print('apply training algorithm...')
# Model initialization
feature_net = utils_resnet.Resnet(pretrained=True)
icarl = utils_icarl.iCaRL(param, feature_net, label_dict)
# Training tools
loss_fn = torch.nn.BCELoss(size_average=False)
optimizer = torch.optim.Adam(icarl.parameters(),
lr=lr,
weight_decay=wght_decay)
scheduler = MultiStepLR(optimizer, milestones=lr_milestones, gamma=0.1)
if (gpu):
icarl = icalr.cuda()
loss_fn = loss_fn.cuda()
scheduler = scheduler.cuda()
# Recording traing process
log = open(work_path + '/log.txt', 'ab', 0)
log.write('epoch time training_loss validation_loss \n'.encode())
for iter_group in range(2): #nb_group
# Loading protoset
if (iter_group == 0):
protoset = dict()
else:
protoset_name = work_path + '/protoset{}'.format(iter_group - 1)
with open(protoset_name, 'rb') as f:
protoset = pickle.load(f)
f.close()
