Zynerji/Ektome-Qwen3-28B-CompressedUncensored
Partial fine-tune vs Qwen/Qwen3-32B. Partial change: 81% of the weights changed (attention, mlp, norms), in 54 of 54 layers.
Tensors changed: Embeddings 0/1, Norms 168/217, MLP 138/162, Attention 180/216, lm_head 0/1