清华大学出版社-图书详情-《PyTorch深度学习简明实战》 (tsinghua.edu.cn)
数据集
income1.csv
Unnamed: 0 Education Income
0 1 10.000000 26.658839
1 2 10.401338 27.306435
2 3 10.842809 22.132410
3 4 11.244147 21.169841
4 5 11.645485 15.192634
5 6 12.086957 26.398951
6 7 12.488294 17.435307
7 8 12.889632 25.507885
8 9 13.290970 36.884595
9 10 13.732441 39.666109
10 11 14.133779 34.396281
11 12 14.535117 41.497994
12 13 14.976589 44.981575
13 14 15.377926 47.039595
14 15 15.779264 48.252578
15 16 16.220736 57.034251
16 17 16.622074 51.490919
17 18 17.023411 61.336621
18 19 17.464883 57.581988
19 20 17.866221 68.553714
20 21 18.267559 64.310925
21 22 18.709030 68.959009
22 23 19.110368 74.614639
23 24 19.511706 71.867195
24 25 19.913043 76.098135
25 26 20.354515 75.775218
26 27 20.755853 72.486055
27 28 21.157191 77.355021
28 29 21.598662 72.118790
29 30 22.000000 80.260571
import torch
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
from torch import nn
# print(torch.__version__)
data = pd.read_csv('./income1.csv')
# print(data.head()) # 数据集只有2列，分别是education和income
# print(data.info()) # 都是float类型，可以直接进行运算。如果是字符串类型，则需要进行更改。
# plt.scatter(data.Education,data.Income)
# plt.xlabel('Education')
# plt.ylabel('Income')
# plt.show()
# X = data.Education.values # 取出1维的ndarray数据。 data.Education是Series
# X = data.Education.values.reshape(-1,1) # 第一个维度-1 表示自动计算，第2个维度指定为1； 生成30个维度为1的输入。
# torch.from_numpy(data.Education.values.reshape(-1,1) # torch.from_numpy 从ndarray创建tensor
X = torch.from_numpy(data.Education.values.reshape(-1,1)).type(torch.FloatTensor) #转换为FloatTensor类型
Y = torch.from_numpy(data.Income.values.reshape(-1,1)).type(torch.FloatTensor)
# print(X.shape)
# print(Y.shape)
# pytorch创建模型，必须自己新定义一个类。这个类必须继承自nn.Module
class EIModel(nn.Module): # 创建一个新的类EIModel，这个类必须继承至父类nn.Module
def __init__(self): # 必须重新定义初始化方法。
super(EIModel, self).__init__()
# 我们的权重 f=wx+b，对应pytorch是linear层，初始化w和b，对输入进行加权。
# linear层就是自己初始化w，乘以输入，再加b
self.linear = nn.Linear(in_features=1,out_features=1)
def forward(self,inputs): # 定义forward方法。 定义如何进行前向传播
logits = self.linear(inputs) # 再self.linear层进行调用，返回输入logits
return logits
model = EIModel() # 类的实例化
loss_fn = nn.MSELoss() # 定义均方差损失函数
opt = torch.optim.SGD(model.parameters(),lr=0.0001)
# 通过model.parameters返回模型的可训练参数. lr 学习率。
for epoch in range(5000):
for x,y in zip(X,Y): # zip函数，X与Y同时迭代
y_pred = model(x) # 调用forward方法，得到输出
loss = loss_fn(y_pred, y)
opt.zero_grad() # 1. 将之前的梯度清零
loss.backward() # 2. 损失反向传播
opt.step() # 3.优化器
print('Down!')
print(list(model.named_parameters())) #输出优化后的参数
# print(model.linear.weight, model.linear.bias)
plt.scatter(data.Education, data.Income, label='real data') #绘制原始散点图
plt.plot(X,model(X).detach().numpy(),c='r',label='predicted line')
plt.xlabel('Education')
plt.ylabel('Income')
plt.legend()
plt.show()
image.png
import torch
import pandas as pd
data = pd.read_csv('./income1.csv')
X = torch.from_numpy(data.Education.values.reshape(-1,1)).type(torch.FloatTensor)
Y = torch.from_numpy(data.Income.values.reshape(-1,1)).type(torch.FloatTensor)
w = torch.randn(1, requires_grad=True)
b = torch.zeros(1, requires_grad=True)
# 模型公式 y = w@x + b
learning_rate=0.0001
for epoch in range(5000):
for x,y in zip(X,Y):
y_pred = torch.matmul(x, w) + b # 1. 使用模型进行预测
loss = (y - y_pred).pow(2).mean() # 2. 根据预算结果计算损失
if not w.grad is None: # 3. 将梯度清零
w.grad.data.zero_()
if not b.grad is None:
b.grad.data.zero_()
loss.backward() # 4. 求梯度
with torch.no_grad():
w.data -= w.grad.data*learning_rate # 5. 优化模型参数
b.data -= b.grad.data*learning_rate
print(w)
print(b)
tensor([4.9734], requires_grad=True)
tensor([-28.3475], requires_grad=True)
网友评论