@metaend.eth

Unfortunately I haven't been able to get better results (worse actually) using Klearu's sparse (patched) with Qwen3.5 0.8b - IMO the model is too small to make a diff

0 reply

0 recast

0 reaction