alt Hacker NewsApparently op cares a lot about speed, which is fine, but ML researchers care about correctness first, speed second. And it makes sense, because they are not as resource constrained as OP.
Replies
nicman23 • today at 6:32 AM
yeah that is why they compute in fp32 lol
Most PyTorch tensor operations are cython not python. So imo rewriting in rust is not going to have an enormous speed up. If that really was the concern we should see a throughput comparison vs PyTorch or something.