{
  "model_type": "causal_lm",
  "architecture": "Transformer",
  "hidden_size": 1024,
  "num_attention_heads": 16,
  "num_hidden_layers": 12,
  "vocab_size": 32000,
  "max_position_embeddings": 2048,
  "activation_function": "gelu",
  "layer_norm_eps": 1e-12,
  "initializer_range": 0.02
}