{
  "model_type": "VerySmollGPT",
  "architectures": [
    "VerySmollGPT"
  ],
  "vocab_size": 104,
  "d_model": 256,
  "n_layers": 6,
  "n_heads": 8,
  "d_ff": 1024,
  "max_seq_len": 128,
  "dropout": 0.1,
  "block_size": 128,
  "tie_word_embeddings": true,
  "training_config": {
    "num_epochs": 3,
    "batch_size": 16,
    "learning_rate": 0.0003,
    "weight_decay": 0.01
  }
}