How to handle QNode using multiprocess for some CPUs

PennyLane Help
1

Hi Team. I’m using pennylane for measuring some quantum states. However, it’s too slow to measure, since I want to get the measurement under different parameters of the circuit. I didn’t find helpful demos in docs, and I’m not familiar with some packages like jit or jax. So I wish to use multiprocess to achieve a speedup by using multiple CPUs since I do not have a GPU. I’m wondering whether qml.QNode supports it? Do I need to init a QNode in each process, like the following code?

def main(n_process):
    processes = []
    for i in range(n_process):
        p = mp.Process(target=measurePorb_basic_process, args=(
            ...# some arguments for measureProb_basic_process
        ))
        processes.append(p)
        p.start()
    for p in processes:
        p.join()

def measurePorb_basic_process():
    qnode = qml.QNode(circuit, device=qml.device("default.qubit", wires=2, shots=100)) # take 2 bit 100 shots for example
    measurements = qnode(args) # get the measurements, suppose it returns qml.Sample(...)
    # do some postprocessing for `measurements` variable
    ...
def circuit(args): # define a circuit
    qml.RX(args, wires=0)   
    qml.CNOT(wires=[0,1])
    return [qml.sample(qml.PauliZ(i)) for i in range(2)]

Or, I just define the QNode out of measureProb_basic_process function, i.e., define it in main function? Thanks for your time.

2

Hi @Peng_Mi ,
You can define your QNode in your main function and use it as many times as you want later.
You can find some performance tips on pennylane.ai/performance. I would generally recommend using lightning.qubit if you’re using more than 10 qubits.

I hope this helps!