marcuscedricridia/etr1o-v1.2

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:

• Lunzima/NQLSG-Qwen2.5-14B-MegaFusion-v8
• YOYO-AI/Qwen2.5-14B-1M-YOYO-V3

Configuration

The following YAML configuration was used to produce this model:

base_model: YOYO-AI/Qwen2.5-14B-1M-YOYO-V3
dtype: bfloat16
merge_method: slerp
parameters:
  t:
    - filter: self_attn
      value: [0.8, 0.7, 0.6, 0.5, 0.4]  # More weight to YOYO (IFEval is instruction following)
    - filter: mlp
      value: [0.4, 0.5, 0.6, 0.7, 0.8]  # More weight to MegaFusion (MuSR + GPQA need better reasoning)
    - filter: input_layernorm|post_attention_layernorm
      value: 0.6  
    - value: 0.5   
slices:
- sources:
  - layer_range: [0, 48]  # Early layers focused on IFEval
    model: YOYO-AI/Qwen2.5-14B-1M-YOYO-V3  
  - layer_range: [0, 48]  # Later layers focused on reasoning (MuSR + GPQA)
    model: Lunzima/NQLSG-Qwen2.5-14B-MegaFusion-v8