{
  "architectures": [
    "MGBertForMaskedLM"
  ],
  "attention_probs_dropout_prob": 0.1,
  "classifier_dropout": 0.1,
  "hidden_dropout_prob": 0.1,
  "hidden_size": 256,
  "intermediate_size": 512,
  "layer_norm_eps": 1e-06,
  "max_position_embeddings": 512,
  "model_type": "mgbert",
  "num_attention_heads": 8,
  "num_hidden_layers": 6,
  "pad_token_id": 0,
  "regressor_dropout": 0.1,
  "tie_word_embeddings": false,
  "torch_dtype": "float32",
  "transformers_version": "4.53.0",
  "use_global_token": true,
  "vocab_size": 17
}