{
  "model_name": "qwen3-14b-sft-merged",
  "base_model": "Qwen/Qwen3-14B",
  "training_method": "FSDP + QLoRA",
  "training_date": "2025-08-23",
  "hardware": "8x H100 80GB HBM3",
  "hyperparameters": {
    "epochs": 1,
    "micro_batch_size": 1,
    "gradient_accumulation_steps": 16,
    "learning_rate": 1e-4,
    "scheduler": "cosine",
    "warmup_ratio": 0.03,
    "optimizer": "adamw_torch",
    "sequence_length": 1024,
    "precision": "bfloat16",
    "flash_attention": true
  },
  "lora_config": {
    "r": 64,
    "alpha": 16,
    "dropout": 0.1,
    "target_modules": "linear_layers",
    "quantization": "4bit"
  },
  "fsdp_config": {
    "sharding_strategy": "full_shard",
    "auto_wrap_policy": "transformer_layer",
    "transformer_layer_cls": "Qwen3DecoderLayer"
  }
}