Code

Updated Extraction of Embeddings#1

Code

Open

ilanit1997 wants to merge 77 commits intoisaacOnline:extract-embeddingsisaacOnline/whisper:extract-embeddingsfrom ilanit1997:extract_embeddings_updatedilanit1997/whisper:extract_embeddings_updatedCopy head branch name to clipboard

+106,801-3,048

Commits

Commits on May 16, 2023

• Fix bug (openai#305)

  Show description for 207a46a
  

  MichaelMonashev

  authored and

  ilanit1997

  committed
  207a46a

• Add package metadata to setup.py (openai#315)

  Show description for 1efaa73
  

  brainwane

  authored and

  ilanit1997

  committed
  1efaa73

• Fix attention caching to make it actually work (openai#370)

  

  vickianand

  authored and

  ilanit1997

  committed
  4117a3a

• suppress generating non-timestamp tokens at the beginning (openai#532)

  

  jumon

  authored and

  ilanit1997

  committed
  77c4d6c

• fix to return only the text token ids

  

  jongwook

  authored and

  ilanit1997

  committed
  3c18acf

• invoking __call__ instead of forward()

  

  jongwook

  authored and

  ilanit1997

  committed
  f20ea99

• fix compression ratio function (openai#561)

  

  jumon

  authored and

  ilanit1997

  committed
  ef25ace

• add large-v2 model

  Show description for 7cc3396
  

  jongwook

  authored and

  ilanit1997

  committed
  7cc3396

• Explicitly closing model file after reading it (openai#630)

  

  paulharter

  authored and

  ilanit1997

  committed
  792e1f0

• Update Hebrew language code to he per IANA registry (openai#401)

  Show description for 678125e
  

  altryne

  authored and

  ilanit1997

  committed
  678125e

• large-v2 figure and arxiv url update

  

  jongwook

  authored and

  ilanit1997

  committed
  69d2a4d

• saving the qk matrix in the attention module for convenience

  

  jongwook

  authored and

  ilanit1997

  committed
  df00c4b

• Revert "saving the qk matrix in the attention module for convenience"

  Show description for 6f602de
  

  jongwook

  authored and

  ilanit1997

  committed
  6f602de

• MultiHeadAttention to return qk as well

  

  jongwook

  authored and

  ilanit1997

  committed
  24185b3

• word-level timestamps in Multilingual_ASR notebook

  

  jongwook

  authored and

  ilanit1997

  committed
  939d1d3

• torch.concatenate -> torch.cat for compatibility

  

  jongwook

  authored and

  ilanit1997

  committed
  9116fdd

• Fix tiny transcribe() docstring typo (openai#857)

  Show description for 8089c63
  

  adamreis

  authored and

  ilanit1997

  committed
  8089c63

• Support batch-dimension in log_mel_spectogram (openai#839)

  

  HennerM

  authored and

  ilanit1997

  committed
  1b23bc3

• Update README.md (openai#804)

  

  mikkovedru

  authored and

  ilanit1997

  committed
  7ac309b

• add github action to run pytest

  

  jongwook

  authored and

  ilanit1997

  committed
  b7535b3

• allow test_transcribe to run on CPU when CUDA is not available

  

  jongwook

  authored and

  ilanit1997

  committed
  4844fc8

• rename GitHub workflow

  

  jongwook

  authored and

  ilanit1997

  committed
  abd105e

• Use ndimage.median_filter instead of signal.medfilter (openai#812)

  Show description for 3021ec0
  

  3 peopleauthored and

  ilanit1997

  committed
  3021ec0

• Add github action to automatically push to pypi on Release x.y.z commit (openai#681)

  Show description for a6ec1a9
  

  2 peopleauthored and

  ilanit1997

  committed
  a6ec1a9

• Release 20230117

  

  jongwook

  authored and

  ilanit1997

  committed
  ca9444b

• print '?' if a letter can't be encoded using the system default encoding (openai#859)

  

  jongwook

  authored and

  ilanit1997

  committed
  141204b

• verbose outputs from pytest

  

  jongwook

  authored and

  ilanit1997

  committed
  8360160

• Fix bug where mm is mistakenly replaced with hmm in e.g. 20mm (openai#659)

  Show description for 5d304e4
  

  2 peopleauthored and

  ilanit1997

  committed
  5d304e4

• use stdout for printing transcription progress (openai#867)

  

  jongwook

  authored and

  ilanit1997

  committed
  91d56d2

• Handle XDG_CACHE_HOME properly for download_root (openai#864)

  Show description for 2990115
  

  2 peopleauthored and

  ilanit1997

  committed
  2990115

• Added --output_format option (openai#333)

  Show description for 91519d4
  

  2 peopleauthored and

  ilanit1997

  committed
  91519d4

• Add TSV formatted output in transcript, using integer start/end times in milliseconds. (openai#228)

  Show description for ff3c02b
  

  2 peopleauthored and

  ilanit1997

  committed
  ff3c02b

• handle printing even if sys.stdout.buffer is not available (openai#887)

  

  jongwook

  authored and

  ilanit1997

  committed
  172ee4c

• Release 20230124

  

  jongwook

  authored and

  ilanit1997

  committed
  00cc3cc

• drop python 3.7 support (openai#889)

  

  jongwook

  authored and

  ilanit1997

  committed
  7a3114b

• Update README.md about Python 3.8+ requirement

  

  jongwook

  authored and

  ilanit1997

  committed
  9f4deb6

• clarify that 3.11 is not supported

  

  jongwook

  authored and

  ilanit1997

  committed
  387da57

• Fix infinite loop caused by incorrect timestamp tokens prediction (openai#914)

