model_config.json

{
  "model_type": "small_transformer",
  "hidden_size": 48,
  "num_attention_heads": 1,
  "num_hidden_layers": 3,
  "vocab_size": 30522,
  "param_estimate": 796850
}