def test_torch(self, tol):
"""Test that the covariance matrix computes the correct
result, and is differentiable, using the Torch interface"""
dev = qml.device("default.qubit", wires=3)
@qml.qnode(dev, interface="torch")
def circuit(weights):
"""Returns the shared probability distribution of ansatz
in the joint basis for obs_list"""
self.ansatz(weights, wires=dev.wires)
for o in self.obs_list:
o.diagonalizing_gates()
return qml.probs(wires=[0, 1, 2])
weights = np.array([0.1, 0.2, 0.3])
weights_t = torch.tensor(weights, requires_grad=True)
probs = circuit(weights_t)
res = fn.cov_matrix(probs, self.obs_list)
expected = self.expected_cov(weights)
assert np.allclose(res.detach().numpy(), expected, atol=tol, rtol=0)
loss = res[0, 1]
loss.backward()
res = weights_t.grad
expected = self.expected_grad(weights)
assert np.allclose(res.detach().numpy(), expected, atol=tol, rtol=0)
def test_tf(self, tol):
"""Test that the covariance matrix computes the correct
result, and is differentiable, using the TF interface"""
dev = qml.device("default.qubit", wires=3)
@qml.qnode(dev, interface="tf")
def circuit(weights):
"""Returns the shared probability distribution of ansatz
in the joint basis for obs_list"""
self.ansatz(weights, wires=dev.wires)
for o in self.obs_list:
o.diagonalizing_gates()
return qml.probs(wires=[0, 1, 2])
weights = np.array([0.1, 0.2, 0.3])
weights_t = tf.Variable(weights)
with tf.GradientTape() as tape:
probs = circuit(weights_t)
cov = fn.cov_matrix(probs, self.obs_list)
loss = cov[0, 1]
expected = self.expected_cov(weights)
assert np.allclose(cov, expected, atol=tol, rtol=0)
grad = tape.gradient(loss, weights_t)
expected = self.expected_grad(weights)
assert np.allclose(grad, expected, atol=tol, rtol=0)
def cov(weights):
probs = circuit(weights)
return fn.cov_matrix(probs, self.obs_list)
def cov(weights):
probs = circuit(weights)
return fn.cov_matrix(probs, obs_list, wires=dev.wires)