lodestone-rock(@LodestoneRock):you don't need to pay too much attention! full attn is overkill and should be used sparingly calibrated krea2 from full attn on 48 heads into 12 heads full attn while the rest is scanline attn (band diagonal attn). scanline attn is block sparse by design (see the gif)

lodestone-rock

@LodestoneRock

Researching pixels @freepik. Independently created the Chroma & Radiance models as personal projects.

Joined April 2023

222 Following 1.3K Followers

lodestone-rock@LodestoneRock

2026.09.17 15:41

you don't need to pay too much attention! full attn is overkill and should be used sparingly calibrated krea2 from full attn on 48 heads into 12 heads full attn while the rest is scanline attn (band diagonal attn). scanline attn is block sparse by design (see the gif)