14模型构造
import torch
from torch import nn
from torch.nn import functional as F
net1 = nn.Sequential(nn.Linear(20, 256), nn.ReLU(), nn.Linear(256,10))
"""
nn.Sequential定义了一种特殊的Module, 即在PyTorch中表示一个块的类,
它维护了一个由Module组成的有序列表。
注意,两个全连接层都是Linear类的实例, Linear类本身就是Module的子类。
另外,到目前为止,我们一直在通过net(X)调用我们的模型来获得模型的输出。
这实际上是net.__call__(X)的简写。这个前向传播函数非常简单: 它将列表中的每个块连接在一起,将每个块的输出作为下一个块的输入。
"""
X1 = torch.rand(2,20)
print(net1(X1))
class MLP(nn.Module):def __init__(self):super().__init__()self.hidden = nn.Linear(20, 256) self.out = nn.Linear(256, 10) def forward(self, X):return self.out(F.relu(self.hidden(X)))
X2 = torch.rand(2,20)
net2 = MLP()
print(net2(X2))
class MySequential(nn.Module):def __init__(self, *args):super().__init__()for idx, module in enumerate(args):self._modules[str(idx)] = moduledef forward(self, X):for block in self._modules.values():X = block(X)return XX3 = torch.rand(2,20)
net3 = MySequential(nn.Linear(20, 256), nn.ReLU(), nn.Linear(256, 10))
print(net3(X3))
class FixedHiddenMLP(nn.Module):def __init__(self):super().__init__()self.rand_weight = torch.rand((20,20), requires_grad=False)self.linear = nn.Linear(20, 20)def forward(self, X):X = self.linear(X)X = F.relu(torch.mm(X, self.rand_weight) + 1)X = self.linear(X)while X.abs().sum() > 1:X = X / 2return X.sum()X4 = torch.rand(2,20)
net4 = FixedHiddenMLP()
print(net4(X4))
class NestMLP(nn.Module):def __init__(self):super().__init__()self.net = nn.Sequential(nn.Linear(20, 64), nn.ReLU(),nn.Linear(64, 32), nn.ReLU())self.linear = nn.Linear(32, 16)def forward(self, X):return self.linear(self.net(X))X5 = torch.rand(2,20)
net5 = nn.Sequential(NestMLP(), nn.Linear(16, 20), FixedHiddenMLP())
print(net5(X5))"""
tensor([[ 0.0843, -0.1867, 0.0457, 0.1082, -0.0236, -0.1245, -0.0184, 0.0233,0.1765, -0.1390],[ 0.0129, -0.1441, 0.1156, -0.0327, 0.0044, -0.0510, 0.0715, -0.0162, 0.0137, -0.1148]], grad_fn=<AddmmBackward>)
tensor([[-0.1180, 0.0799, -0.0260, 0.0529, 0.0055, -0.1481, 0.1311, -0.1334, 0.1224, 0.0713],[-0.0610, 0.0789, -0.0321, 0.0154, 0.0246, -0.1857, 0.0652, -0.0461, 0.1029, 0.1205]], grad_fn=<AddmmBackward>)
tensor([[-0.0571, -0.1119, 0.0851, 0.1362, -0.0945, 0.0078, 0.2157, -0.1273, -0.0017, 0.1981],[-0.0049, -0.0103, 0.0114, -0.0101, -0.1034, 0.0204, 0.1531, 0.0481, 0.1361, -0.0403]], grad_fn=<AddmmBackward>)
tensor(0.3121, grad_fn=<SumBackward0>)
tensor(0.1369, grad_fn=<SumBackward0>)
"""