[TTS] Add tutorial for TTS data prep scripts #6922
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
rlangman
requested review from
XuesongYang,
redoctopus,
racoiaws and
subhankar-ghosh
redoctopus
reviewed
Jun 30, 2023
XuesongYang
previously approved these changes
Jul 8, 2023
| "source": [ | ||
| "In this tutorial, we will prepare a dataset using our [TTS Dataset Processing Scripts](https://github.com/NVIDIA/NeMo/tree/main/scripts/dataset_processing/tts) and use it for training a FastPitch model.\n", | ||
| "\n", | ||
| "**This tutorial uses a different workflow than all other existing TTS tutorials. The scripts and classes used are all experimental and not yet ready for production**" |
There was a problem hiding this comment.
nit: missing a period at the end of the sentence.
| { | ||
| "cell_type": "markdown", | ||
| "source": [ | ||
| "# Dataset Prepration" |
There was a problem hiding this comment.
s/Prepration/Preparation/
| "cell_type": "code", | ||
| "source": [ | ||
| "import IPython.display as ipd\n", | ||
| "from matplotlib.pyplot import imshow" |
There was a problem hiding this comment.
remove since no other places use it.
| "source": [ | ||
| "We can use [create_speaker_map.py](https://github.com/NVIDIA/NeMo/blob/main/scripts/dataset_processing/tts/create_speaker_map.py) to easily create a mapping from speaker ID strings to integer indices that will be used at training time.\n", | ||
| "\n", | ||
| "The script will simply sort the speaker IDs and assign them numbers [0, num_speakers) in alphabetical order." |
There was a problem hiding this comment.
s/[0, num_speakers)/[0, num_speakers)/
| { | ||
| "cell_type": "markdown", | ||
| "source": [ | ||
| "Before training FastPitch, we need to compute some features for every audio file. The default [config file](https://github.com/NVIDIA/NeMo/blob/main/examples/tts/conf/feature/feature_44100.yaml) we will use has parameters for computing the **pitch** and **energy** of every audio frame." |
There was a problem hiding this comment.
nit: I saw inconsistent font formats for pitch and energy, and sometimes pitch and energy.
There was a problem hiding this comment.
I went through the tutorial to try to the formatting more consistent.
- Use bold when it is the first time an important vocab term is mentioned.
- Use
codewhen it refers to specific code, variable name, file, etc. - Use italics to emphasize any other key words.
| "For training it is beneficial for us to *normalize* our features. The most standard approach is to apply **mean-variance normalization** so that each feature has a mean of 0 and variance of 1. To do this we need to compute the *dataset statistics* with the mean and variance of each feature.\n", | ||
| "\n", | ||
| "For TTS it also helps\n", | ||
| "* Normalize features using speaker-level statistics\n", |
Signed-off-by: Ryan <[email protected]>
Signed-off-by: Ryan <[email protected]>
rlangman
force-pushed
the
tts_tutorial
branch
2 times, most recently
from
207a14f to
c3c0b5b
Compare
Signed-off-by: Ryan <[email protected]>
XuesongYang
approved these changes
Jul 12, 2023
XuesongYang
approved these changes
Jul 12, 2023
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
What does this PR do ?
Add a tutorial demonstrating how to do the end to end data preparation and training with the new TTS preprocessing scripts and data loader.
Collection: [TTS]
Changelog
Before your PR is "Ready for review"
Pre checks:
PR Type: