num_workers=0, collate_fn=trainset_valid.collate_fn) net = RetinaNet() net = torch.nn.DataParallel(net, device_ids=[0]) net.cuda() id_net = Idnet() id_net = torch.nn.DataParallel(id_net, device_ids=[0]) id_net.cuda() #MCP = arcface_loss2.Arcface(1024, 3000).cuda() criterion = torch.nn.CrossEntropyLoss().cuda() optimizer = optim.SGD( [{ 'params': id_net.parameters() }], #, {'params':MCP.parameters()}], lr=1e-3, momentum=0.9, weight_decay=1e-4) net.load_state_dict(torch.load("./trained model/originalFAN_model.pth")) net.eval() coder = DataEncoder() def save_model(model, filename): state = model.state_dict() for key in state: state[key] = state[key].clone().cpu() torch.save(state, filename)
collate_fn=testset.collate_fn) #net = torch.nn.DataParallel(net, device_ids=range(torch.cuda.device_count())) #net = torch.nn.DataParallel(net, device_ids=[0]) net = RetinaNet() net = torch.nn.DataParallel(net, device_ids=[0]) net.cuda() id_net = Idnet() id_net = torch.nn.DataParallel(id_net, device_ids=[0]) id_net.cuda() MCP = MarginCosineProduct(1024, 3000).cuda() criterion = torch.nn.CrossEntropyLoss().cuda() optimizer = optim.SGD([{'params': id_net.parameters()}, {'params':MCP.parameters()}], lr=1e-3, momentum=0.9, weight_decay=1e-4) net.load_state_dict(torch.load("./trained model/originalFAN_model.pth")) net.eval() coder = DataEncoder() def save_model(model, filename): state = model.state_dict() for key in state: state[key] = state[key].clone().cpu() torch.save(state, filename) def train(epoch, file_obj, acc):
num_workers=1, collate_fn=testset.collate_fn) #net = torch.nn.DataParallel(net, device_ids=range(torch.cuda.device_count())) #net = torch.nn.DataParallel(net, device_ids=[0]) net = RetinaNet() net = torch.nn.DataParallel(net, device_ids=[0]) net.cuda() id_net = Idnet() id_net = torch.nn.DataParallel(id_net, device_ids=[0]) id_net.cuda() criterion = torch.nn.CrossEntropyLoss().cuda() optimizer = optim.SGD(id_net.parameters(), lr=1e-3, momentum=0.9, weight_decay=1e-4) net.load_state_dict(torch.load("./trained model/originalFAN_model.pth")) net.eval() coder = DataEncoder() def save_model(model, filename): state = model.state_dict() for key in state: state[key] = state[key].clone().cpu() torch.save(state, filename)