AngelSlayer-12B-Unslop-Mell-RPMax-DARKNESS-v2

They say ‘He’ will bring the apocalypse. She seeks understanding, not destruction.

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

This is my eighth model. This is a v2 revision of the original. Check out the original card for extra info. The goal was to rebalance the parameters a bit. Feedback revealed that the model held the character well, but was boring in output, which I myself have observed. Check out redrix/AngelSlayer-12B-Unslop-Mell-RPMax-DARKNESS-v3, (it's more like a v2b, but I named it v3 for differentiation) which has the parameters tuned higher to be more interesting.

Merge Details

Merge Method

This model was merged using the linear DELLA merge method using TheDrummer/UnslopNemo-12B-v4 as a base.

Models Merged

The following models were included in the merge:

• inflatebot/MN-12B-Mag-Mell-R1
• DavidAU/MN-GRAND-Gutenberg-Lyra4-Lyra-12B-DARKNESS
• ArliAI/Mistral-Nemo-12B-ArliAI-RPMax-v1.2

Configuration

The following YAML configuration was used to produce this model:

models:
  - model: ArliAI/Mistral-Nemo-12B-ArliAI-RPMax-v1.2
    parameters:
      weight:
        - filter: self_attn
          value: 0.3
        - filter: mlp
          value: 0.15
        - value: 0.25
      density: 0.6
  - model: inflatebot/MN-12B-Mag-Mell-R1
    parameters:
      weight:
        - filter: self_attn
          value: 0.15
        - filter: mlp
          value: 0.3
        - value: 0.2
      density: 0.7
  - model: TheDrummer/UnslopNemo-12B-v4
    parameters:
      weight:
        - filter: self_attn
          value: 0.25
        - filter: mlp
          value: 0.15
        - value: 0.25
      density: 0.6
  - model: DavidAU/MN-GRAND-Gutenberg-Lyra4-Lyra-12B-DARKNESS
    parameters:
      weight:
        - filter: self_attn
          value: 0.2
        - filter: mlp
          value: 0.30
        - value: 0.2
      density: 0.5
base_model: TheDrummer/UnslopNemo-12B-v4
merge_method: della_linear
dtype: bfloat16
chat_template: "chatml"
tokenizer_source: union
parameters:
  normalize: true
  int8_mask: true
  epsilon: 0.05
  lambda: 1