training_config.json

{
  "model_name": "meta-llama/Llama-3.2-3B-Instruct",
  "task_type": "CAUSAL_LM",
  "training_parameters": {
    "num_epochs": 1,
    "batch_size": 8,
    "learning_rate": 0.0002,
    "max_length": 1024,
    "num_warmup_steps": 128
  },
  "lora_config": {
    "r": 16,
    "lora_alpha": 32,
    "lora_dropout": 0.1,
    "bias": "none",
    "target_modules": [
      "q_proj",
      "k_proj",
      "v_proj",
      "o_proj",
      "gate_proj",
      "up_proj"
    ]
  },
  "training_results": {
    "final_training_loss": 2.3811,
    "final_validation_loss": 2.2514,
    "final_training_perplexity": 10.82,
    "final_validation_perplexity": 9.5,
    "total_steps": 8012,
    "training_time_hours": 8.1
  },
  "hardware": {
    "platform": "TPU v3-8",
    "framework": "PyTorch + torch_xla",
    "environment": "Kaggle"
  }
}