08_P3 pycharm代码
import random # 需要随机化梯度下降和随机初始化参数
import torch
# 定义模型 预测y 即y_hat
def linreg(X, w, b):
"""线性回归模型"""
return torch.matmul(X, w) + b
# 定义合成数据的函数
def synthetic_data(w, b, num_examples):
"""生成y=Xw+b+噪声"""
X = torch.normal(0, 1, (num_examples, len(w))) # 生成均值为0方差为1的随机数,大小为num_examples列数为len(w)
y = torch.matmul(X, w) + b
y += torch.normal(0, 0.01, y.shape) # 加上随机噪音
return X, y.reshape((-1, 1)) # 列数为1,行数需要计算
# 小批量数据读取
def data_iter(batch_size, features, labels):
num_examples = len(features) # num_examples为X中的行数 即多少组数据
indices = list(range(num_examples)) # 生成索引列表[1,2,3,....,500]
random.shuffle(indices) # 随机读取索引 [23,42,103,....,487]
for i in range(0, num_examples, batch_size): # 步长为batch_size
batch_indices = torch.tensor( # 生成数据块,从列表的第i个元素到i+batch_size的元素即每次取10个
indices[i: min(i + batch_size, num_examples)]) # 例 第一次取到[23,42,103,....,368]
yield features[batch_indices], labels[batch_indices] # 不断返回features和batch_indices中第23,42,103,....,368行数据
# 定义损失函数
def squared_loss(y_hat, y):
"""均方损失"""
return (y_hat - y.reshape(y_hat.shape)) ** 2 / 2
# 定义优化算法
def sgd(params, lr, batch_size):
"""小批量随机梯度下降"""
with torch.no_grad():
for param in params:
param -= lr * param.grad / batch_size # 除不除batch_size变化不大
param.grad.zero_() # 避免梯度累计
if __name__ == '__main__':
# 初始化模型参数
w = torch.normal(0, 0.01, size=(2, 1), requires_grad=True) # 生成均值为0方差为0.01的随机向量,2行1列
b = torch.zeros(1, requires_grad=True)
# 给出真实值true_w true_b
true_w = torch.tensor([2, -3.4])
true_b = 4.2
# 构建数据集X作为features,并通过真实值true_w true_b算出y的真实值作为labels
features, labels = synthetic_data(true_w, true_b, 1000) # 生成特征和标注。features, labels分别接收函数的返回值X、y
# 训练的超参数
batch_size = 10
lr = 0.03 # 学习率0.03
num_epochs = 3 # 数据集扫三遍
net = linreg # 训练模型
loss = squared_loss # 均方损失
for epoch in range(num_epochs): # 扫三遍
for X, y in data_iter(batch_size, features, labels): # 拿出批量大小的赋给X,y。 每次拿10行,拿50次。
# X和y的小批量损失
l = loss(net(X, w, b), y) # 把X, w, b放入linreg中做预测,并与真实的y做均方损失
# 因为l形状是(batch_size,1),而不是一个标量。l中的所有元素被加到一起,
# 并以此计算关于[w,b]的梯度
l.sum().backward() # 求和并算梯度
sgd([w, b], lr, batch_size) # 使用参数的梯度更新参数 对w,b进行更新
with torch.no_grad(): # 做预测,不需要梯度
train_l = loss(net(features, w, b), labels) # 放入整个数据
print(f'epoch {epoch + 1}, loss {float(train_l.mean()):f}') # .mean()求均值
print(f'w的估计误差: {true_w - w.reshape(true_w.shape)}')
print(f'b的估计误差: {true_b - b}')
笔记
08_P4 pycharm代码
import torch
from torch.utils import data
from d2l import torch as d2l
from torch import nn
def load_array(data_arrays, batch_size, is_train=True):
"""用(features, labels)构造一个PyTorch数据迭代器"""
dataset = data.TensorDataset(*data_arrays) # 生成数据集dataset
return data.DataLoader(dataset, batch_size, shuffle=is_train) # 随机返回dataset的batch_size行数据
if __name__ == '__main__':
true_w = torch.tensor([2, -3.4])
true_b = 4.2
features, labels = d2l.synthetic_data(true_w, true_b, 1000)
