else:
#if we are using an asym bilstm
options.word_level_dim = sum(options.hidden_dim)
# ===-----------------------------------------------------------------------===
# Set up logging
# ===-----------------------------------------------------------------------===
if not os.path.exists(options.log_dir):
os.mkdir(options.log_dir)
if options.log_to_stdout:
logging.basicConfig(level=logging.INFO)
else:
logging.basicConfig(filename=options.log_dir + "/log.txt", filemode="w", format="%(message)s", level=logging.INFO)
train_dev_cost = utils.CSVLogger(options.log_dir + "/train_dev.log", ["Train.cost", "Dev.cost"])
# ===-----------------------------------------------------------------------===
# Log run parameters
# ===-----------------------------------------------------------------------===
logging.info(options)
logging.info(
"""
Dataset: {dataset}
Num Epochs: {epochs}
LSTM: {layers} layers, {hidden} hidden dim, {word} word level dim
Training set size limit: {sent} sentences or {tokes} tokens
Initial Learning Rate: {lr}
Dropout: {dropout}
def main():
global args, start_epoch
args = parser.parse_args()
test_id = args.net_type + str(args.depth) + '_' + args.expname + '_CAM'
print('test_id : ', test_id)
csv_path = os.path.join('logs', test_id)
checkpoint_path = os.path.join('save_models', test_id)
if not os.path.exists(csv_path):
os.makedirs(csv_path)
if not os.path.exists(checkpoint_path):
os.makedirs(checkpoint_path)
print('csv_path : ', csv_path)
print('models_path : ', checkpoint_path)
# Preprocessing
normalize = transforms.Normalize(mean=[0.485, 0.456, 0.406],
std=[0.229, 0.224, 0.225])
# Dataloader
train_transforms = transforms.Compose([
transforms.Resize((256, 256)),
transforms.RandomResizedCrop(224),
transforms.RandomHorizontalFlip(),
transforms.ToTensor(), normalize
])
val_transform = transforms.Compose([
transforms.Resize((256, 256)),
transforms.CenterCrop((224)),
transforms.ToTensor(), normalize
])
train_dataset = Cub2011(root='../datas/',
train=True,
transform=train_transforms)
train_loader = torch.utils.data.DataLoader(train_dataset,
batch_size=args.batch_size,
shuffle=True,
num_workers=args.workers)
val_dataset = Cub2011(root='../datas/',
train=False,
transform=val_transform)
val_loader = torch.utils.data.DataLoader(val_dataset,
batch_size=args.batch_size,
shuffle=False,
num_workers=args.workers)
numberofclass = 200
model = RN.ResNet(args.dataset, args.depth, numberofclass,
args.bottleneck) # for ResNet
model.cuda()
optimizer = torch.optim.SGD(model.parameters(),
args.lr,
momentum=args.momentum,
weight_decay=args.weight_decay,
nesterov=True)
criterion = nn.CrossEntropyLoss().cuda()
start_epoch = 0
filename = csv_path + '/' + test_id + '.csv'
csv_logger = utils.CSVLogger(csv_path,
args=args,
fieldnames=[
'epoch', 'train_loss', 'train_acc1',
'test_loss', 'test_acc1', 'test_acc5'
],
filename=filename)
if args.resume:
checkpoint_path = checkpoint_path + csv_path
checkpoint = torch.load(checkpoint_path + 'model_best.pth.tar')
model.load_state_dict(checkpoint['state_dict'])
optimizer.load_state_dict(checkpoint['optimizer'])
start_epoch = checkpoint['epoch'] + 1
filename = csv_path + 'log.csv'
csv_logger = utils.CSVLogger(csv_path,
args=args,
fieldnames=[
'epoch', 'train_loss', 'train_acc1',
'test_loss', 'test_acc1', 'test_acc5'
],
filename=filename)
for epoch in range(start_epoch, start_epoch + args.epochs):
utils.adjust_learning_rate(args, optimizer, epoch)
# train for one epoch
train_acc1, train_loss = train(train_loader, model, criterion,
optimizer, epoch)
# evaluate on validation set
acc1, acc5, val_loss = validate(val_loader, model, criterion, epoch)
# loss
train_loss = '%.4f' % (train_loss)
train_acc1 = '%.4f' % (train_acc1)
val_loss = '%.4f' % (val_loss)
test_acc1 = '%.4f' % (acc1)
test_acc5 = '%.4f' % (acc5)
# remember best prec@1 and save checkpoint
is_best = acc1 >= best_acc1
best_acc1 = max(acc1, best_acc1)
if is_best:
best_acc5 = acc5
print('Current best acc (top-1 {0:.3f} and 5 acc {1:.3f})'.format(
best_acc1, best_acc5))
print(' ')
utils.save_checkpoint(
{
'epoch': epoch,
'arch': args.net_type,
'state_dict': model.state_dict(),
'best_acc1': best_acc1,
'best_acc5': best_acc5,
'optimizer': optimizer.state_dict(),
}, is_best, test_id)
row = {
'epoch': str(epoch),
'train_loss': str(train_loss),
'train_acc1': str(train_acc1),
'test_loss': str(val_loss),
'test_acc1': str(test_acc1),
'test_acc5': str(test_acc5)
}
csv_logger.writerow(row)
print('Best accuracy (top-1 {0:.3f} and 5 acc {1:.3f})'.format(
best_acc1, best_acc5))
csv_logger.close()