{
  "architectures": [
    "RoFormerForMaskedLM"
  ],
  "attention_probs_dropout_prob": 0.1,
  "bos_token_id": 128000,
  "embedding_size": 1024,
  "eos_token_id": 128001,
  "hidden_act": "silu",
  "hidden_dropout_prob": 0.1,
  "hidden_size": 1024,
  "initializer_range": 0.02,
  "intermediate_size": 3072,
  "layer_norm_eps": 1e-08,
  "max_position_embeddings": 8192,
  "model_type": "roformer",
  "num_attention_heads": 12,
  "num_hidden_layers": 12,
  "pad_token_id": 128255,
  "rotary_value": false,
  "torch_dtype": "float32",
  "transformers_version": "4.42.3",
  "type_vocab_size": 2,
  "use_cache": true,
  "vocab_size": 128256
}