CODEWITHSUNDEEP
pythonpytorch
Saran Pannasuriyaporn
Saran Pannasuriyaporn

Reputation: 21

count how many tensor in each tensor between 2 tensors are exactly equal

Since I have learned metric 0/1 loss for multi-label classification, and I need to create that metric

so I would like to def check that could count and return how many inside tensor between two tensors are actually matched

For Example

# Only second inside tensor are exactly equal

a = torch.tensor([[1,0,1,1], [1,0,1,1],[1,0,1,1]])
b = torch.tensor([[0,0,1,1], [1,0,1,1],[1,0,1,0]])
check(a,b) -> 1

AND

# Only second and third are exactly equal
c = torch.tensor([[1,0,1,1], [1,0,1,1],[1,0,1,1]])
d = torch.tensor([[0,0,1,1], [1,0,1,1],[1,0,1,1]])

check(c,d) -> 2

My question is that how could we define function check to retrieve 2 tensors and return the following expected output, Thank you in advanced

Upvotes: 0

Views: 278

Answers (1)

Saran Pannasuriyaporn
Saran Pannasuriyaporn

Reputation: 21

I'm answering my question, Thanks to the combination of @Zwang
and @LudvigH advice,

There are many ways you can do this, subtracting one tensor from the other then summing values and finding the non zero tensors, to simply using python's built in == function. What have you tried? - Zwang

I tried one of his methods, and that's answering my question :D, and fixed bug by @Ludvigs

def check(a,b):

    assert a.size() == b.size()      
    diff_tensor =  a - b                  # [1,0,0,0],[0,0,0,0],[0,0,0,0]
    diff_tensor = torch.abs(diff_tensor)  # In case summation of them are neglect to zero, Credit goes to @LudvigH
    diff_scalar = torch.sum(diff_tensor, dim= 1) # [1 , 0 , 0]
      
    return torch.numel(diff_scalar[diff_scalar==0])

Upvotes: 1

Related Questions