README.md

---
base_model: []
tags:
- mergekit
- merge

---
# TestBunny-7B

This is a merge of pre-trained language models created using [mergekit](https://github.com/cg123/mergekit).

Using the updated config file and a rope_alpha=2.5, this should be able to handle a context up to 16384 (formerly would start devolving after 4k).
May be unstable past that - have so far been unable to get coherency out fully to 32k. 

## Merge Details
### Merge Method

This model was merged using the [task arithmetic](https://arxiv.org/abs/2212.04089) merge method using D:/ModelMerge/models/Mistral-7B-v0.1 as a base.

### Models Merged

The following models were included in the merge:
* D:\ModelMerge\merges\TestBunny-7B\bunny-a
* D:/ModelMerge/models/LimaRP-Mistral-7B-v0.1

### Configuration

The following YAML configuration was used to produce this model:

```yaml
base_model: D:/ModelMerge/models/Mistral-7B-v0.1
dtype: float16
merge_method: task_arithmetic
slices:
- sources:
  - layer_range: [0, 32]
    model: D:\ModelMerge\merges\TestBunny-7B\bunny-a
    parameters:
      weight: 1.0
  - layer_range: [0, 32]
    model: D:/ModelMerge/models/LimaRP-Mistral-7B-v0.1
    parameters:
      weight: 0.08
  - layer_range: [0, 32]
    model: D:/ModelMerge/models/Mistral-7B-v0.1
```