README.md

---
base_model:
- Qwen/Qwen2-72B-Instruct
- alpindale/magnum-72b-v1
library_name: transformers
tags:
- mergekit
- merge
- OG_finetune_merge

---

WIP of retouched [alpindale/magnum-72b-v1](https://huggingface.co/alpindale/magnum-72b-v1) but I will not use "Magnum" in the name. Call it FinalMix!

Found some issues, trying to fix them for my own usage and adding more RP data with merging.

You can do your own quantized files with the [imatrix.dat file](https://huggingface.co/Undi95/MG-FinalMix-72B/blob/main/imatrix.dat) done with "[wiki.train.raw](https://cosmo.zip/pub/datasets/wikitext-2-raw/)".

Credits to [Alpin](https://huggingface.co/alpindale) and the gang for [magnum-72b-v1](https://huggingface.co/alpindale/magnum-72b-v1), and [Ikari](https://huggingface.co/ikaridev) for his datasets.

### Prompt template ChatML


```
<|im_start|>system
{system_prompt}<|im_end|>
<|im_start|>user
{prompt}<|im_end|>
<|im_start|>assistant
{output}<|im_end|>
```