mergekit_config.yml

models:
  - model: ./qwen-32b-instruct
    parameters:
      weight: 1
merge_method: ties
base_model: ./qwen-32b
parameters:
  normalize: true
  int8_mask: true
dtype: bfloat16