Code

v0.5.0#1027

Code

Merged

erogol merged 248 commits intomaincoqui-ai/TTS:mainfrom devcoqui-ai/TTS:devCopy head branch name to clipboardJan 3, 2022

+2,978-423

Commits

Commits on Dec 9, 2021

• Save speakers embeddings/ids before starting training

  

  Edresson

  authored and

  erogol

  committed
  8310d19

• Select randomly a speaker from the speaker manager for the test setences

  

  Edresson

  authored and

  erogol

  committed
  234a4aa

• Allow ignore speakers for all multispeaker datasets

  

  Edresson

  authored and

  erogol

  committed
  c9c1960

• Implement training support with d_vecs in the VITS model

  

  Edresson

  authored and

  erogol

  committed
  1efcccd

• Fix pylint issues

  

  Edresson

  authored and

  erogol

  committed
  c9f5838

• Implement multilingual dataloader support

  

  Edresson

  authored and

  erogol

  committed
  829ee55

• Add multilingual training support to the VITS model

  

  Edresson

  authored and

  erogol

  committed
  d0e3647

• Add multilingual inference support

  

  Edresson

  authored and

  erogol

  committed
  8e83a21

• Fix pylint issues

  

  Edresson

  authored and

  erogol

  committed
  32ece5d

• Fix bugs in the non-multilingual VITS inference

  

  Edresson

  authored and

  erogol

  committed
  240356c

• Fix bug in VITS multilingual inference

  

  Edresson

  authored and

  erogol

  committed
  56b5488

• Add voice conversion support for the model VITS trained with external speaker embedding

  

  Edresson

  authored and

  erogol

  committed
  d653227

• Fix the bug in M-AILABS formatter

  

  Edresson

  authored and

  erogol

  committed
  3df5d9a

• Implement vocoder Fine Tuning like SC-GlowTTS paper

  

  Edresson

  authored and

  erogol

  committed
  9071bf3

• Fix pylint issues

  

  Edresson

  authored and

  erogol

  committed
  cfa9910

• Fix unit tests

  

  Edresson

  authored and

  erogol

  committed
  82611cf

• Add VITS multilingual unit test

  

  Edresson

  authored and

  erogol

  committed
  859cf1b

• Add VITS d-vector unit test

  

  Edresson

  authored and

  erogol

  committed
  08da902

• Fix d-vector multispeaker training bug

  

  Edresson

  authored and

  erogol

  committed
  d7042ec

• Fix bug after merge

  

  Edresson

  authored and

  erogol

  committed
  f4abb19

• Fix the optimizer parameters bug in multilingual and multispeaker training

  

  Edresson

  authored and

  erogol

  committed
  256197b

• Active the multispeaker mode in multilingual training

  

  Edresson

  authored and

  erogol

  committed
  2bba769

• Add VITS multispeaker train unit test

  

  Edresson

  authored and

  erogol

  committed
  ecf327a

• Add the ValueError in the restore checkpoint exception to avoid problems with the optimizer restauration when new keys are addition

  

  Edresson

  authored and

  erogol

  committed
  fee01da

• Add H/ASP original checkpoint support

  

  Edresson

  authored and

  erogol

  committed
  a390103

• Add support to use the speaker encoder as loss function in VITS model

  

  Edresson

  authored and

  erogol

  committed
  3cd889a

• Added a notbook for d-vector multilingual VITS

  

  WeberJulian

  authored and

  erogol

  committed
  2be38aa

• Add voice conversion fine tuning mode

  

  Edresson

  authored and

  erogol

  committed
  cd7639c

• Update the VITS model docs

  

  Edresson

  authored and

  erogol

  committed
  5648036

• get_speaker_weighted_sampler

  

  WeberJulian

  authored and

  erogol

  committed
  9d2c445

• freeze vits parts

  

  WeberJulian

  authored and

  erogol

  committed
  de41165

• Add freeze vocoder generator and flow-based decoder option

  

  Edresson

  authored and

  erogol

  committed
  39aff66

• Add audio resample in the speaker consistency loss

  

  Edresson

  authored and

  erogol

  committed
  3ac4283

• Add remove silence VAD script

  

  Edresson

  authored and

  erogol

  committed
  c80cf67

• Merge dataset

  

  WeberJulian

  authored and

  erogol

  committed
  5c89803

• get_aux_input

  

