MiquMaid-v2-2x70B-DPO-GPTQ

safetensors files requiring joining

Note: HF does not support files larger than 50GB. Instructions borrowed from TheBloke on joining files.

To join the files, do the following:

Linux and macOS:

cat gptq_model-4bit-32g.safetensors-part-* > gptq_model-4bit-32g.safetensors && rm gptq_model-4bit-32g.safetensors-part-*

Windows command line:

COPY /B gptq_model-4bit-32g.safetensors-part-a + gptq_model-4bit-32g.safetensors-part-b gptq_model-4bit-32g.safetensors
del gptq_model-4bit-32g.safetensors-part-a gptq_model-4bit-32g.safetensors-part-b


## MiquMaid v2 2x70 DPO

Check out our blogpost about this model series [Here!](https://ikaridevgit.github.io/index.html?blog=blogid-6&bo=true#Miqu-base) - Join our Discord server [Here!](https://discord.gg/Bb8pRUXy3Z)

<center>[<a href="https://huggingface.co/NeverSleep/MiquMaid-v2-70B">V2-70B</a> - <a href="https://huggingface.co/NeverSleep/MiquMaid-v2-70B-DPO">V2-70B-DPO</a> - <a href="https://huggingface.co/NeverSleep/MiquMaid-v2-2x70B">V2-2x70B</a> - <a href="https://huggingface.co/NeverSleep/MiquMaid-v2-2x70B-DPO">V2-2x70B-DPO</a>]
</br>
<div style="width: 100%;">
    <img src="https://cdn-uploads.huggingface.co/production/uploads/63ab1241ad514ca8d1430003/Wbzwoko-IZbOJfvPaImre.png" style="display: block; margin: auto;">
</div></center>

This model uses the Alpaca **prompting format**

Then, we have done a MoE, made of MiquMaid-v2-70B-DPO and Miqu-70B-DPO base, making the model using the finetune AND the base model for each token, working together.

The two model have been trained on DPO for uncensoring, more info on Miqu-70B-DPO [here](https://huggingface.co/Undi95/Miqu-70B-Alpaca-DPO-GGUF)

We have seen a significant improvement, so we decided to share that, even if the model is very big.

## Credits:
- Undi
- IkariDev

## Description

This repo contains FP16 files of MiquMaid-v2-2x70B-DPO.

Switch: [FP16](https://huggingface.co/NeverSleep/MiquMaid-v2-2x70B-DPO) - [GGUF](https://huggingface.co/NeverSleep/MiquMaid-v2-2x70B-DPO-GGUF)

## Training data used:
- [Aesir datasets](https://huggingface.co/MinervaAI)
- [NoRobots](https://huggingface.co/datasets/Doctor-Shotgun/no-robots-sharegpt)
- [limarp](https://huggingface.co/datasets/lemonilia/LimaRP)
- [toxic-dpo-v0.1-sharegpt](https://huggingface.co/datasets/Undi95/toxic-dpo-v0.1-sharegpt)
- [ToxicQAFinal](https://huggingface.co/datasets/NobodyExistsOnTheInternet/ToxicQAFinal)

## DPO training data used:
- [ToxicDPOqa](https://huggingface.co/datasets/NobodyExistsOnTheInternet/ToxicDPOqa)
- [toxic-dpo-v0.1-NoWarning](https://huggingface.co/datasets/Undi95/toxic-dpo-v0.1-NoWarning)

### Custom format:

Instruction:

{system prompt}

Input:

{input}

Response:

{reply}

## Others

Undi: If you want to support us, you can [here](https://ko-fi.com/undiai).

IkariDev: Visit my [retro/neocities style website](https://ikaridevgit.github.io/) please kek