{
  "architectures": [
    "PipelinedRobertaForMaskedLM"
  ],
  "attention_probs_dropout_prob": 0.3,
  "bos_token_id": 3,
  "classifier_dropout": null,
  "eos_token_id": 4,
  "hidden_act": "gelu_new",
  "hidden_dropout_prob": 0.15,
  "hidden_size": 560,
  "initializer_range": 0.02,
  "intermediate_size": 1412,
  "layer_norm_eps": 1e-12,
  "max_position_embeddings": 128,
  "model_type": "roberta",
  "num_attention_heads": 8,
  "num_hidden_layers": 4,
  "pad_token_id": 1,
  "position_embedding_type": "relative_key_query",
  "torch_dtype": "float16",
  "transformers_version": "4.20.1",
  "type_vocab_size": 1,
  "use_cache": true,
  "vocab_size": 8192
}