vcruz305/Qwen3.8-27B-DFlash2-SAGE-EXL3-5.0bpw
Same weights, repacked vs z-lab/Qwen3.8-27B-DFlash2. Same weights in different files (re-saved, re-sharded or new metadata), plus 148 new tensors.
Tensors changed: other 0/2, Norms 0/22, Attention 0/10, MLP 0/10