本文代码为深度学习一书线性回归从零实现对应代码的gpu版本，只需将默认存储在cpu的数据改为存储在gpu即可自动在gpu进行计算
from IPython import display
from matplotlib import pyplot as plt
from mxnet import autograd,nd
import mxnet as mx
import random
import d2lzh#生成数据集
nums_input = 2#特征数
num_examples = 1000#数据个数
true_w = [2,-3.4]#真实权重w
true_b = 4.2#真实权重b
features = nd.random.normal(scale=1,shape=(num_examples,nums_input),ctx=mx.gpu())#标准差为1
labels = true_w[0] * features[:,0] + true_w[1] * features[:,1] + true_b
labels += nd.random.normal(scale=0.01,shape=labels.shape,ctx=mx.gpu())#噪声print(labels)def use_svg_display():#矢量图表示display.set_matplotlib_formats('svg')
def set_figsize(figsize = (3.5,2.5)):use_svg_display()#设置图的尺寸plt.rcParams['figure.figsize'] = figsizedef data_iter(batch_size, features, labels):#读取小批量数据集num_examples = len(features)indices = list(range(num_examples))random.shuffle(indices)#随机for i in range(0,num_examples,batch_size):j = nd.array(indices[i: min(i + batch_size, num_examples)],ctx=mx.gpu())yield features.take(j),labels.take(j)#返回索引元素d2lzh.set_figsize()
plt.scatter(features[:,0].asnumpy(),labels.asnumpy(),1)#加分号只显示图batch_size = 10#初始化权重
w = nd.random.normal(scale=0.01,shape=(nums_input,1),ctx=mx.gpu())
b = nd.zeros(shape=(1,),ctx=mx.gpu())
#创建梯度
w.attach_grad()
b.attach_grad()#定义回归模型
def linreg(X,w,b):return nd.dot(X,w) + b#定义损失函数
def squared_loss(y_hat,y):return (y_hat - y.reshape(y_hat.shape)) ** 2 / 2# return nd.dot(y_hat.T,y) / (2 * n)#定义优化算法
def sgd(params, lr, batch_size):for param in params:param[:] = param - lr * param.grad / batch_sizelr = 0.03
num_epochs = 4#迭代周期
net = linreg
loss = squared_lossfor epoch in range(num_epochs):#每一个迭代周期，使用训练集所有样本一次for X,y in data_iter(batch_size, features,labels):with autograd.record():l = loss(net(X,w,b),y)l.backward()sgd([w,b],lr,batch_size)train_l = loss(net(features,w,b),labels)print('epoch %d, loss %f' % (epoch + 1,train_l.mean().asnumpy()))