I meet a question when I use qml.qnn.KerasLayer. I want to build a hybrid classical-quantum model. But when I run the code, a question appear. Here is my code.
import os,random
os.environ["KERAS_BACKEND"] = "tensorflow"
os.environ["CUDA_VISIBLE_DEVICES"] = "0"
import numpy as np
import matplotlib.pyplot as plt
import scipy.io as scio
import pickle,sys,h5py
import keras
import keras.backend as K
from keras.callbacks import LearningRateScheduler
import tensorflow as tf
import pennylane as qml
import mltools, dataset2016, dataset_train_alldb_allmods2
import csv
# from tensorflow.keras.layers import Input,Dense,Conv1D,MaxPool1D,ReLU,Dropout,Softmax,concatenate,Flatten,Reshape,Activation
#set Keras data format as channels_first
K.set_image_data_format('channels_last')
print(K.image_data_format())
(mods, snrs, lbl), (X_train, Y_train, Y_train_hot), (X_val, Y_val, Y_val_hot), (X_test, Y_test, Y_test_hot), (train_idx, val_idx, test_idx), file_num = \
dataset_train_alldb_allmods2.load_data()
print(X_train.shape)
print(X_test.shape)
classes = mods
# Set up some params
nb_epoch = 3 # number of epochs to train on
batch_size = 512 # training batch size
n_qubits = 2
n_layers = 2
dev = qml.device("default.qubit", wires=n_qubits)
@qml.qnode(dev)
def qnode(inputs, weights):
qml.RX(inputs[0], wires=0)
qml.RX(inputs[1], wires=1)
qml.RY(weights[0,0], wires=0)
qml.RY(weights[1,1], wires=1)
qml.CNOT(wires=[0, 1])
return qml.expval(qml.PauliZ(0)), qml.expval(qml.PauliZ(1))
# model=mcl.DNN3()
dr =0.2
# print(weight_shapes)
weight_shapes = {"weights": (n_layers, n_qubits)}
qlayer = qml.qnn.KerasLayer(qnode, weight_shapes, output_dim=2)
input = tf.keras.layers.Input(shape=(5,))
Dense_1 = tf.keras.layers.Dense(5)
QNN_1 = qml.qnn.KerasLayer(qnode, weight_shapes, output_dim=2, name="QNN_1")
output = tf.keras.layers.Dense(2, activation="softmax")
model = tf.keras.models.Sequential([input, Dense_1, QNN_1, output])
model.compile(loss='categorical_crossentropy',
metrics=['accuracy'],
optimizer='adam')
model.summary()
# Train the framework (model)
filepath = 'weights/weights.h5'
history = model.fit(X_train,
Y_train_hot,
batch_size=batch_size,
epochs=nb_epoch,
verbose=2,
#validation_data=([X_test,X1_test,X2_test],Y_test),
validation_data=(X_val, Y_val_hot),
callbacks = [
keras.callbacks.ModelCheckpoint(filepath, monitor='val_loss', verbose=1, save_best_only=True, mode='auto'),
keras.callbacks.ReduceLROnPlateau(monitor='val_loss',factor=0.5,verbose=1,patince=5,min_lr=0.0000001),
keras.callbacks.EarlyStopping(monitor='val_loss', patience=30, verbose=1, mode='auto')
]
)
mltools.show_history(history)
And the full error message below:
Traceback (most recent call last):
File "E:\Postgraduate\Projects\AMC\AMR-Benchmark-main\HCQDNN3\pytorch_pennelane.py", line 75, in <module>
history = model.fit(X_train,
File "D:\Anaconda\envs\Tensorflow-gpu-python3.8\lib\site-packages\keras\engine\training.py", line 1230, in fit
callbacks.on_epoch_end(epoch, epoch_logs)
File "D:\Anaconda\envs\Tensorflow-gpu-python3.8\lib\site-packages\keras\callbacks.py", line 413, in on_epoch_end
callback.on_epoch_end(epoch, logs)
File "D:\Anaconda\envs\Tensorflow-gpu-python3.8\lib\site-packages\keras\callbacks.py", line 1368, in on_epoch_end
self._save_model(epoch=epoch, batch=None, logs=logs)
File "D:\Anaconda\envs\Tensorflow-gpu-python3.8\lib\site-packages\keras\callbacks.py", line 1422, in _save_model
self.model.save(filepath, overwrite=True, options=self._options)
File "D:\Anaconda\envs\Tensorflow-gpu-python3.8\lib\site-packages\keras\engine\training.py", line 2145, in save
save.save_model(self, filepath, overwrite, include_optimizer, save_format,
File "D:\Anaconda\envs\Tensorflow-gpu-python3.8\lib\site-packages\keras\saving\save.py", line 145, in save_model
hdf5_format.save_model_to_hdf5(
File "D:\Anaconda\envs\Tensorflow-gpu-python3.8\lib\site-packages\keras\saving\hdf5_format.py", line 110, in save_model_to_hdf5
model_metadata = saving_utils.model_metadata(model, include_optimizer)
File "D:\Anaconda\envs\Tensorflow-gpu-python3.8\lib\site-packages\keras\saving\saving_utils.py", line 148, in model_metadata
raise e
File "D:\Anaconda\envs\Tensorflow-gpu-python3.8\lib\site-packages\keras\saving\saving_utils.py", line 145, in model_metadata
model_config['config'] = model.get_config()
File "D:\Anaconda\envs\Tensorflow-gpu-python3.8\lib\site-packages\keras\engine\sequential.py", line 411, in get_config
layer_configs.append(generic_utils.serialize_keras_object(layer))
File "D:\Anaconda\envs\Tensorflow-gpu-python3.8\lib\site-packages\keras\utils\generic_utils.py", line 508, in serialize_keras_object
raise e
File "D:\Anaconda\envs\Tensorflow-gpu-python3.8\lib\site-packages\keras\utils\generic_utils.py", line 503, in serialize_keras_object
config = instance.get_config()
File "D:\Anaconda\envs\Tensorflow-gpu-python3.8\lib\site-packages\keras\engine\base_layer.py", line 727, in get_config
raise NotImplementedError('Layer %s has arguments in `__init__` and '
NotImplementedError: Layer KerasLayer has arguments in `__init__` and therefore must override `get_config`.
Why I can’t build the model? I think I obey the rul when I use ‘qml.qnn.KerasLayer’
If I delete the QNN_1 , let model = tf.keras.models.Sequential([input, Dense_1, output]), the code can run successfully.
. Indeed, upgrading your pennylane version is a good thing to check when you encounter issues. You can see the paper trail of new features, improvements, breaking changes, etc., in the release notes: 
. There are a lot of improvements to KerasLayer!
