Expected min dim=2, found 1

PennyLane Help
1

I write simple qnn qnode code for mnist.
But it throws an error of dimension
I am not getting what is error about.
can you please look into it.

import pennylane as qml
import tensorflow as tf
from tensorflow.keras import layers
import numpy as np

# Define the quantum circuit for encoding
def qnode(params, x):
    qml.templates.AngleEmbedding(x, wires=range(9))
    qml.templates.BasicEntanglerLayers(params, wires=range(9))
    return qml.expval(qml.PauliZ(0))


class QNN(tf.keras.Model):
    def __init__(self):
        super().__init__()
        self.dense = layers.Dense(10)
        self.qnode = qml.QNode(qnode, device=qml.device('default.qubit',  wires=9), interface='tf')
        
    def call(self, inputs):
        # Split the image into patches
        patches = tf.image.extract_patches(inputs, sizes=[1,3,3,1], strides=[1,1,1,1], rates=[1,1,1,1],
                                           padding='VALID')
        patches = tf.reshape(patches, [-1, 9])
        
       weights=np.random.uniform(size=(3,9))
        # Apply the quantum circuit to each patch
        q_results = self.qnode(weights, patches)
        
    
        return self.dense(q_results)

# Load the MNIST dataset
mnist = tf.keras.datasets.mnist
(x_train, y_train), (x_test, y_test) = mnist.load_data()
x_train, x_test = x_train / 255.0, x_test / 255.0
x_train=x_train.reshape(x_train.shape[0], 28,28,1)


model = QNN()


model.compile(optimizer='adam',
              loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True),
              metrics=['accuracy'])

model.fit(x_train, y_train, epochs=10, batch_size=32)

extracted patches from each mnist image and run qnode over it.

2

Hey @Amandeep!

I don’t understand your code that well, but was able to reproduce the error with the following simpler code:

import pennylane as qml
import tensorflow as tf
from tensorflow.keras import layers
import numpy as np

tf.keras.backend.set_floatx('float64')

# Define the quantum circuit for encoding
def qnode(params, x):
    qml.templates.AngleEmbedding(x, wires=range(9))
    qml.templates.BasicEntanglerLayers(params, wires=range(9))
    return qml.expval(qml.PauliZ(0))


class QNN(tf.keras.Model):
    def __init__(self):
        super().__init__()
        self.dense = layers.Dense(10)
        self.qnode = qml.QNode(
            qnode, device=qml.device("default.qubit", wires=9), interface="tf"
        )

    def call(self, inputs):
        # Split the image into patches
        print("abc")
        print("inputs:", inputs)
        patches = tf.image.extract_patches(
            inputs,
            sizes=[1, 3, 3, 1],
            strides=[1, 1, 1, 1],
            rates=[1, 1, 1, 1],
            padding="VALID",
        )
        print("123")

        patches = tf.reshape(patches, [-1, 9])

        weights = np.random.uniform(size=(3, 9))
        # Apply the quantum circuit to each patch
        q_results = self.qnode(weights, patches)

        return self.dense(q_results)

x_train = np.random.uniform(0, 1, size=(2, 28, 28))
x_test = np.random.uniform(0, 1, size=(2, 28, 28))

y_train = np.random.randint(0, 9, size=(2,))
y_test = np.random.randint(0, 9, size=(2,))

x_train, x_test = x_train / 255.0, x_test / 255.0

model = QNN()

model.compile(
    optimizer="adam",
    loss=tf.keras.losses.SparseCategoricalCrossentropy(from_logits=True),
    metrics=["accuracy"],
)

model.fit(x_train, y_train, epochs=1, batch_size=1)

'''
Call arguments received by layer 'qnn_2' (type QNN):
      • inputs=tf.Tensor(shape=(1, 28, 28), dtype=float64)
'''

As the error suggests, something is wrong in your call function, presumably a dimensional issue. Based on the print statements that I put in that function, the error is occurring in the tf.image.extract_patches function. Hope this helps you debug!