BF gates in Tensorflow Quantum
RubensZimbres opened this issue · 2 comments
Hello, I want to develop a neural network considering noise from a real quantum machine with cirq, considering the following Circuit:
`
def make_GHZ_withbitflips(num_qubits,measurements=True):
myqubits = cirq.LineQubit.range(num_qubits)
GHZ_circuit = cirq.Circuit([
cirq.Moment([cirq.H(myqubits[0])]),
cirq.rx(a).on(q0),
cirq.ry(b).on(q1),cirq.ry(a).on(q2), cirq.CNOT(control=q2, target=q3), cirq.CNOT(control=q1, target=q0)
])
for x in range(num_qubits-1):
GHZ_circuit.append([cirq.CNOT(myqubits[x],myqubits[x+1]),cirq.bit_flip(p=0.1)(myqubits[x+1])])
if measurements:
GHZ_circuit.append(cirq.Moment(cirq.measure_each(*myqubits)))
return GHZ_circuit
`
However, it seems that Tensorfow Quantum does not support BF gates, is this right ?
I'm using Tensorflow Quantum 0.6.1 and cirq==1.0.0
The code that is not supported is: cirq.bit_flip(p=0.1)(myqubits[x+1])
The error:
{ "name": "InvalidArgumentError", "message": "Exception encountered when calling layer \"expectation_4\" (type Expectation).\n\nCould not parse gate id: BF. This is likely because a cirq.Channel was used in an op that does not support them. [Op:TfqSimulateExpectation]\n\nCall arguments received:\n • inputs=tf.Tensor(shape=(2,), dtype=string)\n • symbol_names=tf.Tensor(shape=(2,), dtype=string)\n • symbol_values=tf.Tensor(shape=(2, 2), dtype=float32)\n • operators=tf.Tensor(shape=(2, 2), dtype=string)\n • repetitions=None\n • initializer=<keras.initializers.initializers_v2.RandomUniform object at 0x7f403e14c310>", "stack": "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m\n\u001b[0;31mInvalidArgumentError\u001b[0m Traceback (most recent call last)\n\u001b[1;32m/home/theone/other_models/Quantum_Start/Quantum_ML_try_FOI_work3_noise.ipynb Cell 48\u001b[0m in \u001b[0;36m<cell line: 1>\u001b[0;34m()\u001b[0m\n\u001b[0;32m----> <a href='vscode-notebook-cell:/home/theone/other_models/Quantum_Start/Quantum_ML_try_FOI_work3_noise.ipynb#X64sZmlsZQ%3D%3D?line=0'>1</a>\u001b[0m model([datapoint_circuits, commands])\u001b[39m.\u001b[39mnumpy()\n\nFile \u001b[0;32m~/.conda/envs/tf27/lib/python3.9/site-packages/keras/utils/traceback_utils.py:67\u001b[0m, in \u001b[0;36mfilter_traceback.<locals>.error_handler\u001b[0;34m(*args, **kwargs)\u001b[0m\n\u001b[1;32m 65\u001b[0m \u001b[39mexcept\u001b[39;00m \u001b[39mException\u001b[39;00m \u001b[39mas\u001b[39;00m e: \u001b[39m# pylint: disable=broad-except\u001b[39;00m\n\u001b[1;32m 66\u001b[0m filtered_tb \u001b[39m=\u001b[39m _process_traceback_frames(e\u001b[39m.\u001b[39m__traceback__)\n\u001b[0;32m---> 67\u001b[0m \u001b[39mraise\u001b[39;00m e\u001b[39m.\u001b[39mwith_traceback(filtered_tb) \u001b[39mfrom\u001b[39;00m \u001b[39mNone\u001b[39m\n\u001b[1;32m 68\u001b[0m \u001b[39mfinally\u001b[39;00m:\n\u001b[1;32m 69\u001b[0m \u001b[39mdel\u001b[39;00m filtered_tb\n\nFile \u001b[0;32m~/.conda/envs/tf27/lib/python3.9/site-packages/tensorflow_quantum/python/layers/high_level/controlled_pqc.py:264\u001b[0m, in \u001b[0;36mControlledPQC.call\u001b[0;34m(self, inputs)\u001b[0m\n\u001b[1;32m 261\u001b[0m \u001b[39m# this is disabled to make autograph compilation easier.\u001b[39;00m\n\u001b[1;32m 262\u001b[0m \u001b[39m# pylint: disable=no-else-return\u001b[39;00m\n\u001b[1;32m 263\u001b[0m \u001b[39mif\u001b[39;00m \u001b[39mself\u001b[39m\u001b[39m.\u001b[39m_analytic:\n\u001b[0;32m--> 264\u001b[0m \u001b[39mreturn\u001b[39;00m \u001b[39mself\u001b[39;49m\u001b[39m.\u001b[39;49m_layer(model_appended,\n\u001b[1;32m 265\u001b[0m symbol_names\u001b[39m=\u001b[39;49m\u001b[39mself\u001b[39;49m\u001b[39m.