L3.3-70B-Animus-Loki-Hybrid-EXL3

Send me your support to help me feed the data beast!

Support on Ko-fi

Model Description

Hybrid is a high-performance hybrid merge designed to create the ultimate Wings of Fire roleplaying engine. This model utilizes the Della merge method to combine the deep, specialized knowledge of the Animus series with the evocative, flowing prose of Loki.

The primary goal of this hybrid was to:

• Preserve Animus Lore & Knowledge: Retain the strict adherence to Pyrrhian history, tribal politics, and character dynamics established in Animus V12.4 Redux.
• Maintain Formatting: Keep the clean, structured roleplay formatting (Inner thoughts, specific dialogue tags) that Animus is known for.
• Infuse Loki's Prose: Enhance narrative descriptions, sensory details, and emotional depth by blending in the stylistic strengths of Loki V2.0.

Quantized Models

Quantized versions of the hybrid are available below:

Download GGUF Files →

How to Download

You can download specific model quantizations using the Hugging Face Command Line Interface (CLI). This allows you to select the exact version you need.

1. Install huggingface-hub with CLI support:

pip install -U "huggingface_hub[cli]"

2. Download a specific quant:

Use the command below, replacing the revision with the desired model version from the repository's branches.

huggingface-cli download Darkhn-Quants-3/L3.3-Animus-Loki-Hybrid-EXL3 --revision "6.0bpw_H6" --local-dir ./L3.3-Animus-Loki-Hybrid-EXL3

Character Card & Lore Book

For the best roleplaying experience, it is highly recommended to use the provided character card and lore book. These files help guide the model's persona and provide rich, in-universe context.

Download Files →

Sampler Presets

Optimal balance between creativity and narrative coherence for this hybrid merge:

Download SillyTavern Presets →

Manual Sampler Settings:

• Temp: 0.8 - 1.2
• Min P: 0.02
• Dry: 0.8, 1.75, 4

Creative Alternative:

• Temp: 1.0
• Min P: 0.03
• Nsigma: 2.0
• Dry: 0.8, 1.75, 4

Merge Recipe

merge_method: della
base_model: kldzj/Llama-3.3-70B-Instruct-heretic
dtype: float16
out_dtype: bfloat16
tokenizer_source: L3.3-70B-Animus-V12.4-Redux
parameters:
  normalize: false
  int8_mask: true
models:
  - model: L3.3-70B-Animus-V12.4-Redux
    parameters:
      density: 1.0
      weight:
        - filter: self_attn
          value: [1.0, 0.7, 0.7, 1.0]
        - filter: mlp
          value: 1.0 
        - value: 1.0
  - model: L3.3-70B-Loki-V2.0
    parameters:
      density: 0.8
      epsilon: 0.1
      weight:
        - filter: self_attn
          value: [0.1, 0.55, 0.55, 0.1]
        - filter: mlp
          value: [0.1, 0.3, 0.3, 0.1]
        - value: 0.0
model_name: L3.3-70B-Animus-Loki-Hybrid-V16
chat_template: llama3
seed: 42

Chat Template

This model uses the Llama 3.3 instruction template. Ensure your client is configured correctly.

<|begin_of_text|><|start_header_id|>system<|end_header_id|>

system prompt<|eot_id|><|start_header_id|>user<|end_header_id|>

user message<|eot_id|><|start_header_id|>assistant<|end_header_id|>

assistant response<|eot_id|>

Roleplay Format Guide

The model favors a clean, structured style to distinguish between action and thought:

Actions / Descriptions

*He walked across the room and stared out the window.*

Inner Thoughts

*-I wonder what she's thinking.-*

Dialogue

Alex (Curious): "What do you see out there?"

Intended Use & Limitations

• Primary Use: Creative roleplaying within the Wings of Fire universe.
• Versatility: Despite the WoF tuning, users report it is highly effective for general-purpose roleplay and OC settings.
• Safety: This is an uncensored model. Content generation is governed by the user's input.

Acknowledgements

• Credit to kldzj for the Heretic base.
• Credit to CrucibleLab for Loki V2.0.
• Credit to Darkhn for the Animus V12.4 Redux lore backbone.