看了几天的pytorch代码，准备自己手写一个简单的一元一次方程，还是遇到了各种麻烦，学习就这样，先模仿，再自己写，再创新。
import torch
import torch.nn.functional as F
import torch.nn as nn
from torch.utils.data import TensorDataset
import torch.optim as optim
from torch.utils.data import DataLoader
class Linear(nn.Module):
def __init__(self):
super(Linear, self).__init__()
self.lin = nn.Linear(1,1)
def forward(self, x):
return self.lin(x)
x = torch.randn(256).view(-1,1)
y = 3*x + 1
dataset = TensorDataset(x, y)
dataloader = DataLoader(dataset, shuffle=True, batch_size=32)
epochs = 250
lr = 1e-3
model = Linear()
optimizer = optim.SGD(model.parameters(), lr=lr)
loss_function = nn.MSELoss()
for epoch in range(epochs):
for inputs, targets in dataloader:
predict = model(inputs)
loss = loss_function(predict, targets)
optimizer.zero_grad()
loss.backward()
optimizer.step()
print(list(model.parameters()))
输出
[Parameter containing:
tensor([[2.9407]], requires_grad=True), Parameter containing:
tensor([1.0000], requires_grad=True)]
