炼丹时经常需要可视化中间层的结果进行分析，本文采用tensorflow2.1版本，进行代码展示
def show_middle(test_dataset_list, model, checkpoint_dir):tf.keras.backend.clear_session()#读取测试集test_dataset = load_dataset_test()#读取保存的模型参数latest = tf.train.latest_checkpoint(checkpoint_dir)#导入保存的模型结构model = Model(model_name)#加载参数model.load_weights(latest)outputs = []##################visualization_model = models.Model(inputs=model.input,outputs=model.get_layer('tf_op_layer_tmp').output)############# for sample, mask in test_dataset:output = visualization_model.predict(sample)outputs.append(output)outputs = np.mean(outputs, axis = 0)print(outputs)