  WeberJulian

  authored and

  erogol

  committed
  3440c54

• PitchExtractor

  

  WeberJulian

  authored and

  erogol

  committed
  ec83ffb

• fix collate_fn

  

  WeberJulian

  authored and

  erogol

  committed
  21b49c3

• fix test vits

  

  WeberJulian

  authored and

  erogol

  committed
  6ed55ba

• fix f0_cache_path in dataset

  

  WeberJulian

  authored and

  erogol

  committed
  20ac31d

• fix test sentence synthesis

  

  WeberJulian

  authored and

  erogol

  committed
  88d6399

• Add torchaudio in requirements.txt

  

  WeberJulian

  authored and

  erogol

  committed
  e1bdeac

• make style

  

  WeberJulian

  authored and

  erogol

  committed
  3e9ca4b

• fix linter

  

  WeberJulian

  authored and

  erogol

  committed
  215a74b

• fix phonemes per language

  

  WeberJulian

  authored and

  erogol

  committed
  686c738

• fix imports for load_meta_data

  

  WeberJulian

  authored and

  erogol

  committed
  b1df118

• Fix phonemes

  

  WeberJulian

  authored and

  erogol

  committed
  61251bd

• Fix continue path

  

  WeberJulian

  authored and

  erogol

  committed
  5f40e96

• Fix use_speaker_embedding logic

  

  WeberJulian

  authored and

  erogol

  committed
  e045775

• Fix small issues

  

  WeberJulian

  authored and

  erogol

  committed
  868cf64

• fix d-vector

  

  WeberJulian

  authored and

  erogol

  committed
  390096f

• Create a module for the VAD script

  

  Edresson

  authored and

  erogol

  committed
  89019d4

• Remove the unusable fine-tuning model

  

  Edresson

  authored and

  erogol

  committed
  5fc127b

• Turn more clear the VITS loss function

  

  Edresson

  authored and

  erogol

  committed
  86b2536

• Remove unusable speaker manager function

  

  Edresson

  authored and

  erogol

  committed
  ec31dac

• Remove the data from the set_d_vectors_from_file function

  

  Edresson

  authored and

  erogol

  committed
  cad82a9

• Remove the call to get_speaker_manager

  

  Edresson

  authored and

  erogol

  committed
  5ba416e

• Set the new_fields in copy_model_files as None by default

  

  Edresson

  authored and

  erogol

  committed
  d0cb700

• Fix the bug in extract tts spectrograms

  

  Edresson

  authored and

  erogol

  committed
  8ae2272

• Lint fixs

  

  Edresson

  authored and

  erogol

  committed
  9781e4d

• Fix function name

  

  Edresson

  authored and

  erogol

  committed
  1251d04

• remove inference notebook

  

  WeberJulian

  authored and

  erogol

  committed
  346bd93

• Rename ununsed_speakers to ignored_speakers

  

  Edresson

  authored and

  erogol

  committed
  aa1a070

• Fix the bug in multispeaker vits

  

  Edresson

  authored and

  erogol

  committed
  88f4369

• Get the number speaker from the Speaker Manager property

  

  Edresson

  authored and

  erogol

  committed
  5782df8

• Add the language embedding dim in the duration predictor class

  

  Edresson

  authored and

  erogol

  committed
  87059e3

• Add VITS multilingual d-vectors unit test

  

  Edresson

  authored and

  erogol

  committed
  65b5ed4

• Update docstring

  

  WeberJulian

  authored and

  erogol

  committed
  9d8d4e6

• Fix trailing whitespace

  

  WeberJulian

  authored and

  erogol

  committed
  36ddf32

• Remove torchaudio requeriment

  

  Edresson

  authored and

  erogol

  committed
  0359cab

• Remove self.audio_config from VITS

  

  WeberJulian

  authored and

  erogol

  committed
  93dbb67

• Fix pylint checks

  

  Edresson

  authored and

  erogol

  committed
  79f7592

• Fix tests

  

  WeberJulian

  authored and

  erogol

  committed
  9c1bec8

• Fix merge bug

  

  WeberJulian

  authored and

  erogol

  committed
  eff0a5c

• Make a multilingual test use chars

  

  WeberJulian

  authored and

  erogol

  committed
  4a8c344

• Add the SCL resample TODO

  

  Edresson

  authored and

  erogol

  committed
  be8f444

• Add find unique phonemes unit tests

  

