merge

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

Merge Details

Merge Method

This model was merged using the SLERP merge method.

Models Merged

The following models were included in the merge:

• Theros/Qwen2.5-ColdBrew-R1-test3
• bunnycore/Qwen-2.5-7B-Stock-Deep-Bespoke-v2

Configuration

The following YAML configuration was used to produce this model:

slices:
  - sources:
      - model: Theros/Qwen2.5-ColdBrew-R1-test3
        layer_range: [0, 28]
      - model: bunnycore/Qwen-2.5-7B-Stock-Deep-Bespoke-v2
        layer_range: [0, 28]
merge_method: slerp
base_model: Theros/Qwen2.5-ColdBrew-R1-test3
parameters:
  t:
    - filter: self_attn
      value: [0.3, 0.5, 0.6, 0.6, 0.7]  # Avoids extreme low/high fluctuations
    - filter: mlp
      value: [0.7, 0.6, 0.5, 0.4, 0.3]  # Gradual shift, avoiding an early MLP spike
    - value: 0.5
dtype: bfloat16
tokenizer_source: union