Feat/joint diarization and embedding with prepared data#1583

Draft

clement-pages wants to merge 105 commits intopyannote:developfrom clement-pages:feat/joint-diarization-and-embedding-with-prepared-data

+2,931-10

Commits on Jun 8, 2023

fix: raise TypeError on wrong device type in Pipeline.to and Inference.to
chai3
authored

Commits on Jun 12, 2023

Commits on Jun 15, 2023

feat: update FAQtory default answer
hbredin
authored

Commits on Jun 17, 2023

add draft version of the joint diarization and embedding tasks
clement-pages
committed
Merge branch 'develop' of github.com:clement-pages/pyannote-audio into feat/joint-diarization-and-embedding
clement-pages
committed

Commits on Jun 20, 2023

Commits on Jun 21, 2023

Commits on Jun 22, 2023

improve(test): use pyannote.database.registry (pyannote#1413 )
hbredin
authored

Commits on Jun 23, 2023

Commits on Jun 27, 2023

Commits on Jun 28, 2023

Commits on Jul 3, 2023

add validation part into the multi-task
clement-pages
committed

Commits on Jul 4, 2023

Commits on Jul 5, 2023

simplify the way embedding loss is calculated
clement-pages
committed

Commits on Jul 6, 2023

handle case where there is no files from diarization dataset
clement-pages
committed

Commits on Jul 9, 2023

fix(doc): fix typo in diarization docstring
DiaaAj
authored

Commits on Jul 11, 2023

fix size issue in collate_y when building embedding ref
clement-pages
committed

Commits on Jul 18, 2023

Commits on Jul 19, 2023

add V2 of SpeakerEndToEndDiarization
clement-pages
committed

Commits on Jul 26, 2023

Commits on Oct 13, 2023

add prepare_data method in Task class
clement-pages
committed

Commits on Oct 26, 2023

Merge branch 'develop' into feat/data_preparation
hbredin
authored

Commits on Nov 2, 2023

modify organisation of pyannote segmentation tasks
clement-pages
committed
Merge branch 'feat/data_preparation' of github.com:clement-pages/pyannote-audio into feat/data_preparation
clement-pages
committed

Commits on Nov 7, 2023

add two training tests
clement-pages
committed
assign data directly to task in main process, in prepare_data
clement-pages
committed
Merge branch 'develop' into feat/data_preparation
hbredin
authored

Commits on Nov 8, 2023

handle call to Task.prepare_data and Task.setup under different scenarios
clement-pages
committed
Merge branch 'feat/data_preparation' of github.com:clement-pages/pyannote-audio into feat/data_preparation
clement-pages
committed

Commits on Nov 9, 2023

add training tests using task caches
clement-pages
committed
update cache_path type and docstrings
clement-pages
committed
fix classes variable used before assigment
clement-pages
committed

Commits on Nov 14, 2023

Merge branch 'develop' into feat/data_preparation
hbredin
authored

Commits on Nov 15, 2023

Merge branch 'feat/joint-diarization-and-embedding' into feat/joint-diarization-and-embedding-with-prepared-data
clement-pages
committed
fix: fix residual merge problems
clement-pages
committed

Commits on Nov 20, 2023

Commits on Nov 21, 2023

improve code readability
clement-pages
committed

Commits on Nov 27, 2023

improve: use numpy method for w/r task cache instead pickle (#1 )

clement-pages
and
clement-pages
authored
Merge branch 'pyannote:develop' into feat/data_preparation
clement-pages
authored

Commits on Nov 29, 2023

improve: remove complete redefinition of setup in joint task
clement-pages
committed
Merge branch 'feat/joint-diarization-and-embedding-with-prepared-data' of github.com:clement-pages/pyannote-audio into feat/joint-diarization-and-embedding-with-prepared-data
clement-pages
committed
Merge branch 'feat/data-preparation' into feat/joint-diarization-and-embedding-with-prepared-data
clement-pages
committed
improve: remove duplicated attributes in JointSpeakerDiarizationAndEmbedding
clement-pages
committed
update: replace old Task attributes with prepared_data in joint task
clement-pages
committed
improve: handle multi-speaker embeddings in example_output
clement-pages
committed

Commits on Nov 30, 2023

feat: add new end-to-end model for joint speaker diarization and embeddins
clement-pages
committed
fix: fix empty dict issue for metadata_unique_values in prepared_data
clement-pages
committed
improve: add dynamic typing for np array in prepare_data
clement-pages
committed
Merge branch 'feat/data-preparation' into feat/joint-diarization-and-embedding-with-prepared-data
clement-pages
committed

Commits on Dec 4, 2023

improve: check matching bewteen task current protocol and cached protocol
clement-pages
committed
remove: remove unused argument stage in Task.setup
clement-pages
committed
Merge branch 'feat/data-preparation' into feat/joint-diarization-and-embedding-with-prepared-data
clement-pages
committed

Commits on Dec 8, 2023

update: change name of attribute database_ratio to dia_task_rate
clement-pages
committed
wip: attempt to fix issues encountered during training
clement-pages
committed
update: use all the pyannet pretrained model
clement-pages
committed
fix: fix diarization loss calculation condition in training_step
clement-pages
committed

Commits on May 15, 2024

update joint task training_step
clement-pages
committed

Commits on May 27, 2024

fix(task): fiw wrong call to receptive_field in prepare_chunk
clement-pages
committed
Merge branch 'develop' into feat/joint-diarization-and-embedding-with-prepared-data
clement-pages
committed

Commits on May 28, 2024

update(joint task): filter out inactive speaker embeddings from loss computation
clement-pages
committed

Commits on Jun 21, 2024

allow to only compute mean or std in StatsPool
clement-pages
committed
update diarization + embeddings joint task
clement-pages
committed
wip: update joint model
clement-pages
committed
Merge branch 'develop' into feat/joint-diarization-and-embedding-with-prepared-data
clement-pages
authored

Commits on Jul 1, 2024

Merge branch 'pyannote:develop' into feat/joint-diarization-and-embedding-with-prepared-data
clement-pages
authored

Commits on Jul 8, 2024

wip: add pipeline working with joint model
hbredin
committed

Commits on Oct 18, 2024

Commits on Oct 25, 2024

wip: add validation pipeline
clement-pages
committed
clean validation pipeline code
clement-pages
committed

Commits on Oct 28, 2024

handle overlaped segmentation chunks corner case
clement-pages
committed
add some comments
clement-pages
committed

Commits on Nov 18, 2024

replace pyannote.metrics DER by pyannote.audio.torchmetrics one
clement-pages
committed
update joint pipeline
clement-pages
committed