| { | |
| "model_id": "HiTZ/stt_eseu_conformer_transducer_large", | |
| "model_type": "parakeet_rnnt", | |
| "language": "", | |
| "sample_rate": 16000, | |
| "max_audio_seconds": 15.0, | |
| "max_audio_samples": 240000, | |
| "vocab_size": 256, | |
| "blank_id": 256, | |
| "checkpoint": { | |
| "type": "pretrained", | |
| "model_id": "HiTZ/stt_eseu_conformer_transducer_large" | |
| }, | |
| "coreml": { | |
| "compute_precision": "FLOAT32", | |
| "quantization": "6bit_palettize_kmeans" | |
| }, | |
| "components": { | |
| "mel_encoder": { | |
| "path": "parakeet_mel_encoder.mlpackage", | |
| "inputs": { | |
| "audio_signal": [ | |
| 1, | |
| 240000 | |
| ], | |
| "audio_length": [ | |
| 1 | |
| ] | |
| }, | |
| "outputs": { | |
| "encoder": [ | |
| 1, | |
| 512, | |
| 376 | |
| ], | |
| "encoder_length": [ | |
| 1 | |
| ] | |
| } | |
| }, | |
| "decoder": { | |
| "path": "parakeet_decoder.mlpackage", | |
| "inputs": { | |
| "targets": [ | |
| 1, | |
| 1 | |
| ], | |
| "target_length": [ | |
| 1 | |
| ], | |
| "h_in": [ | |
| 1, | |
| 1, | |
| 640 | |
| ], | |
| "c_in": [ | |
| 1, | |
| 1, | |
| 640 | |
| ] | |
| }, | |
| "outputs": { | |
| "decoder": [ | |
| 1, | |
| 640, | |
| 1 | |
| ], | |
| "h_out": [ | |
| 1, | |
| 1, | |
| 640 | |
| ], | |
| "c_out": [ | |
| 1, | |
| 1, | |
| 640 | |
| ] | |
| } | |
| }, | |
| "joint_decision_single_step": { | |
| "path": "parakeet_joint_decision_single_step.mlpackage", | |
| "inputs": { | |
| "encoder_step": [ | |
| 1, | |
| 512, | |
| 1 | |
| ], | |
| "decoder_step": [ | |
| 1, | |
| 640, | |
| 1 | |
| ] | |
| }, | |
| "outputs": { | |
| "token_id": [ | |
| 1, | |
| 1, | |
| 1 | |
| ], | |
| "token_prob": [ | |
| 1, | |
| 1, | |
| 1 | |
| ], | |
| "duration": [ | |
| 1, | |
| 1, | |
| 1 | |
| ] | |
| } | |
| } | |
| } | |
| } |