type=int,
help='Batch size')
parser.add_argument('-restore',
'--restore',
default='model_last.pth',
type=str) # model_last.pth
parser.add_argument('-output_path', '--output_path', default='ckpts', type=str)
parser.add_argument('-prefix_path', '--prefix_path', default='', type=str)
path = os.path.dirname(__file__)
args = parser.parse_args()
args = Parser(args.cfg, log='train').add_args(args)
ckpts = args.makedir()
args.resume = os.path.join(ckpts, args.restore) # specify the epoch
def main():
# os.environ['CUDA_VISIBLE_DEVICES'] = args.gpu
assert torch.cuda.is_available(), "Currently, we only support CUDA version"
torch.manual_seed(args.seed)
# torch.cuda.manual_seed(args.seed)
random.seed(args.seed)
np.random.seed(args.seed)
device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
Network = getattr(models, args.net) #
model = Network(**args.net_params)
model = torch.nn.DataParallel(model).to(device)
parser.add_argument('-snapshot', '--snapshot', default=False, type=str2bool,
help='If True, saving the snopshot figure of all samples.')
parser.add_argument('-restore_prefix', '--restore_prefix', default=argparse.SUPPRESS, type=str,
help='The path to restore the model.') # 'model_epoch_300.pth'
parser.add_argument('-restore_epoch', '--restore_epoch', default='399,499,599,699,799,899,999', type=str)
parser.add_argument('-out_dir', '--out_dir', default='/opt/ml/model/', type=str)
parser.add_argument('-prefix_path', '--prefix_path', default='', type=str)
path = os.path.dirname(__file__)
args = parser.parse_args()
args = Parser(args.cfg, log='train').add_args(args)
# args.gpu = str(args.gpu)
# ckpts = args.makedir()
args.resume = [args.restore_prefix + epoch + '.pth' for epoch in args.restore_epoch.split(',')]
# sample:
# CUDA_VISIBLE_DEVICES=1 python test_all.py --mode=1 --is_out=True --verbose=True --use_TTA=True --postprocess=True --snapshot=False --restore_prefix=./ckpts/DMFNet_pe_all/model_epoch_ --cfg=./ckpts/DMFNet_pe_all/cfg.yaml
def main():
# setup environments and seeds
# os.environ['CUDA_VISIBLE_DEVICES'] = args.gpu
assert torch.cuda.is_available(), "Currently, we only support CUDA version"
torch.manual_seed(args.seed)
torch.cuda.manual_seed(args.seed)
random.seed(args.seed)
np.random.seed(args.seed)
Network = getattr(models, args.net) #
#parser.add_argument('-cfg', '--cfg', default='deepmedic_ce_50_50_c25_redo', type=str)
#parser.add_argument('-cfg', '--cfg', default='deepmedic_ce_50_50_all', type=str)
parser.add_argument('-gpu', '--gpu', default='0', type=str)
parser.add_argument('-out', '--out', default='', type=str)
path = os.path.dirname(__file__)
## parse arguments
args = parser.parse_args()
args = Parser(args.cfg, log='train').add_args(args)
args.gpu = str(args.gpu)
ckpts = args.makedir()
resume = os.path.join(ckpts, 'model_last.tar')
if not args.resume and os.path.exists(resume):
args.resume = resume
def main():
# setup environments and seeds
os.environ['CUDA_VISIBLE_DEVICES'] = args.gpu
torch.manual_seed(args.seed)
torch.cuda.manual_seed(args.seed)
random.seed(args.seed)
np.random.seed(args.seed)
# setup networks
Network = getattr(models, args.net)
model = Network(**args.net_params)
model = model.cuda()
help='Batch size')
parser.add_argument('-restore', '--restore', default='', type=str) # model_last.pth
parser.add_argument('-output_path', '--output_path', default='ckpts', type=str)
parser.add_argument('-prefix_path', '--prefix_path', default='', type=str)
parser.add_argument('-aws', '--aws', default=False, type=bool)
parser.add_argument('-finetune', '--finetune', default=False, type=bool)
path = os.path.dirname(__file__)
## parse arguments
args = parser.parse_args()
args = Parser(args.cfg, log='train').add_args(args)
# args.net_params.device_ids= [int(x) for x in (args.gpu).split(',')]
ckpts = args.makedir()
args.resume = args.restore # specify the epoch
if not args.restore:
# load from /opt/ml/checkpoints
local_path = '/opt/ml/checkpoints/'
list_checkpoints = os.listdir(local_path)
# get last checkpoints?
if list_checkpoints:
for f in list_checkpoints:
try:
shutil.copy2(os.path.join(local_path, f), ckpts)
except:
continue
list_checkpoints.sort(key=len)
list_checkpoints = [checkpoint for checkpoint in list_checkpoints if len(checkpoint) == len(list_checkpoints[-1])]
list_checkpoints = sorted(list_checkpoints)
last_checkpoints = list_checkpoints[-1]
