Update README.md
Browse files
README.md
CHANGED
|
@@ -1,3 +1,18 @@
|
|
| 1 |
---
|
| 2 |
license: apache-2.0
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 3 |
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
---
|
| 2 |
license: apache-2.0
|
| 3 |
+
tags:
|
| 4 |
+
- MoE
|
| 5 |
+
- merge
|
| 6 |
+
- mergekit
|
| 7 |
+
- Mistral
|
| 8 |
+
- Microsoft/WizardLM-2-7B
|
| 9 |
---
|
| 10 |
+
|
| 11 |
+
# WizardLM-2-4x7B-MoE-exl2-3_5bpw
|
| 12 |
+
|
| 13 |
+
This is a quantized version of [WizardLM-2-4x7B-MoE](https://huggingface.co/Skylaude/WizardLM-2-4x7B-MoE) an experimental MoE model made with [Mergekit](https://github.com/arcee-ai/mergekit). Quantization was done using version 0.0.18 of [ExLlamaV2](https://github.com/turboderp/exllamav2).
|
| 14 |
+
|
| 15 |
+
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.
|
| 16 |
+
|
| 17 |
+
For more information see the [original repository](https://huggingface.co/Skylaude/WizardLM-2-4x7B-MoE).
|
| 18 |
+
|