Qwen2.5-7B-Instruct-Math-dare-linear

This is a merge of pre-trained language models created using mergekit.

Performance

Metric	Value
GSM8k (zero-shot)	90.75
HellaSwag (zero-Shot)	80.77
MBPP (zero-shot)	63.08

Merge Details

Merge Method

This model was merged using the Linear DARE merge method using Qwen/Qwen2.5-7B as a base.

Models Merged

The following models were included in the merge:

• Qwen/Qwen2.5-7B-Instruct
• Qwen/Qwen2.5-Math-7B

Configuration

The following YAML configuration was used to produce this model:

base_model: Qwen/Qwen2.5-7B
dtype: bfloat16
merge_method: dare_linear
parameters:
  lambda: 0.7484721287441042
  normalize: 1.0
slices:
- sources:
  - layer_range: [0, 28]
    model: Qwen/Qwen2.5-7B
  - layer_range: [0, 28]
    model: Qwen/Qwen2.5-Math-7B
    parameters:
      density: 0.8456557088847347
      weight: 0.11064925820848412
  - layer_range: [0, 28]
    model: Qwen/Qwen2.5-7B-Instruct
    parameters:
      density: 0.5247829319933462
      weight: 0.6901952279079901