README.md

---
license: apache-2.0
tags:
- MoE
- merge
- mergekit
- Mistral
- Microsoft/WizardLM-2-7B
---

# WizardLM-2-4x7B-MoE

WizardLM-2-4x7B-MoE is an experimental MoE model made with [Mergekit](https://github.com/arcee-ai/mergekit). It was made by combining four [WizardLM-2-7B](https://huggingface.co/microsoft/WizardLM-2-7B) models using the random gate mode. 

Please be sure to set experts per token to 4 for the best results! Context length should be the same as Mistral-7B-Instruct-v0.1 (8k tokens). For instruction templates, Vicuna-v1.1 is recommended.

# Quanitized versions

EXL2 (for fast GPU-only inference): <br />
8_0bpw: https://huggingface.co/Skylaude/WizardLM-2-4x7B-MoE-exl2-8_0bpw (~ 25 GB vram) <br />
6_0bpw: https://huggingface.co/Skylaude/WizardLM-2-4x7B-MoE-exl2-6_0bpw (~ 19 GB vram) <br />
5_0bpw: https://huggingface.co/Skylaude/WizardLM-2-4x7B-MoE-exl2-5_0bpw (~ 16 GB vram) <br />
4_25bpw: https://huggingface.co/Skylaude/WizardLM-2-4x7B-MoE-exl2-4_25bpw (~ 14 GB vram) <br />
3_5bpw: https://huggingface.co/Skylaude/WizardLM-2-4x7B-MoE-exl2-3_5bpw (~ 12 GB vram) <br />
3_0bpw: https://huggingface.co/Skylaude/WizardLM-2-4x7B-MoE-exl2-3_0bpw (~ 11 GB vram)

GGUF (for mixed GPU+CPU inference or CPU-only inference):  <br />
https://huggingface.co/mradermacher/WizardLM-2-4x7B-MoE-GGUF <br />
Thanks to [Michael Radermacher](https://huggingface.co/mradermacher) for making these quants!

# Evaluation

I don't expect this model to be that great since it's something that I made as an experiment. However, I will submit it to the Open LLM Leaderboard to see how it matches up against some other models (particularly WizardLM-2-7B and WizardLM-2-70B). 

# Mergekit config
```
base_model: models/WizardLM-2-7B
gate_mode: random
dtype: float16
experts_per_token: 4
experts:
  - source_model: models/WizardLM-2-7B
  - source_model: models/WizardLM-2-7B
  - source_model: models/WizardLM-2-7B
  - source_model: models/WizardLM-2-7B
```