def train():
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
batch_size = 16
# 如出现“out of memory”的报错信息，可减小batch_size或resize
train_iter, test_iter = d2l.load_data_fashion_mnist(
batch_size, root='Data/softmax/FashionMNIST2065')
lr, num_epochs = 0.001, 5
optimizer = torch.optim.Adam(net.parameters(), lr=lr)
d2l.train_ch5(net, train_iter, test_iter, batch_size, optimizer, device,
num_epochs)
def main():
# 5.5.1 模型
# x = torch.rand(1, 1, 28, 28)
# print(x.shape)
net = LeNet()
# print(net(x))
# 5.5.2 获取数据和训练模型
batch_size = 256
train_iter, test_iter = d2l.load_data_from_fashion_mnist(batch_size)
lr, num_epochs = 0.001, 20
optimizer = torch.optim.Adam(net.parameters(), lr=lr)
d2l.train_ch5(net, train_iter, test_iter, batch_size, optimizer, device,
num_epochs)
def main():
# 5.6.2 模型
net = AlexNet()
print(net)
# 5.6.3 读取数据
batch_size = 128
train_iter, test_iter = d2l.load_data_from_fashion_mnist(batch_size,
resize=224)
# 5.6.4 训练
lr, num_epochs = 0.001, 5
optimizer = torch.optim.Adam(net.parameters(), lr=lr)
d2l.train_ch5(net, train_iter, test_iter, batch_size, optimizer, device,
num_epochs)
mnist_test = torchvision.datasets.FashionMNIST(root=root, train=False, download=True, transform=transform)
train_iter = torch.utils.data.DataLoader(mnist_train, batch_size=batch_size, shuffle=True, num_workers=4)
test_iter = torch.utils.data.DataLoader(mnist_test, batch_size=batch_size, shuffle=False, num_workers=4)
return train_iter, test_iter
batch_size = 128
# 如出现“out of memory”的报错信息，可减小batch_size或resize
train_iter, test_iter = load_data_fashion_mnist(batch_size, resize=224)
lr, num_epochs = 0.001, 5
optimizer = torch.optim.Adam(net.parameters(), lr=lr)
d2l.train_ch5(net, train_iter, test_iter, batch_size, optimizer, device, num_epochs)
# output_size 为输出的步长，即未来几个时刻的数据
self.fc = nn.Linear(args.hidden_size, args.output_size)
def forward(self, x):
# fixme
# x = x.permute(0, 2, 1) # 转置
x = self.conv(
x) # input_shape=(width, height, channel) channel即是通道数，数列的个数
x = x.permute(0, 2, 1)
x, _ = self.lstm(x)
x = self.fc(x)
x = x[:, -1, :]
return x
if __name__ == '__main__':
lr, dropout, num_epochs = 0.001, 0.75, 100
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
loss = torch.nn.MSELoss()
train_iter, test_iter = loan_data()
net = CNN_LSTM(Arg(1, 18, 3, 4, 1, dropout))
optimizer = torch.optim.Adam(net.parameters(), lr=lr, weight_decay=1e-8)
d2l.train_ch5(net, train_iter, test_iter, 0, optimizer, device, num_epochs,
loss)
# net.args={"in_channels":7,"out_channels":6,"hidden_size":3,"num_layers":5}
# net.args.in_channels=7
# print(net.args.num_layers)
