README.md

---
base_model:
- kubernetes-bad/chargen-v2
- NeverSleep/Noromaid-7B-0.4-DPO
- Weyaxi/Einstein-v4-7B
- NousResearch/Hermes-2-Pro-Mistral-7B
library_name: transformers
tags:
- mergekit
- merge
- character_creation
- world-building
- lore_writing
- not-for-all-audiences
- nsfw
language:
- en
pipeline_tag: text-generation
---
# WorldGen-v1.1

WorldGen is a merge designed to enhance the capabilities of the character creation model, CharGen. While CharGen is great for creating characters, even to this day given its age, but it struggles at effectively making Lorebook/WorldInfo entries.</br>
However, by merging it with an RP LLM and two LLMs that contain some scientific data and are good at instruction following I was able to make it more reliable at world-building.

## Prompting Guide
### Formatting
ChatML
### Roles Usage

**User**: Prompts with this role are used to give the AI writing tasks.
<details><summary><b>Examples</b></summary>
  
Lore writing:
```txt
Write an overview of cat demi-humans tongues and how they can't taste sweet things due to their cat biology.
```

Persona writing:</br>
(Do not use a name. Because CharGen is trained on 'User' for personas it's important to use such when creating a persona.)
```txt
Here is a brief overview of a character. Expand it into a detailed description. Include details about character's personality, their outfit and figure. Mention their age and gender, if applicable.
User is a 21-year-old human male. He's very kind and caring especially to his adopted sister, Luna, whom he tends to spoil.
```

Backstory writing:
```txt
Write a backstory on the crescent moon choker User gifted to Luna.
```
</details>


**System**: Prompts for output guidance and feeding important information not in the context.(And for the system prompt, obviously.)
<details><summary><b>Examples</b></summary>

 Guidance prompt:
```txt
When writing the backstory, use the following information:

- It was gifted to Luna on her 14th birthday.
- User spent all the money he got from his part-time job at the time just so he could give her the special gift.
- Ever since then, she kept wearing it ever since.

Give the backstory a header.
Keep the backstory direct and concise with less than 50 words.
```

System prompt:
```txt
You are an expert in creating interesting roleplay characters and lore for the world they reside in.
```

</details>

**Assistant**: The AI's output

<details><summary><b>Examples</b></summary>

```txt
### Cat Demi-Human Tongues:
Much like their fully feline counterparts, cat demi-humans possess a specialized tongue structure that lacks the taste buds responsible for detecting sweet flavors. Their tongues have a higher concentration of taste buds for salty, sour, and umami flavors, which aligns with their natural dietary preferences.
```

</details>

## Merge Details
### Merge Method

This model was merged using the [TIES](https://arxiv.org/abs/2306.01708) merge method using [kubernetes-bad/chargen-v2](https://huggingface.co/kubernetes-bad/chargen-v2) as a base, created using [mergekit](https://github.com/cg123/mergekit).

### Models Merged

<details><summary>The following models were included in the merge:</summary>
<b>Merge Tree:</b>

[kubernetes-bad/chargen-v2](https://huggingface.co/kubernetes-bad/chargen-v2)</br>
&nbsp;&nbsp;&nbsp;&nbsp;⌊ [NeverSleep/Noromaid-7B-0.4-DPO](https://huggingface.co/NeverSleep/Noromaid-7B-0.4-DPO)</br>
&nbsp;&nbsp;&nbsp;&nbsp;⌊ [Weyaxi/Einstein-v4-7B](https://huggingface.co/Weyaxi/Einstein-v4-7B)</br>
&nbsp;&nbsp;&nbsp;&nbsp;⌊ [NousResearch/Hermes-2-Pro-Mistral-7B](https://huggingface.co/NousResearch/Hermes-2-Pro-Mistral-7B)

</details>

### Configuration

<details><summary>The following YAML configuration was used to produce this model:</summary>
<b>Merge Recipe:</b>

```yaml
models:
  - model: kubernetes-bad/chargen-v2
    parameters:
      density: 1
      weight: 1
  - model: NeverSleep/Noromaid-7B-0.4-DPO
    parameters:
      density: 0.5
      weight: 0.2
  - model: Weyaxi/Einstein-v4-7B
    parameters:
      density: [0.6, 0.5, 0.4, 0.5, 0.6]
      weight: 0.15
  - model: NousResearch/Hermes-2-Pro-Mistral-7B
    parameters:
      density: [0.4, 0.5, 0.6, 0.5, 0.4]
      weight: 0.15
merge_method: ties
base_model: kubernetes-bad/chargen-v2
parameters:
  normalize: true
dtype: float32
out_dtype: bfloat16
```

</details>