EngineX-Iluvatar/enginex_bi_series-sherpa-onnx
Archived
16
0
Fork 0
Code Issues Pull Requests Actions Projects Releases Wiki Activity

Add pre-trained models for the Libriheavy dataset (#1122)

Browse Source
This commit is contained in:
Fangjun Kuang
2024-07-13 19:21:13 +08:00
committed by GitHub
parent b5093e27f9
commit 3951a12f8d
5 changed files with 407 additions and 8 deletions
Download Patch File Download Diff File Expand all files Collapse all files

136
.github/workflows/export-libriheavy.yaml vendored Normal file
View File

@@ -0,0 +1,136 @@
name: export-libriheavy-to-onnx
on:
push:
branches:
- libriheavy-model
workflow_dispatch:
concurrency:
group: export-libriheavy-to-onnx-${{ github.ref }}
cancel-in-progress: true
jobs:
export-libriheavy-to-onnx:
if: github.repository_owner == 'k2-fsa' || github.repository_owner == 'csukuangfj'
name: export libriheavy
runs-on: ${{ matrix.os }}
strategy:
fail-fast: false
matrix:
os: [ubuntu-latest]
python-version: ["3.8"]
steps:
- uses: actions/checkout@v4
- name: Setup Python ${{ matrix.python-version }}
uses: actions/setup-python@v5
with:
python-version: ${{ matrix.python-version }}
- name: Run
shell: bash
run: |
cd scripts/icefall
./run-libriheavy.sh
./run-libriheavy-punct-case.sh
- name: Publish to huggingface
env:
HF_TOKEN: ${{ secrets.HF_TOKEN }}
uses: nick-fields/retry@v3
with:
max_attempts: 20
timeout_seconds: 200
shell: bash
command: |
git config --global user.email "csukuangfj@gmail.com"
git config --global user.name "Fangjun Kuang"
for m in large medium small; do
rm -rf huggingface
export GIT_LFS_SKIP_SMUDGE=1
export GIT_CLONE_PROTECTION_ACTIVE=false
src=sherpa-onnx-zipformer-en-libriheavy-20230926-$m
echo "Process $src"
git clone https://huggingface.co/csukuangfj/$src huggingface
cd huggingface
git fetch
git pull
cp -av ../scripts/icefall/$src/* .
git lfs track "*.onnx"
git add .
git commit -m "add large"
git push https://csukuangfj:$HF_TOKEN@huggingface.co/csukuangfj/$src main || true
cd ..
rm -rf huggingface/.git*
mv huggingface $src
tar cjvf $src.tar.bz2 $src
rm -rf $src
ls -lh
done
- name: Publish to huggingface (case and punct)
env:
HF_TOKEN: ${{ secrets.HF_TOKEN }}
uses: nick-fields/retry@v3
with:
max_attempts: 20
timeout_seconds: 200
shell: bash
command: |
git config --global user.email "csukuangfj@gmail.com"
git config --global user.name "Fangjun Kuang"
for m in large medium small; do
rm -rf huggingface
export GIT_LFS_SKIP_SMUDGE=1
export GIT_CLONE_PROTECTION_ACTIVE=false
src=sherpa-onnx-zipformer-en-libriheavy-20230830-$m-punct-case
echo "Process $src"
git clone https://huggingface.co/csukuangfj/$src huggingface
cd huggingface
git fetch
git pull
cp -av ../scripts/icefall/$src/* .
git lfs track "*.onnx"
git add .
git commit -m "add large"
git push https://csukuangfj:$HF_TOKEN@huggingface.co/csukuangfj/$src main || true
cd ..
rm -rf huggingface/.git*
mv huggingface $src
tar cjvf $src.tar.bz2 $src
rm -rf $src
ls -lh
done
- name: Release
uses: svenstaro/upload-release-action@v2
with:
file_glob: true
file: ./*.tar.bz2
overwrite: true
repo_name: k2-fsa/sherpa-onnx
repo_token: ${{ secrets.UPLOAD_GH_SHERPA_ONNX_TOKEN }}
tag: asr-models

2
scripts/apk/generate-asr-2pass-apk-script.py
View File

@@ -80,7 +80,7 @@ def get_2nd_models():
rm -fv README.md
rm -rfv test_wavs
rm model.onnx
rm -fv model.onnx
ls -lh

14
scripts/apk/generate-vad-asr-apk-script.py
View File

@@ -82,7 +82,7 @@ def get_models():
rm -fv README.md
rm -rfv test_wavs
rm model.onnx
rm -fv model.onnx
ls -lh
@@ -189,7 +189,7 @@ def get_models():
pushd $model_name
rm -rfv test_wavs
rm test.py
rm -fv test.py
ls -lh
@@ -208,8 +208,8 @@ def get_models():
rm -fv README.md
rm -fv bpe.model
rm encoder-epoch-12-avg-5.onnx
rm decoder-epoch-12-avg-5.int8.onnx
rm -fv encoder-epoch-12-avg-5.onnx
rm -fv decoder-epoch-12-avg-5.int8.onnx
rm joiner-epoch-12-avg-5.onnx
ls -lh
@@ -229,9 +229,9 @@ def get_models():
rm -fv README.md
rm -fv bpe.model
rm encoder-epoch-99-avg-1.onnx
rm decoder-epoch-99-avg-1.int8.onnx
rm joiner-epoch-99-avg-1.onnx
rm -fv encoder-epoch-99-avg-1.onnx
rm -fv decoder-epoch-99-avg-1.int8.onnx
rm -fv joiner-epoch-99-avg-1.onnx
ls -lh

131
scripts/icefall/run-libriheavy-punct-case.sh Executable file
View File

@@ -0,0 +1,131 @@
#!/usr/bin/env bash
set -ex
cur_dir=$(cd $(dirname $BASH_SOURCE) && pwd)
sherpa_onnx_dir=$(cd $cur_dir/../.. && pwd)
echo "sherpa_onnx_dir: $sherpa_onnx_dir"
pip install sherpa-onnx # for testing
function download_model() {
git lfs install
git clone https://www.modelscope.cn/pkufool/icefall-asr-zipformer-libriheavy-punc-20230830.git
}
function download_test_wavs() {
d=$1
mkdir $d/test_wavs
pushd $d/test_wavs
curl -SL -O https://huggingface.co/csukuangfj/sherpa-onnx-whisper-medium.en/resolve/main/test_wavs/0.wav
curl -SL -O https://huggingface.co/csukuangfj/sherpa-onnx-whisper-medium.en/resolve/main/test_wavs/1.wav
curl -SL -O https://huggingface.co/csukuangfj/sherpa-onnx-whisper-medium.en/resolve/main/test_wavs/8k.wav
curl -SL -O https://huggingface.co/csukuangfj/sherpa-onnx-whisper-medium.en/resolve/main/test_wavs/trans.txt
popd
}
function export_large() {
echo "----------large----------"
src=icefall-asr-zipformer-libriheavy-punc-20230830
dst=sherpa-onnx-zipformer-en-libriheavy-20230830-large-punct-case
mkdir $dst
cp -v $src/data/lang_bpe_756/bpe.model $dst/
cp -v $src/data/lang_bpe_756/tokens.txt $dst/
cp -v $src/exp/*.onnx $dst/
download_test_wavs $dst
ls -lh $dst
ls -lh $dst/test_wavs
sherpa-onnx-offline \
--encoder=$dst/encoder-epoch-16-avg-2.onnx \
--decoder=$dst/decoder-epoch-16-avg-2.onnx \
--joiner=$dst/joiner-epoch-16-avg-2.onnx \
--tokens=$dst/tokens.txt \
$dst/test_wavs/0.wav \
$dst/test_wavs/1.wav \
$dst/test_wavs/8k.wav
sherpa-onnx-offline \
--encoder=$dst/encoder-epoch-16-avg-2.int8.onnx \
--decoder=$dst/decoder-epoch-16-avg-2.onnx \
--joiner=$dst/joiner-epoch-16-avg-2.int8.onnx \
--tokens=$dst/tokens.txt \
$dst/test_wavs/0.wav \
$dst/test_wavs/1.wav \
$dst/test_wavs/8k.wav
}
function export_medium() {
echo "----------medium subset----------"
src=icefall-asr-zipformer-libriheavy-punc-20230830
dst=sherpa-onnx-zipformer-en-libriheavy-20230830-medium-punct-case
mkdir $dst
cp -v $src/data/lang_bpe_756/bpe.model $dst/
cp -v $src/data/lang_bpe_756/tokens.txt $dst/
cp -v $src/exp_medium_subset/*.onnx $dst/
download_test_wavs $dst
ls -lh $dst
ls -lh $dst/test_wavs
sherpa-onnx-offline \
--encoder=$dst/encoder-epoch-50-avg-15.onnx \
--decoder=$dst/decoder-epoch-50-avg-15.onnx \
--joiner=$dst/joiner-epoch-50-avg-15.onnx \
--tokens=$dst/tokens.txt \
$dst/test_wavs/0.wav \
$dst/test_wavs/1.wav \
$dst/test_wavs/8k.wav
sherpa-onnx-offline \
--encoder=$dst/encoder-epoch-50-avg-15.int8.onnx \
--decoder=$dst/decoder-epoch-50-avg-15.onnx \
--joiner=$dst/joiner-epoch-50-avg-15.int8.onnx \
--tokens=$dst/tokens.txt \
$dst/test_wavs/0.wav \
$dst/test_wavs/1.wav \
$dst/test_wavs/8k.wav
}
function export_small() {
echo "----------small subset----------"
src=icefall-asr-zipformer-libriheavy-punc-20230830
dst=sherpa-onnx-zipformer-en-libriheavy-20230830-small-punct-case
mkdir $dst
cp -v $src/data/lang_bpe_756/bpe.model $dst/
cp -v $src/data/lang_bpe_756/tokens.txt $dst/
cp -v $src/exp_small_subset/*.onnx $dst/
download_test_wavs $dst
ls -lh $dst
ls -lh $dst/test_wavs
sherpa-onnx-offline \
--encoder=$dst/encoder-epoch-88-avg-41.onnx \
--decoder=$dst/decoder-epoch-88-avg-41.onnx \
--joiner=$dst/joiner-epoch-88-avg-41.onnx \
--tokens=$dst/tokens.txt \
$dst/test_wavs/0.wav \
$dst/test_wavs/1.wav \
$dst/test_wavs/8k.wav
sherpa-onnx-offline \
--encoder=$dst/encoder-epoch-88-avg-41.int8.onnx \
--decoder=$dst/decoder-epoch-88-avg-41.onnx \
--joiner=$dst/joiner-epoch-88-avg-41.int8.onnx \
--tokens=$dst/tokens.txt \
$dst/test_wavs/0.wav \
$dst/test_wavs/1.wav \
$dst/test_wavs/8k.wav
}
download_model
export_large
export_medium
export_small
rm -rf icefall-asr-zipformer-libriheavy-punc-20230830

