-
Notifications
You must be signed in to change notification settings - Fork 2.3k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
TTS FastPitch Finetuning #2684
TTS FastPitch Finetuning #2684
Conversation
"from nemo.collections.tts.models import HifiGanModel\n", | ||
"from nemo.collections.tts.models import FastPitchModel\n", | ||
"\n", | ||
"hifigan_ckpt_path = \"/home/pneekhara/PreTrainedModels/HifiGan--val_loss=0.08-epoch=899.ckpt\"\n", |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
We probably also need to publish this model to NGC
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Yes, will need to do that.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Let's add a placeholder for the FastPitch checkpoint as well, and add a comment that we plan on releasing .nemo
files for this notebook soon.
" plt.show()" | ||
] | ||
}, | ||
{ |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I think it would also be cool to show how people can fine tune their model on other data. Maybe even use their own voice after recording a few samples.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Can move this to another PR
Signed-off-by: Paarth Neekhara <[email protected]>
Signed-off-by: Paarth Neekhara <[email protected]>
Signed-off-by: Paarth Neekhara <[email protected]>
Signed-off-by: Paarth Neekhara <[email protected]>
* Added filelists and finetuning code * image logging transoposed bug fix in FastPitch * pytorch lightning version requirement * added filelists for multispeaker hifigan * hifigan 44100 Hz * updated filelists, added another configuration of hifigan generator as per jason * added notebooks and some scripts * synthesize samples script update * removed sample filelists * reverted hifigan changes * reverting hifigan.yaml update Signed-off-by: Paarth Neekhara <[email protected]>
Signed-off-by: Paarth Neekhara <[email protected]>
Signed-off-by: Paarth Neekhara <[email protected]>
Signed-off-by: Paarth Neekhara <[email protected]>
Signed-off-by: Paarth Neekhara <[email protected]>
Signed-off-by: Paarth Neekhara <[email protected]>
Signed-off-by: Paarth Neekhara <[email protected]>
Signed-off-by: Paarth Neekhara <[email protected]>
…or fastpitch dataset, comments in configuration files Signed-off-by: Paarth Neekhara <[email protected]>
Signed-off-by: Paarth Neekhara <[email protected]>
Signed-off-by: Paarth Neekhara <[email protected]>
Signed-off-by: Paarth Neekhara <[email protected]>
Signed-off-by: Paarth Neekhara <[email protected]>
…le finetuning Signed-off-by: Paarth Neekhara <[email protected]>
Signed-off-by: Paarth Neekhara <[email protected]>
Signed-off-by: Paarth Neekhara <[email protected]>
* Added speaker in FastPitch2 dataloader Signed-off-by: Paarth Neekhara <[email protected]> * changed fastpitch_align.yaml to have configuration for 441000 Hz Signed-off-by: Paarth Neekhara <[email protected]> * trying to fix g2p Signed-off-by: Paarth Neekhara <[email protected]> * pytorch lightning version requirement Signed-off-by: Paarth Neekhara <[email protected]> * Cleanupfinetuning (NVIDIA#2) * Added filelists and finetuning code * image logging transoposed bug fix in FastPitch * pytorch lightning version requirement * added filelists for multispeaker hifigan * hifigan 44100 Hz * updated filelists, added another configuration of hifigan generator as per jason * added notebooks and some scripts * synthesize samples script update * removed sample filelists * reverted hifigan changes * reverting hifigan.yaml update Signed-off-by: Paarth Neekhara <[email protected]> * added finetuning notebook Signed-off-by: Paarth Neekhara <[email protected]> * removed fastpitch2.py (redundant with fastpitch.py) Signed-off-by: Paarth Neekhara <[email protected]> * restored old fastpitch_align.yaml, made just one finetuning yaml Signed-off-by: Paarth Neekhara <[email protected]> * reverting to old fastpitch align yaml Signed-off-by: Paarth Neekhara <[email protected]> * reverted to old vocabs.py Signed-off-by: Paarth Neekhara <[email protected]> * reverted to old vocabs.py Signed-off-by: Paarth Neekhara <[email protected]> * reverting requirements change Signed-off-by: Paarth Neekhara <[email protected]> * addressed pull request reviews -- updated notebook, speaker loading for fastpitch dataset, comments in configuration files Signed-off-by: Paarth Neekhara <[email protected]> * notebook update Signed-off-by: Paarth Neekhara <[email protected]> * removed redundant configuration file, updated notebook Signed-off-by: Paarth Neekhara <[email protected]> * some more corrections for switching to single configuration file Signed-off-by: Paarth Neekhara <[email protected]> * added url for hifigan dataset in notebook Signed-off-by: Paarth Neekhara <[email protected]> * warning messages if optimizer configuration does not look correct while finetuning Signed-off-by: Paarth Neekhara <[email protected]> * style error fix Signed-off-by: Paarth Neekhara <[email protected]> * dataloader fix after master merge Signed-off-by: Paarth Neekhara <[email protected]> Co-authored-by: Jason <[email protected]>
No description provided.