import torch

def EuclideanDistance(t1,t2):
    dim=len(t1.size())
    if dim==2:
        N,C=t1.size()
        M,_=t2.size()
        dist = -2 * torch.matmul(t1, t2.permute(1, 0))
        dist += torch.sum(t1 ** 2, -1).view(N, 1)
        dist += torch.sum(t2 ** 2, -1).view(1, M)
        dist=torch.sqrt(dist)
        return dist
    elif dim==3:
        B,N,_=t1.size()
        _,M,_=t2.size()
        dist = -2 * torch.matmul(t1, t2.permute(0, 2, 1))
        dist += torch.sum(t1 ** 2, -1).view(B, N, 1)
        dist += torch.sum(t2 ** 2, -1).view(B, 1, M)
        dist=torch.sqrt(dist)
        return dist
    else:
        print('error...')

print(f'dimensional 2.......')
a=torch.Tensor([[0,0],[1,1]])
b=torch.Tensor([[1,0],[3,4]])
print(f'size of a:{a.size()}tsize of b:{b.size()}')
print(f'distance of point a and b is: {EuclideanDistance(a,b)}')
print(f'ndimensional 3.......')
a=torch.unsqueeze(a,dim=0)
b=torch.unsqueeze(b,dim=0)
print(f'size of a:{a.size()}tsize of b:{b.size()}')
print(f'distance of point a and b is: {EuclideanDistance(a,b)}')

d1=-2 * torch.matmul(a, b.permute(0, 2, 1))

d2=torch.sum(a** 2, -1)
d3=torch.sum(b** 2, -1)

d=d1+d2.view(B,N,1)+d3.view(B,1,M)

d=torch.sqrt(d)