我尝试通过mlflow预测为mnist训练的模型

loaded_model = mlflow.pyfunc.load_model(logged_model)

(x_train, y_train), (x_test, y_test) = mnist.load_data()

我尝试通过

x = pd.DataFrame(x_test)

但我得到了

ValueError: Must pass 2-d input. shape=(10000, 28, 28)

但如果我重塑

xtest2 = x_test.reshape(10000, 784)
x = pd.DataFrame(xtest2)
loaded_model.predict(x)

输入未对齐

ValueError: Input 0 of layer "sequential_2" is incompatible with the layer: expected shape=(None, 28, 28), found shape=(None, 784)

这是有意义的，因为该层被设置为

model = tf.keras.Sequential([
        tf.keras.layers.Conv2D(32, 3, activation='relu', input_shape=(28, 28, 1)),

但是如何同时满足panda要求和tensorflow 要求呢？