{
  "activation_dropout": 0.1,
  "architectures": [
    "EmformerForRNNT"
  ],
  "attention_dropout": 0.1,
  "blank_token_id": 4096,
  "final_dropout": 0.1,
  "hidden_act": "gelu",
  "hidden_dropout": 0.1,
  "hidden_size": 768,
  "initializer_range": 0.02,
  "input_dim": 80,
  "intermediate_size": 2048,
  "joiner_activation": "relu",
  "layer_norm_eps": 1e-05,
  "left_context_length": 30,
  "lstm_dropout": 0.3,
  "lstm_hidden_dim": 512,
  "lstm_layer_norm": 0.001,
  "lstm_layer_norm_epsilon": 0.001,
  "model_type": "emformer",
  "num_attention_heads": 8,
  "num_hidden_layers": 20,
  "num_lstm_layers": 3,
  "output_dim": 1024,
  "pad_token_id": 1,
  "right_context_length": 4,
  "segment_length": 16,
  "symbol_embedding_dim": 512,
  "time_reduction_input_dim": 128,
  "time_reduction_stride": 4,
  "torch_dtype": "float32",
  "transformers_version": "4.20.0.dev0",
  "vocab_size": 4097
}