---
base_model:
- HollowMan6/Llama-3.2-3B-SFT-Model-Ocra-500k
- Lyte/Llama-3.2-3B-Overthinker
- EryriLabs/Llama-3.2-SARA-3b
library_name: transformers
tags:
- mergekit
- merge

---
# Llama-3.2-SARAThinker-merged-3b


<figure>
  <img src="SARA-Thinker.png" alt="SARA" width="300">
</figure>


This is a merge of pre-trained language models created using [mergekit](https://github.com/cg123/mergekit). It was developed as part of an ongoing blog series.
I take no responsibility for the ouputs of this model. 

## Merge Details
### Merge Method

This model was merged using the [TIES](https://arxiv.org/abs/2306.01708) merge method using [EryriLabs/Llama-3.2-SARA-3b](https://huggingface.co./EryriLabs/Llama-3.2-SARA-3b) as a base.

### Models Merged

The following models were included in the merge:
* [HollowMan6/Llama-3.2-3B-SFT-Model-Ocra-500k](https://huggingface.co./HollowMan6/Llama-3.2-3B-SFT-Model-Ocra-500k)
* [Lyte/Llama-3.2-3B-Overthinker](https://huggingface.co./Lyte/Llama-3.2-3B-Overthinker)

### Configuration

The following YAML configuration was used to produce this model:

```yaml
models:
  - model: EryriLabs/Llama-3.2-SARA-3b
    parameters:
      density: 0.99  # keeping 90% of base model weights
  - model: Lyte/Llama-3.2-3B-Overthinker
    parameters:
      density: 0.1  # fraction of weights in differences from the base model to retain
      weight:   # weight gradient
        - filter: mlp
          value: 0.1
        - value: 0
  - model: HollowMan6/Llama-3.2-3B-SFT-Model-Ocra-500k
    parameters:
      density: 0.1
      weight: 0.2
merge_method: ties
base_model: EryriLabs/Llama-3.2-SARA-3b
parameters:
  normalize: true
  int8_mask: true
dtype: float16
```