---
license: apache-2.0
library_name: transformers
base_model:
- nbeerbower/Dumpling-Qwen2.5-1.5B-v2
datasets:
- flammenai/MahouMix-v1
---
> 🧪 **Experimental**
>
> Attempt to fix formatting (unsuccessful)

# Mahou-1.5-Qwen2.5-1.5B-E3

Mahou is designed to provide short messages in a conversational context. It is capable of casual conversation and character roleplay.

### Chat Format

This model has been trained to use ChatML format.

```
<|im_start|>system
{{system}}<|im_end|>
<|im_start|>{{char}}
{{message}}<|im_end|>
<|im_start|>{{user}}
{{message}}<|im_end|>
```

### Roleplay Format

- Speech without quotes.
- Actions in `*asterisks*`

```
*leans against wall cooly* so like, i just casted a super strong spell at magician academy today, not gonna lie, felt badass.
```

### SillyTavern Settings

1. Use ChatML for the Context Template.
2. Enable Instruct Mode.
3. Use the [Mahou ChatML Instruct preset](https://huggingface.co/datasets/flammenai/Mahou-ST-ChatML-Instruct/raw/main/Mahou.json).
4. Use the [Mahou Sampler preset](https://huggingface.co/datasets/flammenai/Mahou-ST-Sampler-Preset/raw/main/Mahou.json).

### Method

[QLoRA ORPO finetuned](https://mlabonne.github.io/blog/posts/2024-04-19_Fine_tune_Llama_3_with_ORPO.html) with 2x RTX 3090 for 3 epochs.