part1_sft_metadata.json

{
  "model_id": "Qwen/Qwen2.5-1.5B-Instruct",
  "dataset_id": "medalpaca/medical_meadow_medqa",
  "train_rows": 6106,
  "val_rows": 2036,
  "max_length": 768,
  "train_batch_size": 4,
  "eval_batch_size": 16,
  "gradient_accumulation_steps": 16,
  "effective_batch_size": 64,
  "learning_rate": 0.0002,
  "weight_decay": 0.0,
  "num_epochs": 3,
  "warmup_ratio": 0.03,
  "early_stopping_patience": 1,
  "lora_rank": 16,
  "lora_alpha": 32,
  "lora_dropout": 0.05,
  "lora_target_modules": [
    "q_proj",
    "k_proj",
    "v_proj",
    "o_proj",
    "gate_proj",
    "up_proj",
    "down_proj"
  ],
  "train_runtime_seconds": 999.5917,
  "train_samples_per_second": 18.325,
  "adapter_dir": "/root/SafeGenAI/work/models/model_sft_lora_adapter"
}