  Show description for 1d56e13
  

  2 peopleauthored and

  ilanit1997

  committed
  1d56e13

• Update README.md (openai#894)

  Show description for 368acfe
  

  2 peopleauthored and

  ilanit1997

  committed
  368acfe

• Decoding improvements (openai#1033)

  Show description for f5040f9
  

  jongwook

  authored and

  ilanit1997

  committed
  f5040f9

• word-level timestamps in transcribe() (openai#869)

  Show description for f3f3234
  

  2 peopleauthored and

  ilanit1997

  committed
  f3f3234

• apply formatting with black (openai#1038)

  Show description for 52dca2c
  

  jongwook

  authored and

  ilanit1997

  committed
  52dca2c

• remove auxiliary audio extension (openai#1021)

  Show description for e708fc5
  

  2 peopleauthored and

  ilanit1997

  committed
  e708fc5

• Release 20230306

  

  jongwook

  authored and

  ilanit1997

  committed
  0ae70cd

• Update setup.py

  

  jongwook

  authored and

  ilanit1997

  committed
  2689b60

• Try installing triton only if linux & x86_64 (openai#1051)

  

  jongwook

  authored and

  ilanit1997

  committed
  77a0a3f

• Use triton==2.0.0 (openai#1053)

  

  jongwook

  authored and

  ilanit1997

  committed
  07c4b0b

• attempt to fix the repetition/hallucination issue identified in openai#1046 (openai#1052)

  Show description for 7b8f753
  

  jongwook

  authored and

  ilanit1997

  committed
  7b8f753

• Release 20230307

  

  jongwook

  authored and

  ilanit1997

  committed
  0aa8b79

• fix typo

  

  jongwook

  authored and

  ilanit1997

  committed
  86d33b6

• fix all_tokens handling that caused more repetitions and discrepancy in JSON (openai#1060)

  

  jongwook

  authored and

  ilanit1997

  committed
  88d2811

• kwargs in decode() for convenience (openai#1061)

  Show description for 5c8405c
  

  jongwook

  authored and

  ilanit1997

  committed
  5c8405c

• Release 20230308

  

  jongwook

  authored and

  ilanit1997

  committed
  f433b12

• Use tiktoken (openai#1044)

  Show description for 25eea90
  

  2 peopleauthored and

  ilanit1997

  committed
  25eea90

• Fix alignment between the segments and the list of words (openai#1087)

  Show description for c15c065
  

  guillaumekln

  authored and

  ilanit1997

  committed
  c15c065

• fix github language stats getting dominated by jupyter notebook (openai#1076)

  Show description for 05a8312
  

  3 peopleauthored and

  ilanit1997

  committed
  05a8312

• Fix truncated words list when the replacement character is decoded (openai#1089)

  

  guillaumekln

  authored and

  ilanit1997

  committed
  bf1873f

• abort find_alignment on empty input (openai#1090)

  

  jongwook

  authored and

  ilanit1997

  committed
  11242f3

• Release 20230314

  

  jongwook

  authored and

  ilanit1997

  committed
  80ff7bf

• Update tokenizer.py (openai#1163)

  

  jongwook

  authored and

  ilanit1997

  committed
  ec42661

• python-publish.yml: bump actions version to fix node warning (openai#1211)

  

  kbdharun

  authored and

  ilanit1997

  committed
  4e7a9ad

• Squash long words at window and sentence boundaries. (openai#1114)

  Show description for d390815
  

  2 peopleauthored and

  ilanit1997

  committed
  d390815

• Implement max line width and max line count, and make word highlighting optional (openai#1184)

  Show description for 819074f
  

  2 peopleauthored and

  ilanit1997

  committed
  819074f

• Update README.md to reference tiktoken (openai#1105)

  Show description for 2974581
  

  2 peopleauthored and

  ilanit1997

  committed
  2974581

• Update decoding.py (openai#1155)

  Show description for 4586341
  

  2 peopleauthored and

  ilanit1997

  committed
  4586341

• Update decoding.py (openai#1219)

  

  jongwook

  authored and

  ilanit1997

  committed
  dffc84f

• Python 3.11 (openai#1171)

  Show description for 27784ec
  

  2 peopleauthored and

  ilanit1997

  committed
  27784ec

• Drop ffmpeg-python dependency and call ffmpeg directly. (openai#1242)

  Show description for b045802
  

  2 peopleauthored and

  ilanit1997

  committed
  b045802

• Dropped unused execute bit from mel_filters.npz. (openai#1254)

  

  petterreinholdtsen

  authored and

  ilanit1997

  committed
  07b6909

• Avoid computing higher temperatures on no_speech segments (openai#1279)

  Show description for 792f612
  

  2 peopleauthored and

  ilanit1997

  committed
  792f612

• Updated README.md to provide more insight on BLEU and specific appendices (openai#1236)

  Show description for 4faca12
  

  2 peopleauthored and

  ilanit1997

  committed
  4faca12

• Fix numba depreceation notice (openai#1233)

  Show description for 3dd982e
  

  m3at

  authored and

  ilanit1997

  committed
  3dd982e

• fix condition_on_previous_text (openai#1224)

  Show description for 8099c2e
  

  funboarder13920

  authored and

  ilanit1997

  committed
  8099c2e

• add whisper clones with change to allow save of encoder and decoder embeddings + change pipeline_transcriptions.py to allow save of embedding as npy arrays

  ilanit1997

  committed
  49a0986

• update branch with most recent files from whisper and apply changes to allow embedding extraction

  ilanit1997

  committed
  679584e

• delete local files

  ilanit1997

  committed
  ed80c0a

Commits on May 17, 2023

• fix indentation bug

  ilanit1997

  committed
  94d0e30