Commits

Breadcrumbs

History for

onexperimentalt5tts_s2s

User selector

All users

Datepicker

All time

Commit History

Commits on Feb 5, 2025

• added padding duration

  Show description for a1498d0
  paarthneekhara

  committed
  a1498d0

  Copy full SHA for a1498d0

• fixes

  Show description for 5798d79
  paarthneekhara

  committed
  5798d79

  Copy full SHA for 5798d79

Commits on Feb 4, 2025

• s2s

  Show description for b9f970e
  paarthneekhara

  committed
  b9f970e

  Copy full SHA for b9f970e

Commits on Feb 3, 2025

• scripts update

  Show description for 6eff698
  paarthneekhara

  committed
  6eff698

  Copy full SHA for 6eff698

Commits on Jan 7, 2025

• DPO notes and some script updates (#38)

  Show description for 6b8125f
  shehzeen

  authored
  6b8125f

  Copy full SHA for 6b8125f

Commits on Jan 6, 2025

• kv cache enable for learnablev2

  Show description for 64fa380
  paarthneekhara

  committed
  64fa380

  Copy full SHA for 64fa380

Commits on Jan 3, 2025

• minor updates in eval script

  Show description for 021ee3f
  paarthneekhara

  committed
  021ee3f

  Copy full SHA for 021ee3f

Commits on Dec 23, 2024

• val loss spike fix

  Show description for 4d51f7b
  paarthneekhara

  committed
  4d51f7b

  Copy full SHA for 4d51f7b

Commits on Dec 19, 2024

• move pos emb before weight initialization is applied

  Show description for 643a74d
  paarthneekhara

  committed
  643a74d

  Copy full SHA for 643a74d

• avoid multiple tensorboard logs for same experiment

  Show description for 2ded564
  paarthneekhara

  committed
  2ded564

  Copy full SHA for 2ded564

• decoder dropout fixes..

  Show description for 2da89fa
  paarthneekhara

  committed
  2da89fa

  Copy full SHA for 2da89fa

• Add Low Frame-rate Speech Codec++ modules (#37)

  Edresson

  authored
  1797913

  Copy full SHA for 1797913

• address roy's comments

  Show description for 41bfbb7
  paarthneekhara

  committed
  41bfbb7

  Copy full SHA for 41bfbb7

Commits on Dec 18, 2024

• fixed learnable position embedding implementation and added decoder input dropout option

  Show description for 2a4ed65
  paarthneekhara

  committed
  2a4ed65

  Copy full SHA for 2a4ed65

Commits on Dec 10, 2024

• Add TTS Lhotse dataset (#36)

  Show description for c98b6f2
  Edresson

  authored
  c98b6f2

  Copy full SHA for c98b6f2

Commits on Dec 9, 2024

• bug fix when using only text context manifests in validation

  Show description for 6fe9b94
  paarthneekhara

  committed
  6fe9b94

  Copy full SHA for 6fe9b94

• another missing explicit utf-8 encoding causing issue

  Show description for 0785de1
  paarthneekhara

  committed
  0785de1

  Copy full SHA for 0785de1

Commits on Dec 8, 2024

• fixed ipa g2p erroring on german herteronyms sometimes

  Show description for 8c49fbb
  paarthneekhara

  committed
  8c49fbb

  Copy full SHA for 8c49fbb

Commits on Dec 7, 2024

• bug fix in setup tokenizers

  Show description for c3e6940
  paarthneekhara

  committed
  c3e6940

  Copy full SHA for c3e6940

Commits on Dec 5, 2024

• bug fix cfg

  Show description for 39d1038
  paarthneekhara

  committed
  39d1038

  Copy full SHA for 39d1038

• CFG implementation training and inference (#34)

  Show description for c24a61b
  paarthneekhara

  authored
  c24a61b

  Copy full SHA for c24a61b

Commits on Dec 4, 2024

• cfg inference implementation without kv caching

  Show description for 6070a96
  paarthneekhara

  committed
  6070a96

  Copy full SHA for 6070a96

• removed redundant code in tokenizer

  Show description for fbffb23
  paarthneekhara

  committed
  fbffb23

  Copy full SHA for fbffb23

• Multilingual and sentence-piece tokenizers for T5-TTS (#33)

  Show description for e66777e
  shehzeen

  authored
  e66777e

  Copy full SHA for e66777e

Commits on Dec 2, 2024

• some updates in inference script

  Show description for e4638c5
  paarthneekhara

  committed
  e4638c5

  Copy full SHA for e4638c5

Commits on Dec 1, 2024

• added multi gpu multi node codec extraction script, some minor changes in dataloader setup

  Show description for cdd0894
  paarthneekhara

  committed
  cdd0894

  Copy full SHA for cdd0894

Commits on Nov 30, 2024

• removed duplicate code, infer, train and val step all using the same code to prepare context tensors

  Show description for 2345f9f
  paarthneekhara

  committed
  2345f9f

  Copy full SHA for 2345f9f

Commits on Nov 28, 2024

• added decoder pretraining mode

  Show description for 7fd604e
  paarthneekhara

  committed
  7fd604e

  Copy full SHA for 7fd604e

• bug fix

  Show description for 1a9d5b8
  paarthneekhara

  committed
  1a9d5b8

  Copy full SHA for 1a9d5b8

Commits on Nov 27, 2024

• bug fix for decoder context eos bos id mismatch between model and dataloader

  Show description for 522b12e
  paarthneekhara

  committed
  522b12e

  Copy full SHA for 522b12e

Commits on Nov 26, 2024

• some safeguards in dataloader

  Show description for 988b846
  paarthneekhara

  committed
  988b846

  Copy full SHA for 988b846

• T5TTS DPO Model (#32)

  Show description for c683fcd
  shehzeen

  authored
  c683fcd

  Copy full SHA for c683fcd

Commits on Nov 24, 2024

• added kv caching for inference, speeds up inference. probably more scope to speed up

  Show description for 3f66ae9
  paarthneekhara

  committed
  3f66ae9

  Copy full SHA for 3f66ae9

Commits on Nov 23, 2024

• support fix length context for decoder context models

  Show description for c2fa5f9
  paarthneekhara

  committed
  c2fa5f9

  Copy full SHA for c2fa5f9

• changed text context model to use same encoder as context audio, will break previous text context models

  Show description for 49a40df
  paarthneekhara

  committed
  49a40df

  Copy full SHA for 49a40df

Pagination

Previous

Next