Klein 9B KV Cache FP8

FLUX.2 [klein] 9B KV is a variant of the klein 9B model optimized for fast image editing. It uses KV caching to avoid redundantly recomputing attention over reference image tokens at every denoising step, providing significant speedups when working with reference images.

Please note that this model is the property of Black Forests. It is mainly included to demonstrate the advantages of using KV cache, which speeds up image generation, as described in the paper: