在PyTorch中有一个双线性层时,我无法为计算的完成而烦恼。
这是一个小例子,我试图弄清楚它是如何工作的:
在:
import torch.nn as nn
B = nn.Bilinear(2, 2, 1)
print(B.weight)
退出:
Parameter containing:
tensor([[[-0.4394, -0.4920],
[ 0.6137, 0.4174]]], requires_grad=True)
我正在经历一个零向量和一个向量。
在:
print(B(torch.ones(2), torch.zeros(2)))
print(B(torch.zeros(2), torch.ones(2)))
退出:
tensor([0.2175], grad_fn=<ThAddBackward>)
tensor([0.2175], grad_fn=<ThAddBackward>)
我尝试过各种方式加总权重,但结果却不一样。
谢谢!
答案 0 :(得分:8)
nn.Bilinear
完成的操作是B(x1, x2) = x1*A*x2 + b
(参见doc),
A
存储在nn.Bilinear.weight
b
存储在nn.Bilinear.bias
如果考虑(可选)偏见,则应获得预期的结果。
import torch
import torch.nn as nn
def manual_bilinear(x1, x2, A, b):
return torch.mm(x1, torch.mm(A, x2)) + b
x_ones = torch.ones(2)
x_zeros = torch.zeros(2)
# ---------------------------
# With Bias:
B = nn.Bilinear(2, 2, 1)
A = B.weight
print(B.bias)
# > tensor([-0.6748], requires_grad=True)
b = B.bias
print(B(x_ones, x_zeros))
# > tensor([-0.6748], grad_fn=<ThAddBackward>)
print(manual_bilinear(x_ones.view(1, 2), x_zeros.view(2, 1), A.squeeze(), b))
# > tensor([[-0.6748]], grad_fn=<ThAddBackward>)
print(B(x_ones, x_ones))
# > tensor([-1.7684], grad_fn=<ThAddBackward>)
print(manual_bilinear(x_ones.view(1, 2), x_ones.view(2, 1), A.squeeze(), b))
# > tensor([[-1.7684]], grad_fn=<ThAddBackward>)
# ---------------------------
# Without Bias:
B = nn.Bilinear(2, 2, 1, bias=False)
A = B.weight
print(B.bias)
# None
b = torch.zeros(1)
print(B(x_ones, x_zeros))
# > tensor([0.], grad_fn=<ThAddBackward>)
print(manual_bilinear(x_ones.view(1, 2), x_zeros.view(2, 1), A.squeeze(), b))
# > tensor([0.], grad_fn=<ThAddBackward>)
print(B(x_ones, x_ones))
# > tensor([-0.7897], grad_fn=<ThAddBackward>)
print(manual_bilinear(x_ones.view(1, 2), x_ones.view(2, 1), A.squeeze(), b))
# > tensor([[-0.7897]], grad_fn=<ThAddBackward>)