print("Trainable params: {0:,}".format(trainable_params)) print("Non-trainable params: {0:,}".format(total_params - trainable_params)) print( "---------------------------------------------------------------------------------------------------------------------------" ) print("Input size (MB): %0.2f" % total_input_size) print("Forward/backward pass size (MB): %0.2f" % total_output_size) print("Params size (MB): %0.2f" % total_params_size) print("Estimated Total Size (MB): %0.2f" % total_size) print( "---------------------------------------------------------------------------------------------------------------------------" ) # return summary ############################################################################################################## if __name__ == '__main__': os.environ["CUDA_VISIBLE_DEVICES"] = "2" model = resnet50( ) #Total params: 23,272,266 #Total FLOPS: 62,751,882 #Accuracy: 85.61_98.95 #model = torch.load("model_training").cpu() model = torch.load("model_training_final").cpu() #model = model.module print("model:", model) #summary(model, (1, 28, 28), device="cpu") #summary(model, (3, 32, 32), device="cpu") summary(model, (3, 224, 224), device="cpu")
type=str, default= '/data1/Datasets/ImageNet/ILSVRC2012/ILSVRC2012_img_val_subfolders/', help='test dataset path') parser.add_argument("--parallel", type=int, default=1) parser.set_defaults(autoML=True) parser.set_defaults(train=False) args = parser.parse_args() return args ############################################################################################################## if __name__ == '__main__': os.environ["CUDA_VISIBLE_DEVICES"] = "0,1,2,3" args = get_args() print("args:", args) model = resnet50(pretrained=False).cuda() torch.save(model, "model") print("model_training:", model) if args.parallel == 1: model = torch.nn.DataParallel(model).cuda() fine_tuner = FineTuner_CNN(args.train_path, args.test_path, model) fine_tuner.test() if args.autoML: fine_tuner.autoML()