How could I set value of an array to be its index?

1

Hi,

I would like to do something like this:

a = torch.randint(0, 5, (6,))
inds = torch.randint(0, 7, (5,))
for i, ind in enumerate(inds):
    a[a==ind] = i

The for loop method is likely to be slow, is there better method to do this please ?

2

In your loop you are assigning the “last” i value to a in case inds contains duplicated values and are also overwriting already replaced values.
E.g. assume that inds contains [1, 0]: if a contains any 1s, they would be replaces with a 0 and in the next iteration all 0s would be replaced with 1s again.
Given that, I think a “vectorized” approach might be tricky, as it would not consider this sequential way.
Nevertheless, here is a potentially faster approach, which will not yield the same results, if values are not unique:

input = torch.randint(0, 5, (6,))
a = input.clone()
b = input.clone()
inds = torch.randint(0, 7, (5,))
for i, ind in enumerate(inds):
    a[a==ind] = i
    
idx = (b.unsqueeze(0) == inds.unsqueeze(1)).nonzero()
b[idx[:, 1]] = idx[:, 0]
3

Hi,

It is my fault that I did not describe correctly. What I would like to do is like this:
For example, the inputs are:

ind = torch.tensor([3,5,7,8])
a = torch.tensor([5,5,5,8,8,3,7,3,7])

Here values in ind is unique, but values in a is not unique.
And the expected output is like this:

out = torch.tensor([1,1,1,3,3,0,2,0,2])

I did not mean to do this in sequential way, my implementation above is wrong. Do I have a good way to do this please?

4

Yes, in that case my approach would work and you can compare the results as:

inds = torch.tensor([3,5,7,8])
input = torch.tensor([5,5,5,8,8,3,7,3,7])
a = input.clone()
b = input.clone()
for i, ind in enumerate(inds):
    a[a==ind] = i
    
idx = (b.unsqueeze(0) == inds.unsqueeze(1)).nonzero()
b[idx[:, 1]] = idx[:, 0]

(a == b).all()
> tensor(True)