Easy-to-use Speech Toolkit including SOTA/Streaming ASR with punctuation, influential TTS with text frontend, Speaker Verification System and End-to-End Speech Simultaneous Translation.

kws streaming-asr speech-alignment tts conformer speech-translation voice-recognition sound-classification transformer asr speech-synthesis voice-cloning punctuation-restoration streaming-tts speech-recognition vocoder

Go to file

Luo Tao de212572ed update unittest with comments		8 years ago
data_utils	add mfcc feature for DS2	8 years ago
datasets	Add shuffle type of instance_shuffle and batch_shuffle_clipped.	8 years ago
lm	upload the language model	8 years ago
tests	update unittest with comments	8 years ago
README.md	install libsndfile from /usr to thirdparty	8 years ago
compute_mean_std.py	add mfcc feature for DS2	8 years ago
decoder.py	follow comments in code format	8 years ago
error_rate.py	Follow comments.	8 years ago
evaluate.py	update several scripts to support mfcc	8 years ago
infer.py	update several scripts to support mfcc	8 years ago
model.py	Add function, class and module docs for data parts in DS2.	8 years ago
requirements.txt	install libsndfile from /usr to thirdparty	8 years ago
setup.sh	add soundfile read/write unitest	8 years ago
train.py	add mfcc feature for DS2	8 years ago
tune.py	update several scripts to support mfcc	8 years ago
utils.py	Add shuffle type of instance_shuffle and batch_shuffle_clipped.	8 years ago

README.md

Deep Speech 2 on PaddlePaddle

Installation

Please replace $PADDLE_INSTALL_DIR with your own paddle installation directory.

sh setup.sh
export LD_LIBRARY_PATH=$PADDLE_INSTALL_DIR/Paddle/third_party/install/warpctc/lib:$LD_LIBRARY_PATH

Usage

Preparing Data

cd datasets
sh run_all.sh
cd ..

sh run_all.sh prepares all ASR datasets (currently, only LibriSpeech available). After running, we have several summarization manifest files in json-format.

A manifest file summarizes a speech data set, with each line containing the meta data (i.e. audio filepath, transcript text, audio duration) of each audio file within the data set, in json format. Manifest file serves as an interface informing our system of where and what to read the speech samples.

More help for arguments:

python datasets/librispeech/librispeech.py --help

Preparing for Training

python compute_mean_std.py

It will compute mean and stdandard deviation for audio features, and save them to a file with a default name ./mean_std.npz. This file will be used in both training and inferencing. The default feature of audio data is power spectrum, and the mfcc feature is also supported. To train and infer based on mfcc feature, please generate this file by

python compute_mean_std.py --specgram_type mfcc

and specify --specgram_type mfcc when running train.py, infer.py, evaluator.py or tune.py.

More help for arguments:

python compute_mean_std.py --help

Training

For GPU Training:

CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 python train.py

For CPU Training:

python train.py --use_gpu False

More help for arguments:

python train.py --help

Preparing language model

The following steps, inference, parameters tuning and evaluating, will require a language model during decoding. A compressed language model is provided and can be accessed by

cd ./lm
sh run.sh
cd ..

Inference

For GPU inference

CUDA_VISIBLE_DEVICES=0 python infer.py

For CPU inference

python infer.py --use_gpu=False

More help for arguments:

python infer.py --help

Evaluating

CUDA_VISIBLE_DEVICES=0 python evaluate.py

More help for arguments:

python evaluate.py --help

Parameters tuning

Usually, the parameters \alpha and \beta for the CTC prefix beam search decoder need to be tuned after retraining the acoustic model.

For GPU tuning

CUDA_VISIBLE_DEVICES=0 python tune.py

For CPU tuning

python tune.py --use_gpu=False

More help for arguments:

python tune.py --help

Then reset parameters with the tuning result before inference or evaluating.