This is one of the more amazing papers I've seen

#1
by atrix - opened

What sort of equipment was used to train the hypernetworks on the models from the examples? How long did it take?

Sakana AI org

It took 5 days on 8 H100 GPUs to train each hypernetwork. You can find the training details in the appendix B of the paper.

Rujikorn changed discussion status to closed

Sign up or log in to comment