| { |
| "artifacts": { |
| "embedding.mlmodelc": { |
| "sha256": "a57e6c967d83725c616202b60743b63a6e768c7f6da5159f406e3e2b4fad863d", |
| "size_bytes": 27507286 |
| }, |
| "plda.safetensors": { |
| "sha256": "aff6294b68b66adcbc1c2a402b1379ecfdd98d8d759dc2cca62b5380babea359", |
| "size_bytes": 199624 |
| }, |
| "segmentation.mlmodelc": { |
| "sha256": "ec6f5f62aaa924963f5c11338aaf6bcd0be35b34fe82c9a2957709d8378db07f", |
| "size_bytes": 5988119 |
| } |
| }, |
| "clustering": { |
| "Fa": 0.07, |
| "Fb": 0.8, |
| "ahc_linkage": "centroid", |
| "ahc_metric": "euclidean_on_l2_normalized_embeddings", |
| "algorithm": "VBx", |
| "constrained_assignment": true, |
| "convergence_epsilon": 0.0001, |
| "filter_min_active_ratio": 0.2, |
| "initial_smoothing": 7.0, |
| "kmeans_n_init": 3, |
| "kmeans_random_seed": 42, |
| "max_iterations": 20, |
| "metric": "cosine", |
| "threshold": 0.6, |
| "vbx_speaker_prior_cutoff": 1e-07 |
| }, |
| "embedding": { |
| "compute_precision": "float32", |
| "dimension": 256, |
| "exclude_overlap": true, |
| "fbank": { |
| "center_features_with_global_mean": true, |
| "dither": 0.0, |
| "fft_size": 512, |
| "frame_length_ms": 25.0, |
| "frame_shift_ms": 10.0, |
| "high_frequency_hz": 8000.0, |
| "low_frequency_hz": 20.0, |
| "num_frames": 998, |
| "num_mel_bins": 80, |
| "preemphasis": 0.97, |
| "remove_dc_offset": true, |
| "round_to_power_of_two": true, |
| "snip_edges": true, |
| "use_energy": false, |
| "window": "hamming" |
| }, |
| "masking": { |
| "fallback_to_overlap_inclusive_mask": true, |
| "minimum_embedding_samples": 400, |
| "use_clean_mask_when_active_frames_greater_than": 2 |
| }, |
| "model": "embedding.mlmodelc", |
| "output_name": "embedding", |
| "output_normalization": "none", |
| "output_shape": [ |
| 1, |
| 3, |
| 256 |
| ], |
| "waveform_input_name": "waveform", |
| "waveform_input_shape": [ |
| 1, |
| 1, |
| 160000 |
| ], |
| "weights_input_name": "weights", |
| "weights_input_shape": [ |
| 1, |
| 3, |
| 589 |
| ] |
| }, |
| "license": "cc-by-4.0", |
| "model_type": "pyannote-community-1-coreml", |
| "parity": { |
| "plda_max_absolute_error": 2.4048226445749776e-06, |
| "segmentation": { |
| "agreement": 1.0, |
| "max_absolute_error": 4.38690185546875e-05, |
| "mean_absolute_error": 8.718510798644274e-06 |
| }, |
| "wespeaker": { |
| "agreement": 0.9999999999992925, |
| "max_absolute_error": 5.457550287246704e-06, |
| "mean_absolute_error": 9.935959042195464e-07 |
| } |
| }, |
| "plda": { |
| "dimension": 128, |
| "weights": "plda.safetensors" |
| }, |
| "sample_rate": 16000, |
| "segmentation": { |
| "chunk_duration_seconds": 10.0, |
| "chunk_step_seconds": 1.0, |
| "compute_precision": "float32", |
| "frame_center_start_seconds": 0.0309375, |
| "frame_receptive_field_seconds": 0.0619375, |
| "frame_step_seconds": 0.016875, |
| "input_name": "waveform", |
| "input_shape": [ |
| 1, |
| 1, |
| 160000 |
| ], |
| "max_speakers_per_chunk": 3, |
| "min_duration_off_seconds": 0.0, |
| "model": "segmentation.mlmodelc", |
| "output_name": "log_probabilities", |
| "output_shape": [ |
| 1, |
| 589, |
| 7 |
| ], |
| "powerset_classes": [ |
| [], |
| [ |
| 0 |
| ], |
| [ |
| 1 |
| ], |
| [ |
| 2 |
| ], |
| [ |
| 0, |
| 1 |
| ], |
| [ |
| 0, |
| 2 |
| ], |
| [ |
| 1, |
| 2 |
| ] |
| ], |
| "powerset_max_classes": 2 |
| }, |
| "source_model": "pyannote/speaker-diarization-community-1", |
| "source_revision": "3533c8cf8e369892e6b79ff1bf80f7b0286a54ee", |
| "source_sha256": { |
| "README.md": "2db91f9265bd81f1653ff088b5bff22bf6aebebea03328513af65501643f8a31", |
| "config.yaml": "5ce2bfa9a938dc132cec1172592d65173cbb8f444ea1e4133f10f9391de155be", |
| "embedding/pytorch_model.bin": "6f10ff60898a1d185fa22e1d11e0bfa8a92efec811f11bca48cb8cafebefd929", |
| "plda/plda.npz": "9b77bcd840692710dd3496f62ecfeed8d8e5f002fd991b785079b244eab7d255", |
| "plda/xvec_transform.npz": "325f1ce8e48f7e55e9c8aa47e05d2766b7c48c4b25b8de8dd751e7a4cc5fbe8f", |
| "segmentation/pytorch_model.bin": "7ad24338d844fb95985486eb1a464e32d229f6d7a03c9abe60f978bacf3f816e" |
| } |
| } |
|
|
