|
--- |
|
license: apache-2.0 |
|
language: |
|
- en |
|
tags: |
|
- merge |
|
--- |
|
|
|
 |
|
# To see what will happen |
|
|
|
[Join our Discord!](https://discord.gg/aEGuFph9) |
|
|
|
A merge using the task_arithmetic method of merging. This is a method my team and I haven't experimented with recently, hopefully this will work well. |
|
|
|
You can see the merge script here: |
|
```models: |
|
- model: paulml/NeuralOmniWestBeaglake-7B |
|
parameters: |
|
weight: 1 |
|
- model: FelixChao/Faraday-7B |
|
parameters: |
|
weight: 1 |
|
- model: flemmingmiguel/MBX-7B-v3 |
|
parameters: |
|
weight: 1 |
|
- model: paulml/NeuralOmniBeagleMBX-v3-7B |
|
parameters: |
|
weight: 1 |
|
merge_method: task_arithmetic |
|
base_model: paulml/NeuralOmniBeagleMBX-v3-7B |
|
parameters: |
|
normalize: true |
|
int8_mask: true |
|
dtype: float16 |
|
|
