Commit Graph
2256 Commits
Author SHA1 Message Date
Hervé BREDIN 6c6acc9f29 fix: fix repeated argument 2023-05-17 16:07:54 +02:00
Hervé BREDIN b461dc585a wip: playing with SincNet normalization 2023-05-17 15:27:34 +02:00
Hervé BREDIN 1bab814a2a fix: fix prediction shift logic 2023-05-16 10:34:57 +02:00
Hervé BREDIN e84cac42b9 feat: add hydra task 2023-05-15 15:46:19 +02:00
Hervé BREDIN 4a6ea9c825 wip: add streaming speaker diarization task 2023-05-15 15:42:40 +02:00
Hervé BREDIN a58153662b improve: improve error message on missing specifications (#1371) 2023-05-15 09:37:49 +02:00
Hervé BREDIN 31143e0d44 feat(pipeline): make batch sizes mutable (#1365)
* BREAKING(pipeline): default batch sizes to 1
* BREAKING(pipeline): remove `segmentation_duration` option
2023-05-10 17:19:34 +02:00
Dmitry Mukhutdinov 11b56a137a fix: fix broken initialization of Audio (#1344) 2023-04-25 12:10:59 +02:00
Hervé BREDIN c0b82d037a fix(metric): avoid division by zero 2023-04-24 16:25:28 +02:00
Hervé BREDIN b35dd55bd2 feat(cli): add support for AdamW 2023-04-24 16:09:09 +02:00
Hervé BREDIN 95c5334580 feat: stop training on NaNs 2023-04-24 14:47:33 +02:00
Hervé BREDIN 00e7a117bc feat(pipeline): check version compatibility at load time (#1340) 2023-04-21 17:31:12 +02:00
Hervé BREDIN 444720b987 setup: switch to pyannote.database 5.0.1 2023-04-21 15:53:02 +02:00
Hervé BREDIN 075514d732 doc: remove duplicated changelog file 2023-04-21 15:19:36 +02:00
Hervé BREDIN 2623cf1980 BREAKING(io): update Audio behavior for multi-channel audio
- BREAKING(io): channels are now 0-indexed (used to be 1-indexed)
- BREAKING(io): multi-channel audio is no longer downmixed to mono by default.
  You should update how `pyannote.audio.core.io.Audio` is instantiated:
  * replace `Audio()` by `Audio(mono="downmix")`;
  * replace `Audio(mono=True)` by `Audio(mono="downmix")`;
  * replace `Audio(mono=False)` by `Audio()`.
2023-04-21 15:14:56 +02:00
Hervé BREDIN cbe97e2843 fix: skip batches resulting in NaN loss (#1332) 2023-04-19 17:27:46 +02:00
Paul Butler 6ef2b888d7 chore: replace deprecated np.int with int (#1330) 2023-04-19 09:31:58 +02:00
Hervé BREDIN 1dc2abcc98 fix: fix tutorial
Fixes #1331

Note: Segmentation will be renamed SpeakerDiarization in next release.
2023-04-19 09:17:30 +02:00
Will Proctor bbe13955c4 fix: fix corner case in Inference on Windows 2023-04-17 15:39:22 +02:00
Balazs G. Horvath 00fe29fe0d feat(pipeline): add support for @revision of speechbrain models 2023-04-17 15:35:07 +02:00
Hervé BREDIN 9faa8fc8a8 setup: update most dependencies (#1323)
- setup: switch to torch 2.0+ and lightning 2.0+
- setup: switch to torchaudio 2.0+ and soundfile 0.12+
- setup: switch to pyannote.core 5.0+ and pyannote.database 5.0+
- setup: switch to speechbrain 0.5.14+

- BREAKING(task): rename `Segmentation` task to `SpeakerDiarization`
- BREAKING(task): remove support for variable chunk duration

- BREAKING(pipeline): remove `SpeakerSegmentation` pipeline (in favor of `SpeakerDiarization` pipeline)
- BREAKING(pipeline): remove support `FINCHClustering` and `HiddenMarkovModelClustering`

