{
  "model_name": "distilbert/distilgpt2",
  "epochs": 2,
  "batch_size": 1,
  "learning_rate": 0.0003,
  "grad_accum_steps": 4,
  "precision": "fp32",
  "quantization_bits": 8,
  "lora_r": 4,
  "lora_alpha": 8,
  "lora_dropout": 0.04,
  "target_modules": [
    "c_attn",
    "c_proj"
  ],
  "total_steps": 20,
  "num_train_samples": 80
}