mergekit_config.yml

base_model: Qwen/Qwen3-8B-Base
dtype: bfloat16
merge_method: della
modules:
  default:
    slices:
    - sources:
      - layer_range: [0, 36]
        model: deepseek-ai/deepseek-r1-0528-qwen3-8b
        parameters:
          density: 1.0
          lambda: 0.9
          weight: 1.0
      - layer_range: [0, 36]
        model: Qwen/Qwen3-8B-Base
parameters:
  density: 1.0
  int8_mask: 1.0
  lambda: 0.9
  normalize: 1.0
  weight: 1.0
tokenizer_source: deepseek-ai/deepseek-r1-0528-qwen3-8b