Skip to content

LWprogramming/audiolm-pytorch-training

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

audiolm-pytorch-training

This repository contains my scripts to train AudioLM models, using the audiolm-pytorch library by lucidrains and the my own modifications in the personal_hacks branch. The AudioLM model is a mostly 1-1 reproduction of the paper "AudioLM: a Language Modeling Approach to Audio Generation".

Getting started

# Create necessary directories
export folder_name=itsleonwu # rename this for your own purposes
mkdir /fsx/$folder_name 
cd /fsx/$folder_name 
mkdir audiolm-pytorch-results
mkdir audiolm-pytorch-datasets

# Clone the audiolm-pytorch-training repository
git clone https://github.com/LWprogramming/audiolm-pytorch-training.git

# Create a virtual environment using Python 3.10 and activate it
cd audiolm-pytorch-training
python3.10 -m venv venv
source venv/bin/activate

# Run the hubert_ckpt_download.py script
python hubert_ckpt_download.py

# Run the use_patched_audiolm.py script
python use_patched_audiolm.py personal_hacks

pip install tensorboardX # for some reason not covered in the previous pip install?? installing separately

# Download the dataset
cd ../audiolm-pytorch-datasets
wget https://www.openslr.org/resources/12/dev-clean.tar.gz
tar -xvf dev-clean.tar.gz
mv LibriSpeech LibriSpeech-dev-clean
rm dev-clean.tar.gz

# Create a directory for the sample file
mkdir many_identical_copies_of_cocochorales_single_sample_resampled_24kHz_trimmed_first_second
echo "Remember to upload the sample file to this overfitting dataset!"

About

audiolm-pytorch training code

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published