\u001b[39;49m_symbols,\n\u001b[1;32m 266\u001b[0m symbol_values\u001b[39m=\u001b[39;49minputs[\u001b[39m1\u001b[39;49m],\n\u001b[1;32m 267\u001b[0m operators\u001b[39m=\u001b[39;49mtiled_up_operators)\n\u001b[1;32m 268\u001b[0m \u001b[39melse\u001b[39;00m:\n\u001b[1;32m 269\u001b[0m tiled_up_repetitions \u001b[39m=\u001b[39m tf\u001b[39m.\u001b[39mtile(\u001b[39mself\u001b[39m\u001b[39m.\u001b[39m_repetitions,\n\u001b[1;32m 270\u001b[0m [circuit_batch_dim, \u001b[39m1\u001b[39m])\n\nFile \u001b[0;32m~/.conda/envs/tf27/lib/python3.9/site-packages/tensorflow_quantum/python/layers/circuit_executors/expectation.py:356\u001b[0m, in \u001b[0;36mExpectation.call\u001b[0;34m(self, inputs, symbol_names, symbol_values, operators, repetitions, initializer)\u001b[0m\n\u001b[1;32m 353\u001b[0m \u001b[39mreturn\u001b[39;00m \u001b[39mself\u001b[39m\u001b[39m.\u001b[39m_expectation_op(inputs, symbol_names, symbol_values,\n\u001b[1;32m 354\u001b[0m operators, num_samples)\n\u001b[1;32m 355\u001b[0m \u001b[39melse\u001b[39;00m:\n\u001b[0;32m--> 356\u001b[0m \u001b[39mreturn\u001b[39;00m \u001b[39mself\u001b[39;49m\u001b[39m.\u001b[39;49m_expectation_op(inputs, symbol_names, symbol_values,\n\u001b[1;32m 357\u001b[0m operators)\n\nFile \u001b[0;32m~/.conda/envs/tf27/lib/python3.9/site-packages/tensorflow_quantum/python/differentiators/differentiator.py:156\u001b[0m, in \u001b[0;36mDifferentiator.generate_differentiable_op.<locals>.op_wrapper_analytic\u001b[0;34m(programs, symbol_names, symbol_values, pauli_sums)\u001b[0m\n\u001b[1;32m 153\u001b[0m \u001b[39m@tf\u001b[39m\u001b[39m.\u001b[39mcustom_gradient\n\u001b[1;32m 154\u001b[0m \u001b[39mdef\u001b[39;00m \u001b[39mop_wrapper_analytic\u001b[39m(programs, symbol_names, symbol_values,\n\u001b[1;32m 155\u001b[0m pauli_sums):\n\u001b[0;32m--> 156\u001b[0m forward_pass_vals \u001b[39m=\u001b[39m analytic_op(programs, symbol_names,\n\u001b[1;32m 157\u001b[0m symbol_values, pauli_sums)\n\u001b[1;32m 159\u001b[0m \u001b[39mdef\u001b[39;00m \u001b[39mgradient\u001b[39m(grad):\n\u001b[1;32m 160\u001b[0m \u001b[39mreturn\u001b[39;00m \u001b[39mself\u001b[39m\u001b[39m.\u001b[39m_differentiate_ana(programs, symbol_names,\n\u001b[1;32m 161\u001b[0m symbol_values, pauli_sums,\n\u001b[1;32m 162\u001b[0m forward_pass_vals, grad)\n\nFile \u001b[0;32m~/.conda/envs/tf27/lib/python3.9/site-packages/tensorflow_quantum/core/ops/circuit_execution_ops.py:135\u001b[0m, in \u001b[0;36mget_expectation_op.<locals>.<lambda>\u001b[0;34m(programs, symbol_names, symbol_values, pauli_sums)\u001b[0m\n\u001b[1;32m 131\u001b[0m \u001b[39mif\u001b[39;00m op \u001b[39mis\u001b[39;00m \u001b[39mnot\u001b[39;00m \u001b[39mNone\u001b[39;00m:\n\u001b[1;32m 132\u001b[0m \u001b[39mif\u001b[39;00m quantum_concurrent \u001b[39mis\u001b[39;00m \u001b[39mTrue\u001b[39;00m:\n\u001b[1;32m 133\u001b[0m \u001b[39m# Return an op that does not block graph level parallelism.\u001b[39;00m\n\u001b[1;32m 134\u001b[0m \u001b[39mreturn\u001b[39;00m \u001b[39mlambda\u001b[39;00m programs, symbol_names, symbol_values, pauli_sums: \\\n\u001b[0;32m--> 135\u001b[0m op(programs, symbol_names, symbol_values, pauli_sums)\n\u001b[1;32m 137\u001b[0m \u001b[39m# Return an op that does block graph level parallelism.\u001b[39;00m\n\u001b[1;32m 138\u001b[0m \u001b[39mreturn\u001b[39;00m \u001b[39mlambda\u001b[39;00m programs, symbol_names, symbol_values, pauli_sums: \\\n\u001b[1;32m 139\u001b[0m quantum_context\u001b[39m.\u001b[39m_GLOBAL_OP_LOCK\u001b[39m.