torch.cosine_similarity 能够对两个向量或者张量计算类似度html
>>> input1 = torch.randn(100, 128)
>>> input2 = torch.randn(100, 128)
>>> output = torch.cosine_similarity(input1, input2, dim=1)
print(output.shape)函数
torch.Size([100])
hg1 = torch.FloatTensor(torch.randn([12]))
hg2 = torch.FloatTensor(torch.randn([12]))
torch.cosine_similarity(hg1, hg2, dim=0)code
tensor(-0.1543)
pytorch计算欧式距离htm
torch.dist(hg1, hg2, p=2)input
若是本身写的话就是：（由于很简单，大多数人本身写），Pytorch里封装了这个距离函数it
torch.sqrt(torch.sum((hg1-hg2)**2))class
Hamming 距离，汉密尔顿距离im
torch.dist(hg1, hg2, p=1)di