batch_size = 10
data_iter = load_array((features, labels), batch_size) # 构造并得到小批量数据,是一个矩阵
next(iter(data_iter)) # 将data_iter转成python的iter,通过next不断获取并返回iter的下一条数据,即矩阵的下一行
# Sequential顺序容器,模块将按照在构造函数中传递的顺序添加到模块中。
net = nn.Sequential(nn.Linear(2, 1)) # 导入线性回归模型,输入的维度为2(X的列数为2),输出y_hat维度为1
net[0].weight.data.normal_(0, 0.01) # 模型中第一层(net[0])用随机正态分布改写权重参数。可以不设,使用默认。
net[0].bias.data.fill_(0) # 偏差设为0
loss = nn.MSELoss() # 使用均方误差,也称为平方 L2 范数,默认情况下,它返回所有样本损失的平均值
trainer = torch.optim.SGD(net.parameters(), lr=0.03) # 传入所有待优化的参数(w,b)
# 训练
num_epochs = 3
for epoch in range(num_epochs):
for X, y in data_iter:
l = loss(net(X), y)
trainer.zero_grad()
l.backward()
trainer.step() # 调用优化器来更新模型参数
l = loss(net(features), labels)
print(f'epoch {epoch + 1}, loss {l:f}')
w = net[0].weight.data
print('w的估计误差:', true_w - w.reshape(true_w.shape))
b = net[0].bias.data
print('b的估计误差:', true_b - b)
09_P4 pycharm代码
import torch
from IPython import display
from d2l import torch as d2l
# 动画绘制
class Animator:
"""在动画中绘制数据"""
def __init__(self, xlabel=None, ylabel=None, legend=None, xlim=None,
ylim=None, xscale='linear', yscale='linear',
fmts=('-', 'm--', 'g-.', 'r:'), nrows=1, ncols=1,
figsize=(3.5, 2.5)):
# 增量地绘制多条线
if legend is None:
legend = []
d2l.use_svg_display()
self.fig, self.axes = d2l.plt.subplots(nrows, ncols, figsize=figsize)
if nrows * ncols == 1:
self.axes = [self.axes, ]
# 使用lambda函数捕获参数
self.config_axes = lambda: d2l.set_axes(
self.axes[0], xlabel, ylabel, xlim, ylim, xscale, yscale, legend)
self.X, self.Y, self.fmts = None, None, fmts
def add(self, x, y):
# 向图表中添加多个数据点
if not hasattr(y, "__len__"):
y = [y]
n = len(y)
if not hasattr(x, "__len__"):
x = [x] * n
if not self.X:
self.X = [[] for _ in range(n)]
if not self.Y:
self.Y = [[] for _ in range(n)]
for i, (a, b) in enumerate(zip(x, y)):
if a is not None and b is not None:
self.X[i].append(a)
self.Y[i].append(b)
self.axes[0].cla()
for x, y, fmt in zip(self.X, self.Y, self.fmts):
self.axes[0].plot(x, y, fmt)
self.config_axes()
display.display(self.fig)
display.clear_output(wait=True)
# 定义softmax操作
def softmax(X):
X_exp = torch.exp(X)
partition = X_exp.sum(1, keepdim=True) # 按轴求和,keepdim=True保持原矩阵维度不变
return X_exp / partition # 这里应用了广播机制
# 定义模型
def net(X):
return softmax(torch.matmul(X.reshape((-1, W.shape[0])), W) + b) # -1自动算行数(256),W.shape[0]返回w的行数
'''
# 定义损失函数(交叉熵损失)
y = torch.tensor([0, 2])
y_hat = torch.tensor([[0.1, 0.3, 0.6], [0.3, 0.2, 0.5]])
# z = y_hat[[0, 1], y] # 从y_hat[0]中取出第y[0]个元素,y_hat[0][0].从y_hat[1]中取出第y[1]个元素,y_hat[1][2]
'''
def cross_entropy(y_hat, y):
return - torch.log(y_hat[range(len(y_hat)), y])
# 将预测类别y_hat与真实的y元素进行比较
def accuracy(y_hat, y):
"""计算预测正确的数量"""
if len(y_hat.shape) > 1 and y_hat.shape[1] > 1: # len(y_hat.shape)行数大于1,y_hat.shape[1]列数大于1.
