base_model: Qwen/Qwen2.5-1.5B-Instruct
dtype: float16
merge_method: dare_ties
models:
- model: /kaggle/working/temp_sft_full
  parameters:
    density: 0.30000000000000004
    weight: 1.0