- BREAKING: drop support for Python 3.7
2023-04-17 15:32:37 +02:00
Hervé BREDIN 7121a73f08 feat: add support for label scope and missing classes (#1296)
- setup: switch to pyannote.database 5.0
- feat(task): add support for label scope in speaker diarization task (from pyannote.database 5.0)
- feat(task): add support for missing classes in multi-label segmentation task (from pyannote.database 5.0)
- improve(task): load metadata as tensors/arrays rather than pyannote.core instances
- BREAKING(task): rename `Segmentation` task to `SpeakerDiarization`
- BREAKING(task): remove support for variable chunk duration
2023-04-13 21:15:57 +02:00
Hervé BREDIN 74939acbfa fix: fix faqtory workflow 2023-03-16 15:10:02 +01:00
Hervé BREDIN 39a261a0eb feat: add FAQ based on faqtory (#1283) 2023-03-16 15:04:26 +01:00
Hervé BREDIN b56add2e20 BREAKING(cli): only save top checkpoint 2023-02-20 07:58:58 +01:00
Hervé BREDIN e3dc7d68cc fix: switch back to monitoring DER 2023-02-20 07:54:04 +01:00
Hervé BREDIN 9da0ce8d60 feat: add support for "device" in Pipeline.from_pretrained (#1265) 2023-02-20 07:53:04 +01:00
Hervé BREDIN d4ff0bd23d feat: add option to make progress hook transient (#1264) 2023-02-16 21:17:08 +01:00
Hervé BREDIN 2a9a892726 doc: fix link to pyannote/segmentation 2023-02-16 13:55:58 +01:00
Hervé BREDIN b3385ff709 doc: remove duplicate Colab badge 2023-02-16 13:54:52 +01:00
Hervé BREDIN 733cb433ab Créé avec Colaboratory 2023-02-16 13:49:41 +01:00
Hervé BREDIN 1f83e0b867 feat: add weigh_by_cardinality option (#1256) 2023-02-13 10:34:11 +01:00
Hervé BREDIN 6dab9043b6 Delete suggest-related-links.yml 2023-02-08 09:34:43 +01:00
Hervé BREDIN 38d58e8631 github: add suggest-related-links bot 2023-02-08 09:29:44 +01:00
Hervé BREDIN d065c7181b Merge branch 'develop' of github.com:pyannote/pyannote-audio into develop 2023-02-01 11:25:56 +01:00
Hervé BREDIN 856178a508 fix(pipeline): fix corner case where global number of speakers is smaller than frame-wise one 2023-02-01 11:25:50 +01:00
Hervé BREDIN 9f790f27f6 feat: validate speaker embedding with new EqualErrorRate metric 2023-01-30 16:52:20 +01:00
Hervé BREDIN 559a34a6ad fix: fix embedding validation step (#1245) 2023-01-30 11:06:51 +01:00
FrenchKrab bce740f0b8 fix: fix support for multiple loggers (#1244) 2023-01-30 10:45:56 +01:00
Hervé BREDIN c7325884d8 fix: add missing collate function (#1241) 2023-01-26 09:54:35 +01:00
Hervé BREDIN 5db4783c75 fix: fix corner case with non-speech batch (#1239) 2023-01-24 21:59:45 +01:00
Hervé BREDIN 952e1507fa BREAKING: fix random seed generation
Make sure that multiple nodes use different seeds.
2023-01-24 21:50:41 +01:00
Hervé BREDIN 2dd9af0e6a BREAKING: monitor validation loss instead of DER 2023-01-24 08:58:24 +01:00
Hervé BREDIN f93fe71a88 fix(cli): fix default trainer config 2023-01-20 14:50:52 +01:00
Hervé BREDIN 8f6af1b5de fix(cli): fix default trainer options 2023-01-20 14:27:11 +01:00
Hervé BREDIN 67aa7f0236 fix(cli): remove loss/vad_loss default options 2023-01-20 14:11:44 +01:00
Hervé BREDIN dcfa104d35 Merge branch 'develop' of github.com:pyannote/pyannote-audio into develop 2023-01-20 14:06:29 +01:00
Hervé BREDIN 46563825f1 feat(cli): add "registry" option to training CLI 2023-01-20 14:06:23 +01:00
Ewald Enzinger cb7bfd5a4d setup: switch to pytorch-lightning >= 1.8 (#1210) 2023-01-20 12:46:56 +01:00
Hervé BREDIN 2a4e3bf245 chore: remove main function that is already part of tests 2023-01-20 11:24:56 +01:00