Pytorch之nn.linear层

Pytorch之nn.linear层

1.二维：

注意：改变的是最后一维列数

X = torch.rand(2, 4)
net1 = nn.Sequential(nn.Linear(4, 3))
d = net1(X)
print(d)

输出：
tensor([[-0.4811, -0.2753,  0.4857],
        [-0.2339, -0.1274,  0.4677]], grad_fn=<AddmmBackward>)

---

1.三维：

注意：改变的是最后一维列数

X = torch.rand(3, 2, 4)

输出：
tensor([[[0.1390, 0.2656, 0.5775, 0.0419],
         [0.1601, 0.2080, 0.6637, 0.9871]],

        [[0.3458, 0.6473, 0.2802, 0.8923],
         [0.1926, 0.1190, 0.8961, 0.2184]],

        [[0.5982, 0.2578, 0.3267, 0.9850],
         [0.8681, 0.9482, 0.6696, 0.1949]]])
         
net1 = nn.Sequential(nn.Linear(4, 3))
d= net1(X)
d  
输出：
tensor([[[ 0.0062,  0.3436,  0.0684],
         [-0.0757,  0.4480,  0.4901]],

        [[-0.0655,  0.6235,  0.5633],
         [-0.0134,  0.3415,  0.0643]],

        [[ 0.0011,  0.5260,  0.6617],
         [-0.0137,  0.7969,  0.1548]]], grad_fn=<AddBackward0>)

---