config.json

{
    "architectures": ["BertForMaskedLM"],
    "hidden_size": 256,
    "intermediate_size": 1024,
    "max_position_embeddings": 512,
    "num_attention_heads": 4,
    "num_hidden_layers": 6,
    "vocab_size": 30522,
    "hidden_act": "gelu",
    "initializer_range": 0.02,
    "layer_norm_eps": 1e-12,
    "model_type": "bert",
    "pad_token_id": 0
  }