Implementation of DeepSpeech2 using Baidu Warp-CTC. Creates a network based on the DeepSpeech2 architecture, trained with the CTC activation function.

Several libraries are needed to be installed for training to work. I will assume that everything is being installed in an Anaconda installation on Ubuntu.

Install PyTorch if you haven't already.

Install this fork for Warp-CTC bindings:

git clone https://github.com/SeanNaren/warp-ctc.git
cd warp-ctc
mkdir build; cd build
cmake ..
make
export CUDA_HOME="/usr/local/cuda"
cd pytorch_binding
python setup.py install

Finally:

pip install -r requirements.txt

Currently supports AN4, TEDLIUM and Voxforge.

To download and setup the an4 dataset run below command in the root folder of the repo:

cd data; python an4.py

You have the option to download the raw dataset file manually or through the script (which will cache it). The file is found here.

To download and setup the TEDLIUM_V2 dataset run below command in the root folder of the repo:

cd data; python ted.py # Optionally if you have downloaded the raw dataset file, pass --tar_path /path/to/TEDLIUM_release2.tar.gz

To download and setup the Voxforge dataset run below command in the root folder of the repo:

cd data; python voxforge.py

Note that this dataset does not come with a validation dataset or test dataset.

To create a custom dataset you must create a CSV file containing the locations of the training data. This has to be in the format of:

/path/to/audio.wav,/path/to/text.txt
/path/to/audio2.wav,/path/to/text2.txt
...

The first path is to the audio file, and the second path is to a text file containing the transcript on one line. This can then be used as stated below.

python train.py --train_manifest data/train_manifest.csv --val_manifest data/val_manifest.csv

Use python train.py --help for more parameters and options.

There is also Visdom support to visualise training. Once a server has been started, to use:

python train.py --visdom true