| {"sample_rate": 16000, "hop_length": 160, "n_samples": 240000, "n_rows": 1503, "n_mels": 128, "n_fft": 512, "preemph": 0.97, "n_padded_samples": 240480, "valid_frames": 1500, "enc_frames": 188, "joint_hidden": 640, "pred_hidden": 640, "pred_layers": 2, "vocab_size": 8192, "blank_idx": 8192, "durations": [0, 1, 2, 3, 4], "decode_width": 2, "fused_frontend": true, "decode_lanes": 16, "encode_batch": 6, "precision": "float16", "webgpu_decoder": {"file": "decoder.webgpu.onnx", "decode_lanes": 48}, "encoder_external": "encoder.onnx.data"} |