Correctly Implementing CVNN

PennyLane Help
1

Hello!

I’m trying to correctly implement CVNeuralNetLayers with TorchLayer. For some reason I just can’t get it right.

Below is some example code pulled from the PennyLane website:

import torch
from sklearn.datasets import make_moons

X, y = make_moons(n_samples=200, noise=0.1)
y_ = torch.unsqueeze(torch.tensor(y), 1)
y_hot = torch.scatter(torch.zeros((200, 2)), 1, y_, 1)

import pennylane as qml

dev = qml.device("default.qubit", wires=2)

@qml.qnode(dev)
def qnode(inputs, weights):
    qml.AngleEmbedding(inputs, wires=range(2))
    qml.CVNeuralNetLayers(*weights, wires=range(2))
    return [qml.expval(qml.PauliZ(wires=i)) for i in range(2)]

shapes = qml.CVNeuralNetLayers.shape(n_layers=2, n_wires=2)
weight_shapes = {"weights" : [shapes]}

qlayer = qml.qnn.TorchLayer(qnode, weight_shapes)

clayer_1 = torch.nn.Linear(2, 2)
clayer_2 = torch.nn.Linear(2, 2)
softmax = torch.nn.Softmax(dim=1)
model = torch.nn.Sequential(clayer_1, qlayer, clayer_2, softmax)

opt = torch.optim.SGD(model.parameters(), lr=0.2)
loss = torch.nn.L1Loss()

X = torch.tensor(X, requires_grad=True).float()
y_hot = y_hot.float()

batch_size = 5
batches = 200 // batch_size

data_loader = torch.utils.data.DataLoader(
    list(zip(X, y_hot)), batch_size=5, shuffle=True, drop_last=True
)

for epoch in range(2):

    running_loss = 0

    for xs, ys in data_loader:
        opt.zero_grad()

        loss_evaluated = loss(model(xs), ys)
        loss_evaluated.backward()

        opt.step()

The error that I get back returns as its last error:

---------------------------------------------------------------------------

IndexError                                Traceback (most recent call last)
...
/usr/local/lib/python3.7/dist-packages/pennylane/templates/layers/cv_neural_net.py in <listcomp>(.0)
    116 
    117         # check second dimensions
--> 118         second_dims = [s[1] for s in shapes]
    119         expected = [n_if] * 2 + [n_wires] * 3 + [n_if] * 2 + [n_wires] * 4
    120         if not all(e == d for e, d in zip(expected, second_dims)):

IndexError: tuple index out of range

I’m trying to figure out how to make this work nicely with TorchLayer but I’m all out ideas. Any thoughts on how to correctly implement this?

Any help or advice would be very much appreciated.

2

Hi @QML,

You have a problem here because CVNeuralNetLayers is for continuous variable quantum computing while AngleEmbedding is for qubit-based quantum computing. You cannot use both of them on the same device. Maybe the confusion is regarding CVNeuralNetLayers. What is the functionality that you want to achieve? There may be another function that can help you.