Hervé Bredin
|
4e558df3ee
|
fix: fix Audio.crop with negative start time and mode="pad"
|
2021-12-14 17:03:21 +01:00 |
|
Hervé BREDIN
|
243ca71c6f
|
feat: add support for min/max number of speakers (#827)
|
2021-12-13 15:23:46 +01:00 |
|
Hervé BREDIN
|
ed815c0d0a
|
feat: add support for default pipeline parameters (#826)
|
2021-12-12 09:40:18 +01:00 |
|
Hervé BREDIN
|
3bef756356
|
chore: factorize resegmentation and diarization pipelines
|
2021-12-12 09:34:18 +01:00 |
|
Hervé BREDIN
|
94060b4d32
|
setup: update pytorch and pytorch-lightning dependencies (#728)
Related: #744
|
2021-12-07 15:57:57 +01:00 |
|
Hervé BREDIN
|
0621f25559
|
feat: add offline speaker diarization pipeline (#666)
* BREAKING: repurpose "mode" option to use zero-padding in Audio.crop with out-of-bounds segments
* BREAKING: rename "fixed" option to "duration" in Audio.crop
* BREAKING: remove Segmentation pipeline
* BREAKING: rename returns_cost to return_cost in permutate
* feat: add Inference.{trim | aggregate | stitch} methods
* feat: add pyannote.audio.pipelines.clustering module
* feat: add wrapper for Speechbrain speaker embedding
* feat: add "binarize" utility function
* doc: add documentation for high-level API
* setup: switch to pyannote.core 4.3
|
2021-12-07 15:43:02 +01:00 |
|
 Juanma CoriaandHervé BREDIN
|
eb39c87265
|
fix: do not reset task-dependent layers weights when changing task
Co-authored-by: Hervé BREDIN <hbredin@users.noreply.github.com>
|
2021-12-03 14:33:26 +01:00 |
|
Hervé Bredin
|
39e78dac2f
|
Merge branch 'develop' of https://github.com/pyannote/pyannote-audio into develop
|
2021-12-03 14:18:02 +01:00 |
|
 
|
78ca8c87a6
|
feat: add audio.vad prodigy recipe
Co-authored-by: J-Petiot <jpetiot@sylvebarbe.localdomain>
Co-authored-by: Hervé BREDIN <hbredin@users.noreply.github.com>
|
2021-12-01 15:32:03 +01:00 |
|
Hervé Bredin
|
5058225fbc
|
Merge branch 'develop' of https://github.com/pyannote/pyannote-audio into develop
|
2021-09-21 12:04:20 +02:00 |
|
Hervé Bredin
|
995ddee906
|
fix: fix estimation of number of batches per epoch
|
2021-09-21 12:02:38 +02:00 |
|
Hervé Bredin
|
88ff6e1349
|
Merge branch 'develop' of https://github.com/pyannote/pyannote-audio into develop
|
2021-09-20 10:17:56 +02:00 |
|
Hervé BREDIN
|
860973f6c7
|
improve: homogeneize scheduler API (#757)
|
2021-09-17 10:03:06 +02:00 |
|
Hervé BREDIN
|
17c763d124
|
BREAKING: use AUROC instead of FBeta for validation (#756)
|
2021-09-16 13:45:36 +02:00 |
|
Hervé BREDIN
|
e083e99ed5
|
feat: speed up segmentation pipeline optimization with caching (#752)
* setup: switch to pyannote.pipeline 2.1
* fix: fixes #743
|
2021-09-15 22:17:50 +02:00 |
|
Hervé Bredin
|
6697fc8118
|
Merge branch 'develop' of https://github.com/pyannote/pyannote-audio into develop
|
2021-09-13 15:17:42 +02:00 |
|
Hervé Bredin
|
8424f59edd
|
fix: switch to triangular
|
2021-09-13 15:17:24 +02:00 |
|
Hervé Bredin
|
9f3db894f7
|
Merge branch 'develop' of https://github.com/pyannote/pyannote-audio into develop
|
2021-09-11 11:22:25 +02:00 |
|
Hervé Bredin
|
7ba433a861
|
fix: switch to triangular
|
2021-09-11 11:21:40 +02:00 |
|
Hervé Bredin
|
c492e10263
|
Merge branch 'develop' of https://github.com/pyannote/pyannote-audio into develop
|
2021-09-10 16:07:12 +02:00 |
|
Hervé BREDIN
|
1bf1843c3a
|
feat(cli): add option to configure LR scheduler (#747)
|
2021-09-10 16:06:27 +02:00 |
|
Hervé BREDIN
|
133d6657c1
|
feat: add option to force maximum number of speakers (#746)
