adapter_config.json

{
  "base_model_name_or_path": "unsloth/gemma-7b-bnb-4bit",
  "learning_rate": 0.0003,
  "batch_size": 16,
  "num_train_epochs": 3
}