y_hat = y_hat.argmax(axis=1) # 将y_hat每一行最大的值的索引取出放入y_hat
cmp = y_hat.type(y.dtype) == y # cmp为每个元素都为布尔值的张量
return float(cmp.type(y.dtype).sum()) # 将cmp中数据类型转换成y中的数据类型,false=0.true=1.返回预测正确的个数
class Accumulator:
"""在n个变量上累加"""
def __init__(self, n):
self.data = [0.0] * n
def add(self, *args):
self.data = [a + float(b) for a, b in zip(self.data, args)]
def reset(self):
self.data = [0.0] * len(self.data)
def __getitem__(self, idx):
return self.data[idx]
def evaluate_accuracy(net, data_iter):
"""计算在指定数据集上模型的精度"""
if isinstance(net, torch.nn.Module):
net.eval() # 将模型设置为评估模式
metric = Accumulator(2) # 正确预测数、预测总数
with torch.no_grad():
for X, y in data_iter:
metric.add(accuracy(net(X), y), y.numel())
return metric[0] / metric[1]
def updater(batch_size):
return d2l.sgd([W, b], lr, batch_size)
# 训练 迭代一次的流程 .nn还是还是自定义
def train_epoch_ch3(net, train_iter, loss, updater):
"""训练模型一个迭代周期(定义见第3章)"""
# 将模型设置为训练模式
if isinstance(net, torch.nn.Module):
net.train()
# 长度为3的迭代器累加信息。 训练损失总和、训练准确度总和、样本数
metric = Accumulator(3)
for X, y in train_iter: # 循环235次
# 计算梯度并更新参数
# print(X.shape) [256,1,28,28]
# print(y.shape) [256]
y_hat = net(X)
l = loss(y_hat, y)
if isinstance(updater, torch.optim.Optimizer):
# 使用PyTorch内置的优化器和损失函数
updater.zero_grad()
l.mean().backward()
updater.step()
else:
# 使用定制的优化器和损失函数
l.sum().backward()
updater(X.shape[0])
metric.add(float(l.sum()), accuracy(y_hat, y), y.numel())
# 返回训练损失(所有loss的累加/样本总数)和训练精度(所有分类正确样本数/样本总数)
return metric[0] / metric[2], metric[1] / metric[2]
# 训练
def train_ch3(net, train_iter, test_iter, loss, num_epochs, updater):
"""训练模型(定义见第3章)"""
animator = Animator(xlabel='epoch', xlim=[1, num_epochs], ylim=[0.3, 0.9],
legend=['train loss', 'train acc', 'test acc'])
for epoch in range(num_epochs): # 扫10遍
train_metrics = train_epoch_ch3(net, train_iter, loss, updater) # train_metrics训练误差
test_acc = evaluate_accuracy(net, test_iter) # test_acc在测试数据集上的精度
animator.add(epoch + 1, train_metrics + (test_acc,)) # 可视化
train_loss, train_acc = train_metrics
assert train_loss < 0.5, train_loss
assert train_acc <= 1 and train_acc > 0.7, train_acc
assert test_acc <= 1 and test_acc > 0.7, test_acc
# 预测
def predict_ch3(net, test_iter, n=12):
"""预测标签(定义见第3章)"""
for X, y in test_iter:
break
trues = d2l.get_fashion_mnist_labels(y)
preds = d2l.get_fashion_mnist_labels(net(X).argmax(axis=1))
titles = [true+pred for true, pred in zip(trues, preds)]
d2l.show_images(
X[0:n].reshape((n, 28, 28)), 1, n, titles=titles[0:n])
d2l.plt.show()
if __name__ == '__main__':
# 初始化参数模型
batch_size = 256
num_inputs = 784 # 把28*28的图片拉成一个784的向量
num_outputs = 10 # 因为有10种类型,故输出为10
# 权重将构成一个 784×10 的矩阵W, 偏置将构成一个 1×10 的行向量b
W = torch.normal(0, 0.01, size=(num_inputs, num_outputs), requires_grad=True)
b = torch.zeros(num_outputs, requires_grad=True)
lr = 0.1
num_epochs = 10
train_iter, test_iter = d2l.load_data_fashion_mnist(batch_size) # 加载所有数据,每256个为一组
# print(len(train_iter)) 235
train_ch3(net, train_iter, test_iter, cross_entropy, num_epochs, updater)
d2l.plt.show()
# 预测
predict_ch3(net, test_iter)
09_P5 pycharm代码
import torch
from torch import nn
from d2l import torch as d2l
# 初始化模型参数
def init_weights(m):
if type(m) == nn.Linear:
nn.init.normal_(m.weight, std=0.01) # 将权重初始化为均值为0,方差0.01的随机值
if __name__ == '__main__':
batch_size = 256
num_epochs = 10
train_iter, test_iter = d2l.load_data_fashion_mnist(batch_size)
# PyTorch不会隐式地调整输入的形状。因此,我们在线性层前定义了展平层(flatten),来调整网络输入的形状
# 把任何维度的tensor转成向量(二维) 线性层Linear输入784输出10 Sequential类构造器
net = nn.Sequential(nn.Flatten(), nn.Linear(784, 10))
net.apply(init_weights) # 每一层都跑一下此函数
# 交叉熵损失函数
loss = nn.CrossEntropyLoss(reduction='none')
# 优化算法
trainer = torch.optim.SGD(net.parameters(), lr=0.1)
# 训练
d2l.train_ch3(net, train_iter, test_iter, loss, num_epochs, trainer)
d2l.plt.show()