ssh -i "qutrunk-asia.pem" ubuntu@18.141.145.32复制代码

ubuntu@ip-192-168-150-172:~$ pip3 install --upgrade pip  -i https://pypi.tuna.tsinghua.edu.cn/simpleubuntu@ip-192-168-150-172:~$ pip3 install qutrunk  -i https://pypi.tuna.tsinghua.edu.cn/simple复制代码

ubuntu@ip-192-168-150-172:~$ python3Python 3.10.7 (main, Dec 20 2022, 07:32:05) [GCC 9.4.0] on linuxType "help", "copyright", "credits" or "license" for more information.
>>> import qutrunk>>> qutrunk.run_check()============QuSL instruction:===========qreg q[2]creg c[2]H * q[0]MCX(1) * (q[0], q[1])Measure * q[0]Measure * q[1]===============Draw circuit=============      ┌───┐      ┌─┐q[0]: ┤ H ├──■───┤M├───      └───┘┌─┴──┐└╥┘┌─┐q[1]: ─────┤ CX ├─╫─┤M├           └────┘ ║ └╥┘ c: 2/════════════╩══╩═                  0  1==========circuit running result=========[{"0b00": 57}, {"0b11": 43}]===========circuit running info=========={"backend": "BackendLocal", "task_id": "7478c36b8ec941dbaed9b14f64a854a3", "status": "success", "arguments": {"shots": 100}}qutrunk v0.2.0 is installed successfully! You can use QuTrunk now.复制代码

ubuntu@ip-192-168-150-172:~$ jupyter notebook passwordEnter password:Verify password:[NotebookPasswordApp] Wrote hashed password to /home/ubuntu/.jupyter/jupyter_notebook_config.json复制代码

$ jupyter notebook  --ip=0.0.0.0复制代码

import random
import numpy as npimport tensorflow as tffrom matplotlib import pyplot as plt
from qutrunk.circuit import QCircuitfrom qutrunk.circuit.gates import Rx, Ry, Rz, PauliZ复制代码

def datapoint_circuit(rotations):    beta0, beta1, beta2 = [float(f) for f in rotations]    circuit = QCircuit()    qreg = circuit.allocate(1)    Rx(beta0) * qreg[0]    Ry(beta1) * qreg[0]    Rz(beta2) * qreg[0]    return circuit复制代码

def model_circuit():    circuit = QCircuit()    q = circuit.allocate(1)    angles = ["theta-1", "theta-2", "theta-3"]    params = circuit.create_parameters(angles)
    Rz(params[0]) * q[0]    Ry(params[1]) * q[0]    Rx(params[2]) * q[0]
    return circuit复制代码

def expectation():    circuit = QCircuit()    qreg = circuit.allocate(1)    return PauliZ(qreg[0])复制代码

def join_circuit(datapoint_cir, model_cir, inputs):    params = {"theta-" + str(i): inputs[i] for i in range(len(inputs))}    model_cir = model_cir.bind_parameters(params)    datapoint_cir.append_circuit(model_cir)    return datapoint_cir复制代码

class CustomGradientTape:    def __init__(self, inputs, exp_op, shift=np.pi / 2):        self.inputs = inputs        self.exp_op = exp_op        self.shift = shift
    def gradient(self, loss, var_list):        params = [var.numpy() for var in var_list]
        gradients = []        for i in range(len(params)):            shift_right = np.copy(params)            shift_right[i] += self.shift            shift_left = np.copy(params)            shift_left[i] -= self.shift
            circuit = join_circuit(datapoint_circuit(self.inputs), model_circuit(), shift_right)            expectation_right = -1 * circuit.expval_pauli(self.exp_op)
            circuit = join_circuit(datapoint_circuit(self.inputs), model_circuit(), shift_left)            expectation_left = -1 * circuit.expval_pauli(self.exp_op)
            gradient = expectation_right - expectation_left            gradients.append(tf.convert_to_tensor(gradient, dtype=tf.float32))        return gradients复制代码

class ControlledPQC(tf.keras.layers.Layer):
    def __init__(self, optimizer, exp_op, var_list, grad):        super(ControlledPQC, self).__init__()        self.optimizer = optimizer        self.exp_op = exp_op        self.var_list = var_list
        self.grad = grad
    def call(self, inputs):        params_list = [var.numpy() for var in self.var_list]        circuit = join_circuit(datapoint_circuit(inputs), model_circuit(), params_list)        loss = -1 * circuit.expval_pauli(self.exp_op)        self.optimizer.minimize(loss, self.var_list, self.grad)
        return loss复制代码

ITR = 200LR = 0.02rand = random.Random()random_rotations = tf.convert_to_tensor(np.random.uniform(0, 2 * np.pi, 3))op = expectation()
opt = tf.keras.optimizers.Adam(learning_rate=LR)复制代码

control_params = [tf.Variable(rand.uniform(0, 2 * np.pi)) for _ in range(3)]model = ControlledPQC(opt, op, control_params, CustomGradientTape(random_rotations, op))
loss_list = []for it in range(ITR):    ls = model(random_rotations)    loss_list.append(ls)    if it % 10 == 0:        print(f"{it}: {ls:.4f}")复制代码

0: -0.239910: -0.620320: -0.868930: -0.974640: -0.996950: -0.996760: -0.997370: -0.999180: -0.999990: -1.0000100: -1.0000110: -1.0000120: -1.0000130: -1.0000140: -1.0000150: -1.0000160: -1.0000170: -1.0000180: -1.0000190: -1.0000复制代码

plt.plot(loss_list)plt.title("Learning to Control a Qubit")plt.xlabel("Iterations")plt.ylabel("loss")复制代码