51 lines
1.1 KiB
Markdown
51 lines
1.1 KiB
Markdown
|
---
|
||
|
|
base_model:
|
||
|
|
- Qwen/Qwen2.5-7B-Instruct
|
||
|
|
library_name: transformers
|
||
|
|
tags:
|
||
|
|
- mergekit
|
||
|
|
- merge
|
||
|
|
|
||
|
|
---
|
||
|
|
# phase_f4b
|
||
|
|
|
||
|
|
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 TIES](https://arxiv.org/abs/2311.03099) merge method using [Qwen/Qwen2.5-7B-Instruct](https://huggingface.co/Qwen/Qwen2.5-7B-Instruct) as a base.
|
||
|
|
|
||
|
|
### Models Merged
|
||
|
|
|
||
|
|
The following models were included in the merge:
|
||
|
|
* ./outputs/phase_d/merged_model
|
||
|
|
* ./outputs/phase_e2b/merged_model
|
||
|
|
|
||
|
|
### Configuration
|
||
|
|
|
||
|
|
The following YAML configuration was used to produce this model:
|
||
|
|
|
||
|
|
```yaml
|
||
|
|
# F4-b: DARE-TIES Merge (Phase D 0.7 + E2-b 0.3)
|
||
|
|
# More conservative: Phase D dominant, E2-b as supplement
|
||
|
|
# Hypothesis: Lighter E2-b influence avoids ALF regression
|
||
|
|
|
||
|
|
models:
|
||
|
|
- model: ./outputs/phase_d/merged_model
|
||
|
|
parameters:
|
||
|
|
density: 0.5
|
||
|
|
weight: 0.7
|
||
|
|
- model: ./outputs/phase_e2b/merged_model
|
||
|
|
parameters:
|
||
|
|
density: 0.5
|
||
|
|
weight: 0.3
|
||
|
|
merge_method: dare_ties
|
||
|
|
base_model: Qwen/Qwen2.5-7B-Instruct
|
||
|
|
parameters:
|
||
|
|
normalize: true
|
||
|
|
int8_mask: true
|
||
|
|
dtype: bfloat16
|
||
|
|
|
||
|
|
```
|