config.json

{
  "architectures": [
    "PinyinCodeForMaskedLM"
  ],
  "auto_map": {
    "AutoConfig": "configuration_pinyin_code.PinyinCodeConfig",
    "AutoModel": "modeling_pinyin_code.PinyinCodeEncoderModel",
    "AutoModelForMaskedLM": "modeling_pinyin_code.PinyinCodeForMaskedLM",
    "AutoTokenizer": [
      "tokenization_pinyin_code.EncodedMandarinTokenizer",
      null
    ]
  },
  "block_size": 512,
  "bos_token_id": 2,
  "cls_token_id": 2,
  "dropout": 0.1,
  "dtype": "float32",
  "eos_token_id": 3,
  "evaluation_backend": "masked_language_modeling",
  "hidden_size": 512,
  "is_decoder": false,
  "mask_token_id": 14,
  "max_position_embeddings": 512,
  "model_type": "pinyin_code",
  "n_embd": 512,
  "n_head": 8,
  "n_layer": 8,
  "num_attention_heads": 8,
  "num_hidden_layers": 8,
  "pad_token_id": 0,
  "patch_pathlib_utf8_open": true,
  "sep_token_id": 3,
  "training_model_type": "bert",
  "transformers_version": "5.10.2",
  "unk_token_id": 1,
  "use_cache": false,
  "vocab_size": 16000
}