.checkpoint))
if args.n_gpus > 1:
gpus = []
for i in range(args.n_gpus):
gpus.append(i)
model = nn.DataParallel(model, device_ids=gpus)
optimizer = optim.Adam(model.parameters(), lr=args.lr)
criterion = nn.CrossEntropyLoss()
if args.mode == "train":
for epoch in range(args.n_epochs):
train(epoch, model, optimizer, criterion, train_loader, device)
if (epoch + 1) % args.save_every == 0:
print("saving model...")
torch.save(the_model.state_dict(), args.checkpoint)
elif args.mode == "test":
test(model, criterion, test_loader, device)
else:
raise NotImplementedError
elif backbone_network == 'WideResNet':
from models import WideResNet
model = WideResNet(n_layers=n_layers,
widening_factor=opt.widening_factor,
dataset=opt.dataset,
attention=opt.attention_module,
group_size=opt.group_size)
model = nn.DataParallel(model).to(device)
criterion = nn.CrossEntropyLoss()
if dataset_name in ['CIFAR10', 'CIFAR100']:
optim = torch.optim.SGD(model.parameters(),
lr=opt.lr,
momentum=opt.momentum,
weight_decay=opt.weight_decay)
milestones = [150, 225]
elif dataset_name == 'ImageNet':
optim = torch.optim.SGD(model.parameters(),
lr=opt.lr,
momentum=opt.momentum,
weight_decay=opt.weight_decay)
milestones = [30, 60]
elif dataset_name == 'SVHN':
optim = torch.optim.SGD(model.parameters(),
def train_net(args):
torch.manual_seed(7)
np.random.seed(7)
checkpoint = args.checkpoint
start_epoch = 0
best_acc = 0
writer = SummaryWriter()
epochs_since_improvement = 0
# Initialize / load checkpoint
if checkpoint is None:
if args.network == 'r18':
model = resnet18(args)
elif args.network == 'r34':
model = resnet34(args)
elif args.network == 'r50':
model = resnet50(args)
elif args.network == 'r101':
model = resnet101(args)
elif args.network == 'r152':
model = resnet152(args)
elif args.network == 'mobile':
model = MobileNet(1.0)
else:
model = resnet_face18(args.use_se)
model = nn.DataParallel(model)
metric_fc = ArcMarginModel(args)
metric_fc = nn.DataParallel(metric_fc)
if args.optimizer == 'sgd':
# optimizer = torch.optim.SGD([{'params': model.parameters()}, {'params': metric_fc.parameters()}],
# lr=args.lr, momentum=args.mom, weight_decay=args.weight_decay)
optimizer = InsightFaceOptimizer(
torch.optim.SGD([{
'params': model.parameters()
}, {
'params': metric_fc.parameters()
}],
lr=args.lr,
momentum=args.mom,
weight_decay=args.weight_decay))
else:
optimizer = torch.optim.Adam([{
'params': model.parameters()
}, {
'params': metric_fc.parameters()
}],
lr=args.lr,
weight_decay=args.weight_decay)
else:
checkpoint = torch.load(checkpoint)
start_epoch = checkpoint['epoch'] + 1
epochs_since_improvement = checkpoint['epochs_since_improvement']
model = checkpoint['model']
metric_fc = checkpoint['metric_fc']
optimizer = checkpoint['optimizer']
logger = get_logger()
# Move to GPU, if available
model = model.to(device)
metric_fc = metric_fc.to(device)
# Loss function
if args.focal_loss:
criterion = FocalLoss(gamma=args.gamma).to(device)
else:
criterion = nn.CrossEntropyLoss().to(device)
# Custom dataloaders
train_dataset = ArcFaceDataset('train')
train_loader = torch.utils.data.DataLoader(train_dataset,
batch_size=args.batch_size,
shuffle=True,
num_workers=8)
# Epochs
for epoch in range(start_epoch, args.end_epoch):
# One epoch's training
train_loss, train_top1_accs = train(train_loader=train_loader,
model=model,
metric_fc=metric_fc,
criterion=criterion,
optimizer=optimizer,
epoch=epoch,
logger=logger)
print('\nCurrent effective learning rate: {}\n'.format(optimizer.lr))
