else:
print("Error: no model version " + model_name)
exit(0)
model = Network(cfg)
use_cuda = torch.cuda.is_available()
logger.info('cuda available: {}'.format(use_cuda))
assert use_cuda
model = model.cuda()
# logger.info(model)
logger.info('#classifier parameters (model): {}'.format(
sum([x.nelement() for x in model.parameters()])))
##### model_fn (criterion)
model_fn = model_fn_decorator(test=True)
##### load model
utils.checkpoint_restore(
model,
cfg.exp_path,
cfg.config.split('/')[-1][:-5],
use_cuda,
cfg.test_epoch,
dist=False,
f=cfg.pretrain
) # resume from the latest epoch, or specify the epoch to restore
##### evaluate
test(model, model_fn, data_name, cfg.test_epoch)
sum([x.nelement() for x in model.parameters()])))
##### optimizer
if cfg.optim == 'Adam':
optimizer = optim.Adam(filter(lambda p: p.requires_grad,
model.parameters()),
lr=cfg.lr)
elif cfg.optim == 'SGD':
optimizer = optim.SGD(filter(lambda p: p.requires_grad,
model.parameters()),
lr=cfg.lr,
momentum=cfg.momentum,
weight_decay=cfg.weight_decay)
##### model_fn (criterion)
model_fn = model_fn_decorator()
import data.scannetv2_inst
dataset = data.scannetv2_inst.Dataset()
dataset.trainLoader()
dataset.valLoader()
##### dataset
'''if cfg.dataset == 'scannetv2':
if data_name == 'scannet':
else:
print("Error: no data loader - " + data_name)
exit(0)'''
##### resume
start_epoch = utils.checkpoint_restore(