\u001b[39mexecute(\u001b[39mlambda\u001b[39;00m: op(\n\u001b[1;32m 140\u001b[0m programs, symbol_names, symbol_values, pauli_sums))\n\nFile \u001b[0;32m~/.conda/envs/tf27/lib/python3.9/site-packages/tensorflow_quantum/core/ops/tfq_simulate_ops.py:44\u001b[0m, in \u001b[0;36mtfq_simulate_expectation\u001b[0;34m(programs, symbol_names, symbol_values, pauli_sums)\u001b[0m\n\u001b[1;32m 22\u001b[0m \u001b[39mdef\u001b[39;00m \u001b[39mtfq_simulate_expectation\u001b[39m(programs, symbol_names, symbol_values, pauli_sums):\n\u001b[1;32m 23\u001b[0m \u001b[39m\"\"\"Calculate the expectation value of circuits wrt some operator(s)\u001b[39;00m\n\u001b[1;32m 24\u001b[0m \n\u001b[1;32m 25\u001b[0m \u001b[39m Args:\u001b[39;00m\n\u001b[0;32m (...)\u001b[0m\n\u001b[1;32m 42\u001b[0m \u001b[39m (after resolving the corresponding parameters in).\u001b[39;00m\n\u001b[1;32m 43\u001b[0m \u001b[39m \"\"\"\u001b[39;00m\n\u001b[0;32m---> 44\u001b[0m \u001b[39mreturn\u001b[39;00m SIM_OP_MODULE\u001b[39m.\u001b[39;49mtfq_simulate_expectation(\n\u001b[1;32m 45\u001b[0m programs, symbol_names, tf\u001b[39m.\u001b[39;49mcast(symbol_values, tf\u001b[39m.\u001b[39;49mfloat32), pauli_sums)\n\nFile \u001b[0;32m<string>:48\u001b[0m, in \u001b[0;36mtfq_simulate_expectation\u001b[0;34m(programs, symbol_names, symbol_values, pauli_sums, name)\u001b[0m\n\n\u001b[0;31mInvalidArgumentError\u001b[0m: Exception encountered when calling layer \"expectation_4\" (type Expectation).\n\nCould not parse gate id: BF. This is likely because a cirq.Channel was used in an op that does not support them. [Op:TfqSimulateExpectation]\n\nCall arguments received:\n • inputs=tf.Tensor(shape=(2,), dtype=string)\n • symbol_names=tf.Tensor(shape=(2,), dtype=string)\n • symbol_values=tf.Tensor(shape=(2, 2), dtype=float32)\n • operators=tf.Tensor(shape=(2, 2), dtype=string)\n • repetitions=None\n • initializer=<keras.initializers.initializers_v2.RandomUniform object at 0x7f403e14c310>" }
Bit flips are a supported noise channel. You can see all supported noise channels via: tfq.util.get_supported_channels() which will return (ignore the values):
{cirq.amplitude_damp(gamma=0.01): 1,
cirq.asymmetric_depolarize(error_probabilities={'I': 0.94, 'X': 0.01, 'Y': 0.02, 'Z': 0.03}): 1,
cirq.bit_flip(p=0.01): 1,
cirq.depolarize(p=0.01): 1,
cirq.generalized_amplitude_damp(p=0.01,gamma=0.02): 1,
cirq.phase_damp(gamma=0.01): 1,
cirq.phase_flip(p=0.01): 1,
cirq.ResetChannel(): 1}
I just verified this with something similar to your GHZ example:
import tensorflow_quantum as tfq
import cirq
def make_GHZ(qubits):
c = cirq.Circuit()
c += cirq.H(qubits[0])
for i in range(len(qubits) - 1):
c += cirq.CNOT(qubits[i], qubits[i + 1])
return c
qs = [cirq.GridQubit(0, i) for i in range(3)]
noise_test = make_GHZ(qs).with_noise(cirq.bit_flip(p=0.05))
noiseless_test = make_GHZ(qs)
ops = [cirq.Z(i) for i in qs]
exp = tfq.layers.Expectation()(noiseless_test, symbol_names=[], symbol_values=[[]], operators=ops)
print(exp)
exp = tfq.layers.SampledExpectation(backend='noisy')(noise_test, symbol_names=[], symbol_values=[[]], operators=ops, repetitions=1000)
print(exp)Which returned
tf.Tensor([[0. 0. 0.]], shape=(1, 3), dtype=float32)
tf.Tensor([[ 0.002 -0.024 0.028]], shape=(1, 3), dtype=float32)
For more information on simulating noise with TFQ, I recommend: https://www.tensorflow.org/quantum/tutorials/noise
I would guess there is something else in the code causing your error, but I cannot tell from what is provided. I would recommend linking to the minimal viable example to demonstrate the error. Additionally, you can format the code neater using three of the ` symbols e.g. ```python CODE ```
Thanks very much, I will work on that.
UPDATE: I used tfq.layers.NoisyControlledPQC and it worked.