🛠️ We compressed a 31B video+audio diffusion model to 14B, 6 steps, INT8
Summary
Hyperflow-H3-Flash is a compressed video and audio diffusion model that reduces a 31B parameter model down to 14B parameters using 6 steps and INT8 quantization.
Why it’s interesting
It achieves significant model compression and efficiency for a large video and audio diffusion model.
Source metrics: Points 2 · Comments 0
HN discussion · Project
Source: #HackerNews / Show HN
Summary
Hyperflow-H3-Flash is a compressed video and audio diffusion model that reduces a 31B parameter model down to 14B parameters using 6 steps and INT8 quantization.
Why it’s interesting
It achieves significant model compression and efficiency for a large video and audio diffusion model.
Source metrics: Points 2 · Comments 0
HN discussion · Project
Source: #HackerNews / Show HN