{
  "model_name": "Qwen/Qwen3-0.6B",
  "num_heads": 4,
  "head_layer_indices": [
    5,
    11,
    16,
    22
  ],
  "quantization": "none",
  "hidden_size": 1024,
  "vocab_size": 151936,
  "num_hidden_layers": 28,
  "training_config": {
    "dataset_name": "wikitext",
    "dataset_config_name": "wikitext-2-raw-v1",
    "batch_size": 4,
    "gradient_accumulation_steps": 32,
    "max_steps": 500,
    "lr": 0.005,
    "max_length": 512
  }
}