| { | |
| "architectures": [ | |
| "ParakeetForRNNT" | |
| ], | |
| "blank_token_id": 1024, | |
| "decoder_hidden_size": 640, | |
| "dtype": "float32", | |
| "encoder_config": { | |
| "activation_dropout": 0.1, | |
| "attention_bias": true, | |
| "attention_dropout": 0.1, | |
| "conv_kernel_size": 9, | |
| "convolution_bias": true, | |
| "dropout": 0.1, | |
| "dropout_positions": 0.0, | |
| "hidden_act": "silu", | |
| "hidden_size": 512, | |
| "initializer_range": 0.02, | |
| "intermediate_size": 2048, | |
| "layerdrop": 0.1, | |
| "max_position_embeddings": 5000, | |
| "model_type": "parakeet_encoder", | |
| "num_attention_heads": 8, | |
| "num_hidden_layers": 17, | |
| "num_key_value_heads": 8, | |
| "num_mel_bins": 80, | |
| "scale_input": true, | |
| "subsampling_conv_channels": 256, | |
| "subsampling_conv_kernel_size": 3, | |
| "subsampling_conv_stride": 2, | |
| "subsampling_factor": 8 | |
| }, | |
| "hidden_act": "relu", | |
| "initializer_range": 0.02, | |
| "is_encoder_decoder": true, | |
| "max_symbols_per_step": 10, | |
| "model_type": "parakeet_rnnt", | |
| "num_decoder_layers": 1, | |
| "pad_token_id": 0, | |
| "transformers_version": "5.17.0", | |
| "vocab_size": 1025 | |
| } | |