training:
  data: weights/sample_data/sample_data_final.pkl
  output_dir: weight/training
  epochs: 10
  train_steps: 8
  validation_steps: 2
  validation_fraction: 0.2
  batch_size: 1
  learning_rate: 1.0e-5
  weight_decay: 1.0e-6
  gradient_clip: 1.0
  patience: 3
  seed: 2026
  train_modules: decoder