  Edresson

  authored and

  erogol

  committed
  c382a3f

• Add remove silence vad script Unit test

  

  Edresson

  authored and

  erogol

  committed
  d79531c

• Add Docstring for TorchSTFT

  

  Edresson

  authored and

  erogol

  committed
  b909a3b

• Move multilingual logic out of the trainer

  

  WeberJulian

  authored and

  erogol

  committed
  352b4be

• Fix trailing space

  

  WeberJulian

  authored and

  erogol

  committed
  4001322

• Revert init multispeaker change

  

  WeberJulian

  authored and

  erogol

  committed
  0f64d45

• Prevent weighted sampler use when num_gpus > 1

  

  WeberJulian

  authored and

  erogol

  committed
  3f3505c

• Add test for language_weighted_sampler

  

  WeberJulian

  authored and

  erogol

  committed
  22c7be5

• Fix seed in test_samplers to avoid random fails

  

  WeberJulian

  authored and

  erogol

  committed
  7b81c16

• Add support for multi-lingual models in CLI

  

  WeberJulian

  authored and

  erogol

  committed
  4706583

• Add recipe for multi-lingual VITS

  

  WeberJulian

  authored and

  erogol

  committed
  846b4a1

• Fix zoo tests

  

  WeberJulian

  authored and

  erogol

  committed
  b4bb0ac

Commits on Dec 10, 2021

• Make style

  erogol

  committed
  66b6e9b

• Remove redundant code

  erogol

  committed
  bbea9b3

• Add 👑YourTTS docs

  erogol

  committed
  84bbe02

Commits on Dec 13, 2021

• Use torchaudio for ResNet speaker encoder

  erogol

  committed
  4b06e3e

• Fix synthesizer reading use_language_embedding

  erogol

  committed
  f73573c

• Rename setup_model to setup_speaker_encoder_model

  erogol

  committed
  6d7199d

• Fixup

  erogol

  committed
  6274d5e

• Fixup

  erogol

  committed
  9ec6238

Commits on Dec 16, 2021

• Add addtional datasets

  loganhart02

  committed
  103c010

• Add functions to get and check and argument in config and config.model_args

  erogol

  committed
  abedfd5

• Change speaker_idx to speaker_name

  erogol

  committed
  08a1cf3

• Use speaker_encoder from speaker manager in Vits

  erogol

  committed
  1ddf245

• Implement init_speaker_encoder and change arg names

  erogol

  committed
  95ca2ef

• Add torchaudio to requirements.txt

  erogol

  committed
  223ffe2

Commits on Dec 17, 2021

• Upload sdist and wheel to PyPI

  reuben

  committed
  7810085

• Include usage instructions in README

  reuben

  committed
  859ac1a

• Remove broken link

  reuben

  committed
  0057494

• Try to speed up CI install times

  reuben

  committed
  0635e8b

Commits on Dec 20, 2021

• Merge pull request #1022 from coqui-ai/papercuts

  Show description for 6b8ed18
  erogol

  authored
  6b8ed18

• Merge pull request #1021 from loganhart420/dataset_downloaders

  Show description for 3780346
  erogol

  authored
  3780346

• Save speakers embeddings/ids before starting training

  

  Edresson

  authored and

  erogol

  committed
  eb3e8af

• Select randomly a speaker from the speaker manager for the test setences

  

  Edresson

  authored and

  erogol

  committed
  e0ad838

• Allow ignore speakers for all multispeaker datasets

  

  Edresson

  authored and

  erogol

  committed
  6a7db67

• Implement training support with d_vecs in the VITS model

  

  Edresson

  authored and

  erogol

  committed
  d91c595

• Fix pylint issues

  

  Edresson

  authored and

  erogol

  committed
  5f1c181

• Implement multilingual dataloader support

  

  Edresson

  authored and

  erogol

  committed
  f996afe

• Add multilingual training support to the VITS model

  

  Edresson

  authored and

  erogol

  committed
  dcb2374

• Add multilingual inference support

  

  Edresson

  authored and

  erogol

  committed
  ac9416f

• Fix pylint issues

  

  Edresson

  authored and

  erogol

  committed
  3fbbebd

• Fix bugs in the non-multilingual VITS inference

  

  Edresson

  authored and

  erogol

  committed
  7c0b8ec

• Fix bug in VITS multilingual inference

  

