---
base_model:
- Qwen/Qwen2.5-Coder-32B-Instruct
- Qwen/Qwen2.5-32B
- Qwen/QwQ-32B
library_name: transformers
tags:
- mergekit
- merge

---
# QwQ-Qwen2.5-Coder-Instruct-32B-MW

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

## Merge Details
### Merge Method

This model was merged using the [SCE](https://arxiv.org/abs/2408.07990) merge method using [Qwen/Qwen2.5-32B](https://huggingface.co/Qwen/Qwen2.5-32B) as a base.

### Models Merged

The following models were included in the merge:
* [Qwen/Qwen2.5-Coder-32B-Instruct](https://huggingface.co/Qwen/Qwen2.5-Coder-32B-Instruct)
* [Qwen/QwQ-32B](https://huggingface.co/Qwen/QwQ-32B)

### Configuration

The following YAML configuration was used to produce this model:

```yaml
models:
  - model: Qwen/Qwen2.5-32B
  - model: Qwen/QwQ-32B
  - model: Qwen/Qwen2.5-Coder-32B-Instruct

merge_method: sce
base_model: Qwen/Qwen2.5-32B
parameters:
  select_topk: 0.6
  layers_weights:
    - pattern: "transformer.h.[0-9]|1[0-5]"  # Regex for layers 0-15
      value: [0.0, 0.7, 0.3]
    - pattern: "transformer.h.(1[6-9]|[2-4][0-9]|5[0-9])"  # Regex for layers 16-59
      value: [0.0, 0.9, 0.1]
    - pattern: "transformer.h.(6[0-3])"  # Regex for layers 60-63
      value: [0.0, 0.8, 0.2]
    - pattern: "lm_head"
      value: [0.0, 0.8, 0.2]
dtype: bfloat16

```