Show HN: We compressed a 31B video+audio diffusion model to 14B, 6 steps, INT8github.com
29 points by jiant 5 hours ago | 9 comments