  Edresson

  authored and

  erogol

  committed
  e997889

• Add voice conversion support for the model VITS trained with external speaker embedding

  

  Edresson

  authored and

  erogol

  committed
  c334d39

• Fix the bug in M-AILABS formatter

  

  Edresson

  authored and

  erogol

  committed
  f1f0163

• Implement vocoder Fine Tuning like SC-GlowTTS paper

  

  Edresson

  authored and

  erogol

  committed
  c53693c

• Fix pylint issues

  

  Edresson

  authored and

  erogol

  committed
  36dcd11

• Fix unit tests

  

  Edresson

  authored and

  erogol

  committed
  7ef3ddc

• Add VITS multilingual unit test

  

  Edresson

  authored and

  erogol

  committed
  959cc8f

• Add VITS d-vector unit test

  

  Edresson

  authored and

  erogol

  committed
  e68b042

• Fix d-vector multispeaker training bug

  

  Edresson

  authored and

  erogol

  committed
  76251b6

• Fix bug after merge

  

  Edresson

  authored and

  erogol

  committed
  9be5b75

• Fix the optimizer parameters bug in multilingual and multispeaker training

  

  Edresson

  authored and

  erogol

  committed
  de78556

• Active the multispeaker mode in multilingual training

  

  Edresson

  authored and

  erogol

  committed
  92f7f4f

• Add VITS multispeaker train unit test

  

  Edresson

  authored and

  erogol

  committed
  bbdb5c3

• Add the ValueError in the restore checkpoint exception to avoid problems with the optimizer restauration when new keys are addition

  

  Edresson

  authored and

  erogol

  committed
  0bdfd3c

• Add H/ASP original checkpoint support

  

  Edresson

  authored and

  erogol

  committed
  9b011b1

• Add support to use the speaker encoder as loss function in VITS model

  

  Edresson

  authored and

  erogol

  committed
  690b37d

• Added a notbook for d-vector multilingual VITS

  

  WeberJulian

  authored and

  erogol

  committed
  6ac31e4

• Add voice conversion fine tuning mode

  

  Edresson

  authored and

  erogol

  committed
  eeb8ac0

• Update the VITS model docs

  

  Edresson

  authored and

  erogol

  committed
  9de4539

• get_speaker_weighted_sampler

  

  WeberJulian

  authored and

  erogol

  committed
  005bba6

• freeze vits parts

  

  WeberJulian

  authored and

  erogol

  committed
  2b952d8

• Add freeze vocoder generator and flow-based decoder option

  

  Edresson

  authored and

  erogol

  committed
  1c6bcda

• Add audio resample in the speaker consistency loss

  

  Edresson

  authored and

  erogol

  committed
  1bd1a05

• Add remove silence VAD script

  

  Edresson

  authored and

  erogol

  committed
  10ff90d

• Merge dataset

  

  WeberJulian

  authored and

  erogol

  committed
  b3abd01

• get_aux_input

  

  WeberJulian

  authored and

  erogol

  committed
  9a2f913

• PitchExtractor

  

  WeberJulian

  authored and

  erogol

  committed
  78c2d12

• fix collate_fn

  

  WeberJulian

  authored and

  erogol

  committed
  2a2b576

• fix test vits

  

  WeberJulian

  authored and

  erogol

  committed
  3b5592a

• fix f0_cache_path in dataset

  

  WeberJulian

  authored and

  erogol

  committed
  0804806

• fix test sentence synthesis

  

  WeberJulian

  authored and

  erogol

  committed
  4d721bc

• Add torchaudio in requirements.txt

  

  WeberJulian

  authored and

  erogol

  committed
  e22f7a2

• make style

  

  WeberJulian

  authored and

  erogol

  committed
  1472b6d

• fix linter

  

  WeberJulian

  authored and

  erogol

  committed
  e995a63

• fix phonemes per language

  

  WeberJulian

  authored and

  erogol

  committed
  1340938

• fix imports for load_meta_data

  

  WeberJulian

  authored and

  erogol

  committed
  846bf16

• Fix phonemes

  

  WeberJulian

  authored and

  erogol

  committed
  120332d

• Fix continue path

  

  WeberJulian

  authored and

  erogol

  committed
  23d789c

• Fix use_speaker_embedding logic

  

  WeberJulian

  authored and

  erogol

  committed
  e8af6a9

• Fix small issues

  

  WeberJulian

  authored and

  erogol

  committed
  da6c1e8

• fix d-vector

  

