def load_model(model_file):
model_file = model_file
if args.student_model == '3CNN':
model = net_distill.Net_3CNN(int(args.num_classes), int(args.num_channels))
elif args.student_model == '5CNN':
model = net_distill.Net_5CNN(int(args.num_classes), int(args.num_channels))
elif args.student_model == '7CNN':
model = net_distill.Net_7CNN(int(args.num_classes), int(args.num_channels))
checkpoint = torch.load(model_file, map_location=lambda storage, loc: storage)
state_dict = {str.replace(k,'module.',''): v for k,v in checkpoint['state_dict'].items()}
model.load_state_dict(state_dict)
model.eval()
return model
def load_model(path, student_model, num_classes_extract_step, num_channels):
model_file = path
if student_model == '3CNN'
model = net_distill.Net_3CNN(num_classes_extract_step, num_channels)
elif student_model == '5CNN'
model = net_distill.Net_5CNN(num_classes_extract_step, num_channels)
elif student_model == '7CNN'
model = net_distill.Net_7CNN(num_classes_extract_step, num_channels)
checkpoint = torch.load(model_file, map_location=lambda storage, loc: storage)
state_dict = {str.replace(k,'module.',''): v for k,v in checkpoint['state_dict'].items()}
#model.load_state_dict(checkpoint['state_dict'])
model.load_state_dict(state_dict)
for i, (name, module) in enumerate(model._modules.items()):
module = recursion_change_bn(model)
model.eval()
return model