thlurte's picture
Upload README.md with huggingface_hub
e8954bf verified
|
Raw
History Blame Contribute Delete
1.15 kB
---
license: apache-2.0
base_model:
- Qwen/Qwen2.5-Coder-1.5B
- Qwen/Qwen2.5-Math-1.5B
tags:
- merge
- mergekit
- dare_ties
- code
- math
pipeline_tag: text-generation
---
# Qwen2.5-CodeMath-1.5B 🧮💻
**Qwen2.5-CodeMath-1.5B** is a fused foundation model created by merging **[`Qwen/Qwen2.5-Coder-1.5B`](https://huggingface.co/Qwen/Qwen2.5-Coder-1.5B)** and **[`Qwen/Qwen2.5-Math-1.5B`](https://huggingface.co/Qwen/Qwen2.5-Math-1.5B)** using **DARE-TIES**.
## 📌 Architecture & Merge Configuration
- **Base Anchor**: `Qwen/Qwen2.5-1.5B`
- **Method**: DARE-TIES (Drop And REscale with Task Vector Interference Elimination)
- **Model 1**: `Qwen/Qwen2.5-Coder-1.5B` (Weight: `0.60`, Density: `0.80`)
- **Model 2**: `Qwen/Qwen2.5-Math-1.5B` (Weight: `0.40`, Density: `0.80`)
- **Precision**: `bfloat16`
```yaml
merge_method: dare_ties
base_model: Qwen/Qwen2.5-1.5B
models:
- model: Qwen/Qwen2.5-Coder-1.5B
parameters:
weight: 0.6
density: 0.8
- model: Qwen/Qwen2.5-Math-1.5B
parameters:
weight: 0.4
density: 0.8
parameters:
int8_mask: true
dtype: bfloat16
tokenizer_source: Qwen/Qwen2.5-Coder-1.5B
```