EveryVoiceTTS/EveryVoice

Pull Requests Commits

chore: fix configs

roedoejet committed 2 years ago

fda18e23

feat(cli): add test, train, tune, and preprocess commands to cli

roedoejet committed 2 years ago

ad2dd62a

fix: initialize output audio as none

roedoejet committed 2 years ago

4ab67b02

roedoejet committed 2 years ago

2e5d7c95

refactor: change to absolute imports and package structure

roedoejet committed 2 years ago

64d53100

refactor(train): change to lj as default

roedoejet committed 2 years ago

d0b7cd90

fix(preprocessor): move pyworld import down to avoid installation errors

roedoejet committed 2 years ago

eba9561f

fix(vocoder): fix weight normalization

roedoejet committed 2 years ago

efaefe32

fix: log datasets in experiment folder

roedoejet committed 2 years ago

43af9339

refactor(logging): add logging for all losses and fix step count for audio and images

roedoejet committed 2 years ago

86907b45

refactor: set lj as default task

roedoejet committed 2 years ago

20bf6207

fix(tests): update tests to new changes

roedoejet committed 2 years ago

d7a3ceae

feat(preprocessor): automatically process input and output sampling rates in preprocessor

roedoejet committed 2 years ago

b6a2b283

fix(model): fix audio logging sample rate

roedoejet committed 2 years ago

78c10efa

fix(dataloader): load segmented trainset and unsegmented valset

roedoejet committed 2 years ago

f5a443ed

feat(vocoder): allow for changing activation function as hyperparameter

roedoejet committed 2 years ago

2c33505e

fix: update audio sampling rate

roedoejet committed 2 years ago

c21584a3

feat(train): added checkpointing in train script

roedoejet committed 2 years ago

204233b3

fix(utils): fixed error in utils

roedoejet committed 2 years ago

c4982250

feat(lj): add full lj filelist

roedoejet committed 2 years ago

87db91f1

feat(upsampling): allow calculation of y_g_hat_mel to vary depending on upsampled sampling rate

roedoejet committed 2 years ago

c336a5d0

chore: merge plus added tests

roedoejet committed 2 years ago

3f2cdb1c

feat(vocoder): allow for depthwise separable convolutions

roedoejet committed 2 years ago

e2923e8a

refactor(config): allow for upsample sampling rate to differ from input sampling rate for vocoder

roedoejet committed 2 years ago

e22462dc

fix(preprocessor): fix exists typos and minor refactoring

roedoejet committed 2 years ago

bf76c260

fix: re-add tabulate package

roedoejet committed 2 years ago

21773efa

feat(tune): add autotune

roedoejet committed 2 years ago

50b7a2f9

refactor(vocoder): change logger and model

roedoejet committed 2 years ago

91a46836

refactor: removed librosa-based mel spec transform

roedoejet committed 2 years ago

7b39ec18

feat(vocoder): add spectral transform to vocoder

roedoejet committed 2 years ago

24c3b143

Newer Older