README.md

---
base_model:
- Sao10K/32B-Qwen2.5-Kunou-v1
- Kaoeiri/Qwenwify-32B-v3
- Qwen/QwQ-32B-Preview
- Dans-DiscountModels/Qwen2.5-32B-ChatML
library_name: transformers
tags:
- mergekit
- merge
- not-for-all-audiences
license: cc-by-nc-nd-4.0

---
# merge

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 [DARE](https://arxiv.org/abs/2311.03099) [TIES](https://arxiv.org/abs/2306.01708) merge method using [Qwen/QwQ-32B-Preview](https://huggingface.co/Qwen/QwQ-32B-Preview) as a base.

### Models Merged

The following models were included in the merge:
* [Sao10K/32B-Qwen2.5-Kunou-v1](https://huggingface.co/Sao10K/32B-Qwen2.5-Kunou-v1)
* [Kaoeiri/Qwenwify-32B-v3](https://huggingface.co/Kaoeiri/Qwenwify-32B-v3)
* [Dans-DiscountModels/Qwen2.5-32B-ChatML](https://huggingface.co/Dans-DiscountModels/Qwen2.5-32B-ChatML)

### Configuration

The following YAML configuration was used to produce this model:

```yaml
models:
  - model: Kaoeiri/Qwenwify-32B-v3 # Backbone model with 11 pre-merged models
    parameters:
      weight: 1.0
      density: 0.911
  - model: Sao10K/32B-Qwen2.5-Kunou-v1 # Synthetic roleplay model with scenarios
    parameters:
      weight: 0.15
      density: 0.814
  - model: Dans-DiscountModels/Qwen2.5-32B-ChatML # Reasoning and conversational model
    parameters:
      weight: 0.30
      density: 0.871
merge_method: dare_ties
base_model: Qwen/QwQ-32B-Preview # Latest Qwen model for Q&A tasks
parameters:
  density: 0.90 # Target density for the merged model
  epsilon: 0.05 # Small adjustment factor for fine-tuning
  lambda: 1.35 # Balancing factor for weight distribution
dtype: bfloat16 # Efficient precision format for memory optimization
tokenizer_source: union # Combines tokenizers from all models to maximize vocabulary coverage

```