Re:Stability AI Launches the First of Its StableLM Suite of ...
by lhl on Wednesday April 19, 2023 @05:36PM
(#35633619)
https://arxiv.org/abs/2205.14135 - Section 5 suggests that the biggest limitation is that custom CUDA kernels need to be coded on a per-GPU architecture basis.
[ Parent
| on HN ]
There are no comments on this comment yet.