pytorch-优化器比较

最新推荐文章于 2023-01-16 08:08:31 发布

马凡乐

最新推荐文章于 2023-01-16 08:08:31 发布

阅读量183

点赞数

分类专栏：深度学习 pytorch 文章标签： pytorch 深度学习 python

本文链接：https://blog.csdn.net/qq_43960768/article/details/125594352

版权

深度学习同时被 2 个专栏收录

8 篇文章 0 订阅

订阅专栏

pytorch

5 篇文章 0 订阅

订阅专栏

import torch
import torch.utils.data as data
import torch.nn as nn
import torch.optim as optim
import torch.nn.functional as F
import matplotlib.pyplot as plt

lr = 1e-2
batch_size = 32
epochs = 12
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
# 生成训练数据
x = torch.unsqueeze(torch.linspace(-1, 1, 1000), dim=1)
y = torch.pow(x, 2) + 0.1 * torch.normal(torch.zeros_like(x))
# print(x.shape)
# print(y.shape)

dataset = data.TensorDataset(x, y)

loader = data.DataLoader(dataset=dataset, batch_size=batch_size, shuffle=True)


# 构建神经网络
class Net(nn.Module):
    def __init__(self):
        super(Net, self).__init__()
        self.hidden = nn.Linear(1, 20)
        self.predict = nn.Linear(20, 1)

    def forward(self, x):
        x = F.relu(self.hidden(x))
        x = self.predict(x)
        return x


# 使用多种优化器
net_SGD = Net()
net_Momentum = Net()
net_RMSProp = Net()
net_Adam = Net()

nets = [net_SGD, net_Momentum, net_RMSProp, net_Adam]

opt_SGD = optim.SGD(net_SGD.parameters(), lr=lr)
opt_Momentum = optim.SGD(net_SGD.parameters(), lr=lr, momentum=0.9)
opt_RMSProp = optim.RMSprop(net_SGD.parameters(), lr=lr, alpha=0.9)
opt_Adam = optim.Adam(net_SGD.parameters(), lr=lr, betas=(0.9, 0.99))

optimizers = [opt_SGD, opt_Momentum, opt_RMSProp, opt_Adam]

# 训练模型
loss_func = nn.MSELoss()
loss_his = [[], [], [], []]
for epoch in range(epochs):
    for step, (batch_x, batch_y) in enumerate(loader):
        for net, opt, l_his in zip(nets, optimizers, loss_his):
        	# if epoch % 3 == 0:
            # 	opt.param_groups[0]['lr'] *= 0.99
            batch_x = batch_x.to(device)
            batch_y = batch_y.to(device)
            out = net(batch_x)
            loss = loss_func(out, batch_y)
            l_his.append(loss.data.numpy())
            opt.zero_grad()
            loss.backward()
            opt.step()

# 可视化结果
labels = ['SGD', 'Momentum', 'RMSProp', 'Adam']
for i, l_his in enumerate(loss_his):
    plt.plot(l_his, label=labels[i])
plt.legend(loc='best')
plt.xlabel('Steps')
plt.ylabel('Loss')
plt.ylim((0, 1))
plt.show()