  WeberJulian

  authored and

  erogol

  committed
  631addf

• Create a module for the VAD script

  

  Edresson

  authored and

  erogol

  committed
  352aa69

• Remove the unusable fine-tuning model

  

  Edresson

  authored and

  erogol

  committed
  6fc3b9e

• Turn more clear the VITS loss function

  

  Edresson

  authored and

  erogol

  committed
  8c22d5a

• Remove unusable speaker manager function

  

  Edresson

  authored and

  erogol

  committed
  9daa33d

• Remove the data from the set_d_vectors_from_file function

  

  Edresson

  authored and

  erogol

  committed
  b769b49

• Remove the call to get_speaker_manager

  

  Edresson

  authored and

  erogol

  committed
  34749f8

• Set the new_fields in copy_model_files as None by default

  

  Edresson

  authored and

  erogol

  committed
  2b2ceca

• Fix the bug in extract tts spectrograms

  

  Edresson

  authored and

  erogol

  committed
  85418ff

• Lint fixs

  

  Edresson

  authored and

  erogol

  committed
  45d0b04

• Fix function name

  

  Edresson

  authored and

  erogol

  committed
  f34596d

• remove inference notebook

  

  WeberJulian

  authored and

  erogol

  committed
  9ca3af9

• Rename ununsed_speakers to ignored_speakers

  

  Edresson

  authored and

  erogol

  committed
  90eac13

• Fix the bug in multispeaker vits

  

  Edresson

  authored and

  erogol

  committed
  f394d60

• Get the number speaker from the Speaker Manager property

  

  Edresson

  authored and

  erogol

  committed
  4196a42

• Add the language embedding dim in the duration predictor class

  

  Edresson

  authored and

  erogol

  committed
  1296853

• Add VITS multilingual d-vectors unit test

  

  Edresson

  authored and

  erogol

  committed
  06d89f9

• Update docstring

  

  WeberJulian

  authored and

  erogol

  committed
  ffc269e

• Fix trailing whitespace

  

  WeberJulian

  authored and

  erogol

  committed
  2e51686

• Remove torchaudio requeriment

  

  Edresson

  authored and

  erogol

  committed
  d39200e

• Remove self.audio_config from VITS

  

  WeberJulian

  authored and

  erogol

  committed
  4cd0e4e

• Fix pylint checks

  

  Edresson

  authored and

  erogol

  committed
  78a23e1

• Fix tests

  

  WeberJulian

  authored and

  erogol

  committed
  09eda31

• Fix merge bug

  

  WeberJulian

  authored and

  erogol

  committed
  8b52fb8

• Make a multilingual test use chars

  

  WeberJulian

  authored and

  erogol

  committed
  54e33bf

• Add the SCL resample TODO

  

  Edresson

  authored and

  erogol

  committed
  67dda0a

• Add find unique phonemes unit tests

  

  Edresson

  authored and

  erogol

  committed
  e068fab

• Add remove silence vad script Unit test

  

  Edresson

  authored and

  erogol

  committed
  a57ddfb

• Add Docstring for TorchSTFT

  

  Edresson

  authored and

  erogol

  committed
  818dc4c

• Move multilingual logic out of the trainer

  

  WeberJulian

  authored and

  erogol

  committed
  6b03943

• Fix trailing space

  

  WeberJulian

  authored and

  erogol

  committed
  9cfbacc

• Revert init multispeaker change

  

  WeberJulian

  authored and

  erogol

  committed
  74cedfa

• Prevent weighted sampler use when num_gpus > 1

  

  WeberJulian

  authored and

  erogol

  committed
  2bbcb55

• Add test for language_weighted_sampler

  

  WeberJulian

  authored and

  erogol

  committed
  6f01eed

• Fix seed in test_samplers to avoid random fails

  

  WeberJulian

  authored and

  erogol

  committed
  8b3769c

• Add support for multi-lingual models in CLI

  

  WeberJulian

  authored and

  erogol

  committed
  a564eb9

• Add recipe for multi-lingual VITS

  

  WeberJulian

  authored and

  erogol

  committed
  6700bb1

• Fix zoo tests

  

