XReyRobert/Qwen3.8-27B-W4A16-AutoRound-INT8-Head-Embeddings
Same weights, repacked vs dbirks/Qwen3.8-27B-W4A16-AutoRound. Same weights in different files (re-saved, re-sharded or new metadata), plus 6 new tensors.
Tensors changed: Attention 0/976, MLP 0/691, Norms 0/325, Embeddings 0/4, other 0/1