How efficient is torch.cat? I am assuming that it reallocates everything that is passed as an argument. Is that true ?
1 Like
yes, it pre-allocates the full tensor and then copy into it each element.
1 Like