{
  "model_type": "transformer",
  "architectures": ["CausalLM"],
  "hidden_size": 2048,
  "num_hidden_layers": 24,
  "num_attention_heads": 16,
  "intermediate_size": 8192,
  "vocab_size": 50000,
  "max_position_embeddings": 4096,
  "layer_norm_eps": 1e-5,
  "initializer_range": 0.02
}