print('Step num: {}\n'.format(optimizer.step_num))
writer.add_scalar('model/train_loss', train_loss, epoch)
writer.add_scalar('model/train_accuracy', train_top1_accs, epoch)
writer.add_scalar('model/learning_rate', optimizer.lr, epoch)
# One epoch's validation
megaface_acc = megaface_test(model)
writer.add_scalar('model/megaface_accuracy', megaface_acc, epoch)
# Check if there was an improvement
is_best = megaface_acc > best_acc
best_acc = max(megaface_acc, best_acc)
if not is_best:
epochs_since_improvement += 1
print("\nEpochs since last improvement: %d\n" %
(epochs_since_improvement, ))
else:
epochs_since_improvement = 0
# Save checkpoint
save_checkpoint(epoch, epochs_since_improvement, model, metric_fc,
optimizer, best_acc, is_best)
def train_net(args):
torch.manual_seed(7) #torch的随机种子，在torch.randn使用
np.random.seed(7)
checkpoint = args.checkpoint
start_epoch = 0
best_acc = 0
writer = SummaryWriter() #tensorboard
epochs_since_improvement = 0
# Initialize / load checkpoint
if checkpoint is None:
if args.network == 'r18':
model = resnet18(args)
elif args.network == 'r34':
model = resnet34(args)
elif args.network == 'r50':
model = resnet50(args)
elif args.network == 'r101':
model = resnet101(args)
elif args.network == 'r152':
model = resnet152(args)
elif args.network == 'mobile':
model = MobileNet(1.0)
else:
model = resnet_face18(args.use_se)
model = nn.DataParallel(model)
metric_fc = ArcMarginModel(args)
metric_fc = nn.DataParallel(metric_fc)
if args.optimizer == 'sgd':
# optimizer = torch.optim.SGD([{'params': model.parameters()}, {'params': metric_fc.parameters()}],
# lr=args.lr, momentum=args.mom, weight_decay=args.weight_decay)
optimizer = InsightFaceOptimizer(
torch.optim.SGD([{
'params': model.parameters()
}, {
'params': metric_fc.parameters()
}],
lr=args.lr,
momentum=args.mom,
weight_decay=args.weight_decay))
else:
optimizer = torch.optim.Adam([{
'params': model.parameters()
}, {
'params': metric_fc.parameters()
}],
lr=args.lr,
weight_decay=args.weight_decay)
else:
checkpoint = torch.load(checkpoint)
#这里还需要自己加载进去
start_epoch = checkpoint['epoch'] + 1
epochs_since_improvement = checkpoint['epochs_since_improvement']
model = checkpoint['model']
metric_fc = checkpoint['metric_fc']
optimizer = checkpoint['optimizer']
logger = get_logger()
# Move to GPU, if available
model = model.to(device)
metric_fc = metric_fc.to(device)
# Loss function
if args.focal_loss:
criterion = FocalLoss(gamma=args.gamma).to(device)
else:
criterion = nn.CrossEntropyLoss().to(device)
# Custom dataloaders
train_dataset = Dataset(root=args.train_path,
phase='train',
input_shape=(3, 112, 112))
train_loader = torch.utils.data.DataLoader(train_dataset,
batch_size=args.batch_size,
shuffle=True,
num_workers=8)
# Epochs
for epoch in range(start_epoch, args.end_epoch):
# One epoch's training
# 这里写一个训练函数十分简练，值得学习
train_loss, train_top1_accs = train(train_loader=train_loader,
model=model,
metric_fc=metric_fc,
criterion=criterion,
optimizer=optimizer,
epoch=epoch,
logger=logger)
print('\nCurrent effective learning rate: {}\n'.format(optimizer.lr))
print('Step num: {}\n'.format(optimizer.step_num))
writer.add_scalar('model/train_loss', train_loss, epoch)
writer.add_scalar('model/train_accuracy', train_top1_accs, epoch)
writer.add_scalar('model/learning_rate', optimizer.lr, epoch)
# Save checkpoint
if epoch % 10 == 0:
save_checkpoint(epoch, epochs_since_improvement, model, metric_fc,
optimizer, best_acc)