README.md

---
base_model:
- Sao10K/Frostwind-v2.1-m7
- SanjiWatsuki/Kunoichi-DPO-v2-7B
- macadeliccc/WestLake-7B-v2-laser-truthy-dpo
license: cc-by-nc-4.0
tags:
- moe
- merge
- Roleplay
---
> [!IMPORTANT]
> This model may buggy. Final version is realease. Thank you for using this model.
> 
> For TextGenWebUI user only: Use Transformers to infer/use for all version of this series. I have just known that there is a issue in llamaccp in TextGenWebUI. Please don't use the GGUF version because it have some buggy pointer.
# Experimental 3x7B model

A Experimental MoE Model that custom for all-rounded Roleplay. Well understand Character Card and high logic.

Thank all the origin model author: Sao10K, SanjiWatsuki, macadeliccc, for create those model. Pardon me that I want hide the recipe. :(

If you want 32k context length capable, you could try those versions:

  - [V2](https://huggingface.co/Alsebay/HyouKan-3x7B-V2-32k)   
  - [V2.1](https://huggingface.co/Alsebay/HyouKan-3x7B-V2.1-32k)

Other version: lower the expert to 2:

- https://huggingface.co/Alsebay/Hyou-2x7B

# You may want see this: https://huggingface.co/Alsebay/My_LLMs_Leaderboard
# It's ridiculous that I can run this original version in 4bit, but can't run in GGUF version. Maybe my GPU can't handle it? 

Have try from Q2 to fp16, no hope. 😥 Seem that there is a bug in model pointer. maybe cause by Sao10K/Frostwind-v2.1-m7 because it a experimental model.

Link here: https://huggingface.co/Alsebay/HyouKan-GGUF

# Thank [mradermacher](https://huggingface.co/mradermacher) for quantizing again my model.

[mradermacher](https://huggingface.co/mradermacher) version, he do all the rest of quantization model, and also imatrix: https://huggingface.co/mradermacher/HyouKan-3x7B-GGUF/

# Is this model good? Want more dicussion? Let's me know in community tab! ヾ(≧▽≦*)o