Protocol-Phantom-12B

[!CAUTION] <span style="color:red; font-weight:bold">⚠️ Warning:</span> This model can produce narratives and RP that contain violent and graphic erotic content. Adjust your system prompt accordingly, and use ChatML chat template.

🧞‍ Protocol Phantom 12B

This is a merge of pre-trained language models created using mergekit.

This model was merged using the nuslerp method using mistralai/Mistral-Nemo-Instruct-2407 as a base.

Protocol Phantom has some refusals and may require jailbreaks or ablation to fully uncensor.

Several methods were tested including karcher and della, but they produced broken output. Models such as Impish Bloodmoon, SakuraKaze, Irix, VelvetCafe, Arsenic Shahrazad, and others had to be omitted from the merge.

Configuration

The following YAML configuration was used to produce this model:

architecture: MistralForCausalLM
base_model: B:/12B/mistralai--Mistral-Nemo-Instruct-2407
models:
  - model: B:/12B/LatitudeGames--Wayfarer-2-12B
    parameters:
      weight: 0.5
  - model: B:/12B/WokeAI--Tankie-DPE-12b-SFT-v2
    parameters:
      weight: 0.5
merge_method: nuslerp
parameters:
  nuslerp_flatten: false     # Flattens tensors to treat them as high-dimensional vectors
  nuslerp_row_wise: true   # Set to true if you want to interpolate per-row instead of per-tensor
dtype: float32
out_dtype: bfloat16
tokenizer:
  source: union