Dot product

高维度计算

from torch import nn
class DotProductSimilarity(nn.Module):
    def __init__(self,scale_output=True):
        super(DotProductSimilarity,self).__init__()
        self.scale_output=scale_output
    def forward(self,tensor_1,tensor_2):
        result=(tensor_1*tensor_2).sum(dim=-1)
        if (self.scale_output):
            result /= math.sqrt(tensor_1.size())
        return  result
input_1=torch.ones(2,1,720,1280,requires_grad=True)
input_2=torch.ones(2,1,720,1280,requires_grad=True)
input_3=torch.ones(2,1,3,5,requires_grad=True)
con= DotProductSimilarity()
CS=con(input_1,input_2)
print(CS.shape)
CS = CS.unsqueeze(3)
s = CS*input_1
# print(torch.exp(s))
print(input_1.shape,input_2.shape,CS.shape,s.shape)
print(s)