{
  "architectures": [
    "GPT"
  ],
  "vocab_size": 65536,
  "n_layer": 24,
  "n_head": 16,
  "hidden_size": 2048,
  "max_position_embeddings": 2048,
  "model_type": "gpt",
  "torch_dtype": "bfloat16"
}