base_model: /kaggle/working/gpt2
dtype: float32
merge_method: nearswap
models:
- model: /kaggle/working/debias_gpt2
parameters:
  t: 0.0