Skip to content

Latest commit

 

History

History
125 lines (82 loc) · 3.44 KB

pretrained-models.rst

File metadata and controls

125 lines (82 loc) · 3.44 KB

Pre-trained models

Pre-trained models for different projects

Project Pretrained models
k2-fsa/sherpa Click here <k2_fsa_sherpa_pretrained_models>
k2-fsa/sherpa-onnx Click here <sherpa-onnx-pre-trained-models>
k2-fsa/sherpa-ncnn Click here <sherpa-ncnn-pre-trained-models>

How to download

We are hosting our pre-trained models on Huggingface as git repositories managed by Git LFS.

There are at least two methods for downloading:

  • Using git lfs
  • Using wget

In the following, we use the pre-trained model sherpa-onnx-wenetspeech-2023-06-15-streaming as an example.

Using git lfs

Please first install git-lfs by following https://git-lfs.com/.

Linux

# apt/deb
sudo apt-get install git-lfs

# yum/rpm
sudo yum install git-lfs

Please see https://github.com/git-lfs/git-lfs/blob/main/INSTALLING.md for details.

macOS

brew install git-lfs

Windows

Please visit https://gitforwindows.org/ to download and install git-lfs.

Then use the following commands to download pre-trained models:

Linux/macOS

GIT_LFS_SKIP_SMUDGE=1 git clone https://huggingface.co/pkufool/icefall-asr-zipformer-streaming-wenetspeech-20230615
cd icefall-asr-zipformer-streaming-wenetspeech-20230615
git lfs pull --include "exp/*chunk-16-left-128.*onnx"

Windows (Powershell)

$env:GIT_LFS_SKIP_SMUDGE="1"
git clone https://huggingface.co/pkufool/icefall-asr-zipformer-streaming-wenetspeech-20230615
cd icefall-asr-zipformer-streaming-wenetspeech-20230615
git lfs pull --include "exp/*chunk-16-left-128.*onnx"

Windows (cmd)

set GIT_LFS_SKIP_SMUDGE="1"
git clone https://huggingface.co/pkufool/icefall-asr-zipformer-streaming-wenetspeech-20230615
cd icefall-asr-zipformer-streaming-wenetspeech-20230615
git lfs pull --include "exp/*chunk-16-left-128.*onnx"

Note

It is very important to set the environment variable GIT_LFS_SKIP_SMUDGE to 1. We don't recommend using git lfs install as it will download many large files that we don't need.

Using wget

First, let us visit the huggingface git repository of the pre-trained model:

Visit the pre-traind model git repo

Click Files and versions and navigate to the directory containing files for downloading:

Get the URL for downloading

Right click the arrow that indicates downloading and copy the link address. After that, you can use, for instance, wget to download the file with the following command:

wget https://huggingface.co/pkufool/icefall-asr-zipformer-streaming-wenetspeech-20230615/resolve/main/exp/decoder-epoch-12-avg-4-chunk-16-left-128.int8.onnx

Repeat the process until you have downloaded all the required files.