tensorflow2.1 输出中间层结果

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)