---
library_name: transformers
license: apache-2.0
base_model:
- nbeerbower/Mahou-1.5-mistral-nemo-12B-lorablated
datasets:
- nbeerbower/Arkhaios-DPO
- nbeerbower/Purpura-DPO
---

![image/png](https://huggingface.co/nbeerbower/Mistral-Nemo-Prism-12B/resolve/main/prism-cover.png?download=true)

> 🧪 **Just Another Model Experiment**
>
> This is one of many experimental iterations I'm sharing publicly while I mess around with training parameters and ideas. It's not a "real" release - just me being transparent about my learning process. Feel free to look under the hood, but don't expect anything production-ready!

# Mistral-Nemo-Prism-12B

[Mahou-1.5-mistral-nemo-12B-lorablated](https://huggingface.co/nbeerbower/Mahou-1.5-mistral-nemo-12B-lorablated) finetuned on [Arkhaios-DPO](https://huggingface.co/datasets/nbeerbower/Arkhaios-DPO) and [Purpura-DPO](https://huggingface.co/datasets/nbeerbower/Purpura-DPO).

The goal was to reduce archaic language and purple prose in a completely uncensored model.

### Method

[ORPO tuned](https://mlabonne.github.io/blog/posts/2024-04-19_Fine_tune_Llama_3_with_ORPO.html) with 2x A100 for 5 epochs.

The learning rate was lowered to 3e-6 for this version. In addition, a system prompt was introduced to further augment the prompts and encourage responses to match the data.