Two of the best "Uncensored RolePlay Models" toguether, I'm not responsible of the use of this. The "Base Model Used" was "AngelSlayer at 0.6", "Krix" model was "0.4".

It works, estabilizing the "AngelSlayer model" in a good way.

Models used:

DreadPoor/Krix-12B-Model_Stock for clean, uncensored talk, general‑purpose.
redrix/AngelSlayer-12B-Unslop-Mell-RPMax-DARKNESS-v3 for dark RP.
Quantized Model, Here RicardoEstep/AngelSlayerKrix-12B-GGUF.

YAML Configuration Used:

yaml

models:
  - model: redrix/AngelSlayer-12B-Unslop-Mell-RPMax-DARKNESS-v3
  - model: DreadPoor/Krix-12B-Model_Stock
merge_method: slerp
base_model: redrix/AngelSlayer-12B-Unslop-Mell-RPMax-DARKNESS-v3
dtype: bfloat16
parameters:
  normalize: true
  t: 0.4    # Going only 0.4 at Krix.
tokenizer:
  source: union