|
--- |
|
base_model: |
|
- CultriX/SeQwence-14B |
|
- VAGOsolutions/SauerkrautLM-v2-14b-DPO |
|
- v000000/Qwen2.5-Lumen-14B |
|
- CultriX/Qwen2.5-14B-Wernicke |
|
- Qwen/Qwen2.5-14B |
|
- CultriX/Qwen2.5-14B-MegaMerge-pt2 |
|
library_name: transformers |
|
tags: |
|
- mergekit |
|
- merge |
|
license: apache-2.0 |
|
language: |
|
- en |
|
metrics: |
|
- accuracy |
|
pipeline_tag: text-generation |
|
--- |
|
# 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/Qwen2.5-14B](https://huggingface.co/Qwen/Qwen2.5-14B) as a base. |
|
|
|
### Models Merged |
|
|
|
The following models were included in the merge: |
|
* [CultriX/SeQwence-14B](https://huggingface.co/CultriX/SeQwence-14B) |
|
* [VAGOsolutions/SauerkrautLM-v2-14b-DPO](https://huggingface.co/VAGOsolutions/SauerkrautLM-v2-14b-DPO) |
|
* [v000000/Qwen2.5-Lumen-14B](https://huggingface.co/v000000/Qwen2.5-Lumen-14B) |
|
* [CultriX/Qwen2.5-14B-Wernicke](https://huggingface.co/CultriX/Qwen2.5-14B-Wernicke) |
|
* [CultriX/Qwen2.5-14B-MegaMerge-pt2](https://huggingface.co/CultriX/Qwen2.5-14B-MegaMerge-pt2) |
|
|
|
### Configuration |
|
|
|
The following YAML configuration was used to produce this model: |
|
|
|
```yaml |
|
|
|
models: |
|
- model: CultriX/Qwen2.5-14B-Wernicke |
|
parameters: |
|
weight: 0.35 # Strong performance in GPQA, MUSR, and MMLU-PRO |
|
density: 0.6 # Retain 60% of significant parameters |
|
- model: VAGOsolutions/SauerkrautLM-v2-14b-DPO |
|
parameters: |
|
weight: 0.30 # Exceptional IFEval and MATH Level 5 capabilities |
|
density: 0.6 # Retain 60% of significant parameters |
|
- model: CultriX/Qwen2.5-14B-MegaMerge-pt2 |
|
parameters: |
|
weight: 0.20 # Balanced contributions to Truthful QA and MMLU |
|
density: 0.5 # Retain 50% of significant parameters |
|
- model: CultriX/SeQwence-14B |
|
parameters: |
|
weight: 0.15 # Provides diverse data and generalization |
|
density: 0.4 # Retain 40% of significant parameters |
|
- model: v000000/Qwen2.5-Lumen-14B |
|
parameters: |
|
weight: 0.10 # Enhances creative and narrative tasks |
|
density: 0.5 # Retain 50% for task diversity |
|
base_model: Qwen/Qwen2.5-14B |
|
merge_method: dare_ties |
|
parameters: |
|
normalize: true # Ensures parameter scaling compatibility |
|
int8_mask: true # Optimizes memory and computational efficiency |
|
dtype: bfloat16 |
|
tokenizer_source: Qwen/Qwen2.5-14B-Instruct |
|
|
|
|
|
``` |
