keenanpepper/Kimi-K2.6-FP8-fused
Same weights, repacked vs RedHatAI/Kimi-K2.6-FP8-BLOCK. Same weights in different files (re-saved, re-sharded or new metadata), plus 606 new tensors.
Tensors changed: Norms 0/357, Attention 0/359, MLP 0/411, lm_head 0/1, Embeddings 0/4, other 0/58