How to apply element-wise map function with tensor?

1

I have a dict H，which key is（i，j） and value is a tensor of 4 elements.
And a one-dim tensor e, such as e =[1,2,3]
I want to construct a 3-dim tensor T:
T = torch.tensor(H[(e[i] ,e[j])] for i in range(3) for j in range(3)])
that is to say T_ij = H[(e[i],e[j])].
T will be a tensor with size (3,3,4)
is there any possible way to parallel this process?
Thanks very much!!

2

I don’t quite understand what H is since you are mentioning a dict, but are trying to call it with two indices. Are you using a nested list instead or a tensor?

1 Like
3

Thanks your reply!
H is a python dictionary，which key is a tuple.
Like H[(0,0)] = [1,2,3,4]
H[(0,1)] = [1,2,3,4]