How to convert ndarray of shape (row, y, z) from mnist to pandas dataframe for tensorflow model pred-CodePudding

I am trying to predict model trained for mnist via mlflow

loaded_model = mlflow.pyfunc.load_model(logged_model)

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

I tried to create dataframe via

x = pd.DataFrame(x_test)

but i got

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

but if I reshape

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

I get input not aligned

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

this make sense since, the layer is setup like

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

but how do I satisfy both pandas requirement and tensorflow requirement?

CodePudding user response：

You could try reshaping before calling model.predict:

x = pd.DataFrame(xtest2)

model.predict(tf.keras.layers.Reshape((28, 28, 1))(x))