k2-fsa/sherpa

Stars
539
Rank 82,402 (Top 2 %)
Language
C++
License
Apache License 2.0
Created over 2 years ago
Updated about 1 month ago

k2-fsa/sherpa

k2-fsa

There are no reviews yet. Be the first to send feedback to the community and the maintainers!

Speech-to-text server framework with next-gen Kaldi

Try sherpa from within your browser without installing anything: https://huggingface.co/spaces/k2-fsa/automatic-speech-recognition

See https://k2-fsa.github.io/sherpa/python/huggingface/ for more details.

sherpa

sherpa is an open-source speech-to-text (i.e., speech recognition) framework, focusing exclusively on end-to-end (E2E) models, namely transducer- and CTC-based models.

Note: There is no plan to support attention-based encoder-decoder (AED) models.

Installation

Please first install:

git clone https://github.com/k2-fsa/sherpa
cd sherpa
mkdir build
cd build
cmake \
  -DCMAKE_BUILD_TYPE=Release \
  -DCMAKE_INSTALL_PREFIX=$HOME/software/sherpa \
  ..
make -j6 install/strip

# If you don't want to strip the binaries and libraries, you can
# use "make -j6 install"

export PATH=$HOME/software/sherpa/bin:$PATH

# To uninstall sherpa, use
#  rm -rf $HOME/software/sherpa

or

git clone https://github.com/k2-fsa/sherpa
cd sherpa

python3 setup.py bdist_wheel
pip install ./dist/k2_sherpa-*.whl

# Please don't use `python3 setup.py install`.
# Otherwise, you won't have access to pre-compiled binaries

# To uninstall sherpa, use
#  pip uninstall k2-sherpa

Using Docker

docker build . -f Dockerfile -t sherpa_server:latest
docker run --rm --gpus all --name sherpa_server --net host -it sherpa_server:latest

To check that you have installed sherpa successfully, you can run the following binaries:

sherpa-version

sherpa-offline --help
sherpa-online --help
sherpa-online-microphone --help

sherpa-offline-websocket-server --help
sherpa-offline-websocket-client --help

sherpa-online-websocket-server --help
sherpa-online-websocket-client --help
sherpa-online-websocket-client-microphone --help

Usages

See documentation at https://k2-fsa.github.io/sherpa/ for more usages.

sherpa-onnx

Speech-to-text, text-to-speech, speaker recognition, and VAD using next-gen Kaldi with onnxruntime without Internet connection. Support embedded systems, Android, iOS, Raspberry Pi, RISC-V, x86_64 servers, websocket server/client, C/C++, Python, Kotlin, C#, Go, NodeJS, Java, Swift, Dart, JavaScript, Flutter, Object Pascal, Lazarus, Rust

k2

FSA/FST algorithms, differentiable, with PyTorch compatibility.

sherpa-ncnn

Real-time speech recognition and voice activity detection (VAD) using next-gen Kaldi with ncnn without Internet connection. Support iOS, Android, Linux, macOS, Windows, Raspberry Pi, VisionFive2, LicheePi4A etc.

icefall

libriheavy

Libriheavy: a 50,000 hours ASR corpus with punctuation casing and context

snowfall

Moved to https://github.com/k2-fsa/icefall

fast_rnnt

A torch implementation of a recursion which turns out to be useful for RNN-T.

text_search

Some fast-ish algorithms for batch text search in moderate-sized collections, intended for data cleanup

kaldifst

Python wrapper for OpenFST and its extensions from Kaldi. Also support reading/writing ark/scp files

multi_quantization

next-gen-kaldi-wechat

kaldi-decoder

Decoders from Kaldi using OpenFst

colab

Colab notebooks for Next-gen Kaldi

Jupyter Notebook

analyze_diagnostics

Scripts for analyzing the output of icefall's "diagnostics.py" code (--print-diagnostics=True option)

k2-fsa-www

Source for next-gen Kaldi home page.

sherpa-torch-cpp-makefile-example

divide_lm

sherpa-onnx-go

sherpa-onnx Go package for speech recognition without network access, supporting Linux, macOS, Windows

sherpa-onnx-go-windows

sherpa-onnx Go package for Windows

sherpa-ncnn-go-linux

sherpa-ncnn Go package for Linux