|
2021-09-10 09:58:00 +02:00 |
|
Hervé Bredin
|
23c4eaaed5
|
Merge branch 'develop' of https://github.com/pyannote/pyannote-audio into develop
|
2021-09-08 16:06:12 +02:00 |
|
Hervé Bredin
|
39633d9d3a
|
fix: remove augmentation tests
|
2021-09-08 14:34:31 +02:00 |
|
Hervé BREDIN
|
8ab7ebe044
|
fix: fix fine-tuning and transfer learning API (#737)
* fix: fixes #736
* feat: add more unit tests
|
2021-09-08 14:23:01 +02:00 |
|
Hervé BREDIN
|
1c6fd10710
|
Update README.md
|
2021-09-06 15:43:10 +02:00 |
|
Hervé BREDIN
|
12c1574754
|
feat: add support for multi-GPU for training/validating speaker embedding
* BREAKING: use AUROC as validation metric (instead of EER)
* fix: fixes #591
* feat: add basic speaker verification pipeline
* feat: add basic script to compute EER
|
2021-08-11 12:12:27 +02:00 |
|
Hervé BREDIN
|
9a011f99ac
|
fix: fix embedding train dataloader (#725)
|
2021-08-09 22:11:39 +02:00 |
|
Hervé BREDIN
|
235a1db9df
|
fix: update API to PL 1.4 (#723)
|
2021-08-09 11:53:44 +02:00 |
|
Hervé BREDIN
|
4b8a6b6fdd
|
feat: switch to torchmetrics (#722)
|
2021-08-09 10:00:40 +02:00 |
|
Hervé BREDIN
|
d8112ea849
|
fix: fix demo link
|
2021-08-09 09:20:52 +02:00 |
|
Hervé Bredin
|
a188f1d305
|
Merge branch 'develop' of https://github.com/pyannote/pyannote-audio into develop
|
2021-08-04 14:37:47 +02:00 |
|
Hervé BREDIN
|
b2cea93fc7
|
fix: fix link to demo
|
2021-07-16 10:43:48 +02:00 |
|
Hervé Bredin
|
436216c26d
|
Merge branch 'develop' of https://github.com/pyannote/pyannote-audio into develop
|
2021-06-22 14:42:14 +02:00 |
|
Hervé BREDIN
|
2ecdc23c21
|
chore: make tensorboard logs lighter (#687)
|
2021-06-22 14:41:57 +02:00 |
|
Hervé BREDIN
|
03484f0dd7
|
doc: link to pyannote's AMI diarization setup fork
|
2021-06-17 14:12:57 +02:00 |
|
Hervé BREDIN
|
b810445728
|
doc: link to pyannote models
|
2021-06-17 14:11:08 +02:00 |
|
Hervé BREDIN
|
20213e7157
|
fix: fix resegmentation pipeline (#683)
|
2021-06-09 14:19:32 +02:00 |
|
Hervé Bredin
|
246318153d
|
Merge branch 'develop' of https://github.com/pyannote/pyannote-audio into develop
|
2021-06-08 21:07:06 +02:00 |
|
Hervé BREDIN
|
c535d3aa16
|
chore: simplify Inference codebase
|
2021-06-08 13:56:10 +02:00 |
|
Hervé BREDIN
|
4c1da2cfad
|
hack: fix huggingface.co download counter (#681)
|
2021-06-08 13:46:39 +02:00 |
|
Hervé BREDIN
|
3c988c028d
|
fix: fix xvector architecture and training (#676)
|
2021-06-06 13:16:59 +02:00 |
|
Hervé BREDIN
|
c69e92df77
|
chore: remove multi-task support (#675)
|
2021-06-04 16:35:32 +02:00 |
|
Hervé Bredin
|
062bd7bd9a
|
git: add AMI-diarization-setup submodule
|
2021-04-27 13:44:58 +02:00 |
|
 Loreto ParisiandHervé BREDIN
|
20e567359d
|
feat: add cache_dir option to {Model|Pipeline}.from_pretrained
Co-authored-by: Hervé BREDIN <hbredin@users.noreply.github.com>
|
2021-04-16 09:09:27 +02:00 |
|
Hervé Bredin
|
de97df6990
|
Merge branch 'develop' of https://github.com/pyannote/pyannote-audio into develop
|
2021-04-13 15:40:13 +02:00 |
|
Harry Coultas Blum
|
3c4d0a68a3
|
feat: add support for file-like objects (#640)
|
2021-04-12 11:31:13 +02:00 |
|
Hervé BREDIN
|
37a6760738
|
feat: add option to pass arguments to inference (#655)
|
2021-04-09 13:29:15 +02:00 |
|
Hervé BREDIN
|
7f90409756
|
fix: update installation instructions and requirements (#653)
|
2021-04-07 14:43:46 +02:00 |
|
Hervé BREDIN
|
e4d6896500
|
fix: bring back notebook-based tests (#654)
|
2021-04-07 13:55:38 +02:00 |
|