Read news on INT8 with our app.
Read more in the app
Show HN: We compressed a 31B video+audio diffusion model to 14B, 6 steps, INT8