DavidAU/Qwen3-The-Josiefied-Omega-Directive-22B-uncensored-abliterated
Partial fine-tune (fewer layers) vs Goekdeniz-Guelmez/Josiefied-Qwen3-14B-abliterated-v3. Goekdeniz-Guelmez/Josiefied-Qwen3-14B-abliterated-v3 matches part of DavidAU/Qwen3-The-Josiefied-Omega-Directive-22B-uncensored-abliterated (443 of its 685 tensors; the rest of DavidAU/Qwen3-The-Josiefied-Omega-Directive-22B-uncensored-abliterated has no counterpart). On the shared part: Partial change: 59% of the weights changed (attention, lm_head, mlp, norms), in 24 of 40 layers.
Tensors changed: Embeddings 0/1, Norms 96/161, MLP 72/120, Attention 96/160, lm_head 1/1