  WeberJulian

  authored and

  erogol

  committed
  54b7fb4

• Make style

  erogol

  committed
  704dddc

• Remove redundant code

  erogol

  committed
  649dc9e

• Add 👑YourTTS docs

  erogol

  committed
  45b2f8e

• Use torchaudio for ResNet speaker encoder

  erogol

  committed
  7a987db

• Fix synthesizer reading use_language_embedding

  erogol

  committed
  35a781f

• Rename setup_model to setup_speaker_encoder_model

  erogol

  committed
  79de38c

• Fixup

  erogol

  committed
  3818bd0

• Fixup

  erogol

  committed
  3c6d7f4

• Add functions to get and check and argument in config and config.model_args

  erogol

  committed
  4c50f6f

• Change speaker_idx to speaker_name

  erogol

  committed
  4d13b88

• Use speaker_encoder from speaker manager in Vits

  erogol

  committed
  d29c378

• Implement init_speaker_encoder and change arg names

  erogol

  committed
  473414d

• Add torchaudio to requirements.txt

  erogol

  committed
  e57c117

• Remove commented code

  erogol

  committed
  a25269d

• Add more listing options to ModelManager

  erogol

  committed
  f769595

Commits on Dec 21, 2021

• Fix multi-speaker init in Synthesizer

  erogol

  committed
  c9c1fa0

• Merge branch 'pr/Edresson/731-rebased' of https://github.com/coqui-ai/TTS into pr/Edresson/731-rebased

  erogol

  committed
  cc514b3

• Fix speaker encoder init

  erogol

  committed
  56378b1

• Add UK vocoder models

  erogol

  committed
  8d2bb28

Commits on Dec 22, 2021

• Implement RMS volume normalization

  erogol

  committed
  633dcc9

• Init speaker manager for speaker encoder

  erogol

  committed
  5c5ddd2

• Update .gitignore

  erogol

  committed
  55ce7f0

Commits on Dec 23, 2021

• Update bug template

  erogol

  committed
  9ff0bfb

• Fix #1037

  erogol

  committed
  275c759

Commits on Dec 29, 2021

• Add VITS model tests

  erogol

  committed
  2033e17

Commits on Dec 30, 2021

• Fix Training HiFiGan -- avg loss not decreasing #1003

  erogol

  committed
  6189fdf

• Update Speaker Encoder models

  erogol

  committed
  638091f

• Update VITS model

  erogol

  committed
  7129b04

• Add custom asserts to tests

  erogol

  committed
  497332b

• Fix speaker encoder test

  erogol

  committed
  348b5c9

• Fix resnet speaker encoder

  erogol

  committed
  36cef59

Commits on Dec 31, 2021

• Add the YourTTS entry to the models

  erogol

  committed
  8100135

• Fix your_tts inference from the listed models

  erogol

  committed
  61874bc

Commits on Jan 1, 2022

• Print testing model

  erogol

  committed
  254c110

• Print urls when BadZipError

  erogol

  committed
  8fd1ee1

• Update yourTTS url

  erogol

  committed
  33711af

• Merge branch 'pr/Edresson/731-rebased' into dev

  erogol

  committed
  d37cfe4

• Update uk vocoder url

  erogol

  committed
  c5512af

• Merge branch 'dev' of https://github.com/coqui-ai/TTS into dev

  erogol

  committed
  38f5a11

• Make linter

  erogol

  committed
  e55f5ee

• Updake default vocoder for uk model

  erogol

  committed
  7ef458a

• Fix phoneme language

  WeberJulian

  committed
  a63998c

• Fix VITS multispeaker recipe

  WeberJulian

  committed
  9d0a732

Commits on Jan 2, 2022

• Merge pull request #1044 from WeberJulian/dev

  Show description for d82fa05
  erogol

  authored
  d82fa05

• Fix language assignment

  erogol

  committed
  d724984

• Merge pull request #1047 from coqui-ai/fix_lang_assign

  Show description for b0d6883
  erogol

  authored
  b0d6883

Commits on Jan 3, 2022

• Fix if else statement

  erogol

  committed
  7fad969

• Prioritize the given encoder path over config

  erogol

  committed
  fc09e31

• Merge pull request #1050 from coqui-ai/fix_synthesizer_init

  Show description for 03bcae1
  erogol

  authored
  03bcae1

• Keep proj_dim in speaker encoder models

  erogol

  committed
  5840d89

• Bump up to v0.5.0

  erogol

  committed
  58c38de

• Fix train_tts.py and uncomment code (#1051)

  Show description for e1accb6
  WeberJulian

  authored
  e1accb6