Current torch.min() does not support multiple dimensions?

2

Hi @Kevin96,

You should do:

a.view(-1,2*2).min(axis=1)

where view guarantees no memory copy, and almost no overhead.

5 Likes