|
|
--- |
|
|
base_model: |
|
|
- Theros/Qwen2.5-ColdBrew-R1 |
|
|
library_name: transformers |
|
|
tags: |
|
|
- mergekit |
|
|
- merge |
|
|
|
|
|
--- |
|
|
# 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 Passthrough merge method. |
|
|
|
|
|
### Models Merged |
|
|
|
|
|
The following models were included in the merge: |
|
|
* [Theros/Qwen2.5-ColdBrew-R1](https://huggingface.co/Theros/Qwen2.5-ColdBrew-R1) |
|
|
|
|
|
### Configuration |
|
|
|
|
|
The following YAML configuration was used to produce this model: |
|
|
|
|
|
```yaml |
|
|
name: Q2.5-ColdBrew-R1-Obsidian |
|
|
const_tag: &scale_factor 0.7071067812 # 1/sqrt(2) scaling for stability |
|
|
|
|
|
attenuate-env: &attenuated_env |
|
|
parameters: |
|
|
scale: |
|
|
- filter: q_proj |
|
|
value: *scale_factor |
|
|
- filter: k_proj |
|
|
value: *scale_factor |
|
|
- value: 1.0 |
|
|
|
|
|
slices: |
|
|
- sources: |
|
|
- model: Theros/Qwen2.5-ColdBrew-R1 |
|
|
layer_range: [0, 8] # Retaining foundational knowledge and language structure. |
|
|
|
|
|
- sources: |
|
|
- model: Theros/Qwen2.5-ColdBrew-R1 |
|
|
layer_range: [9, 19] # Full-strength mid-range layers. |
|
|
|
|
|
- sources: |
|
|
- model: Theros/Qwen2.5-ColdBrew-R1 |
|
|
layer_range: [10, 18] # Targeted reinforcement, slightly attenuated to avoid over-dominance. |
|
|
<<: *attenuated_env |
|
|
|
|
|
- sources: |
|
|
- model: Theros/Qwen2.5-ColdBrew-R1 |
|
|
layer_range: [19, 28] # Keeping higher-level abstract processing untouched for stability. |
|
|
|
|
|
merge_method: passthrough |
|
|
dtype: bfloat16 |
|
|
normalize: true |
|
|
int8_mask: true |
|
|
|
|
|
``` |
|
|
|
