| name: &name "SpeakerIdentificationInfer" | |
| data: | |
| enrollment_manifest: ??? | |
| test_manifest: ??? | |
| out_manifest: './infer_output.json' | |
| sample_rate: 16000 | |
| backend: | |
| backend_model: cosine_similarity # supported backends are cosine_similarity and neural_classifier | |
| cosine_similarity: | |
| model_path: titanet_large # or path to .nemo file | |
| batch_size: 32 | |
| neural_classifier: | |
| model_path: ??? # path to neural model trained/finetuned with enrollment dataset | |
| batch_size: 32 | |
| # json manifest line example | |
| # | |
| # enrollment_manifest: | |
| # {"audio_filepath": "/path/to/audio_file", "offset": 0, "duration": null, "label": "<speaker_label>"} | |
| # | |
| # test_manifest: | |
| # {"audio_filepath": "/path/to/audio_file", "offset": 0, "duration": null, "label": "infer"} | |
| # | |