132
scripts/icefall/run-libriheavy.sh Executable file
View File

@@ -0,0 +1,132 @@
#!/usr/bin/env bash
set -ex
cur_dir=$(cd $(dirname $BASH_SOURCE) && pwd)
sherpa_onnx_dir=$(cd $cur_dir/../.. && pwd)
echo "sherpa_onnx_dir: $sherpa_onnx_dir"
pip install sherpa-onnx # for testing
function download_model() {
git lfs install
git clone https://www.modelscope.cn/pkufool/icefall-asr-zipformer-libriheavy-20230926.git
}
function download_test_wavs() {
d=$1
mkdir $d/test_wavs
pushd $d/test_wavs
curl -SL -O https://huggingface.co/csukuangfj/sherpa-onnx-whisper-medium.en/resolve/main/test_wavs/0.wav
curl -SL -O https://huggingface.co/csukuangfj/sherpa-onnx-whisper-medium.en/resolve/main/test_wavs/1.wav
curl -SL -O https://huggingface.co/csukuangfj/sherpa-onnx-whisper-medium.en/resolve/main/test_wavs/8k.wav
curl -SL -O https://huggingface.co/csukuangfj/sherpa-onnx-whisper-medium.en/resolve/main/test_wavs/trans.txt
popd
}
function export_large() {
echo "----------large----------"
src=icefall-asr-zipformer-libriheavy-20230926
dst=sherpa-onnx-zipformer-en-libriheavy-20230926-large
mkdir $dst
cp -v $src/data/lang_bpe_500/bpe.model $dst/
cp -v $src/data/lang_bpe_500/tokens.txt $dst/
cp -v $src/exp/*.onnx $dst/
download_test_wavs $dst
ls -lh $dst
ls -lh $dst/test_wavs
sherpa-onnx-offline \
--encoder=$dst/encoder-epoch-16-avg-3.onnx \
--decoder=$dst/decoder-epoch-16-avg-3.onnx \
--joiner=$dst/joiner-epoch-16-avg-3.onnx \
--tokens=$dst/tokens.txt \
$dst/test_wavs/0.wav \
$dst/test_wavs/1.wav \
$dst/test_wavs/8k.wav
sherpa-onnx-offline \
--encoder=$dst/encoder-epoch-16-avg-3.int8.onnx \
--decoder=$dst/decoder-epoch-16-avg-3.onnx \
--joiner=$dst/joiner-epoch-16-avg-3.int8.onnx \
--tokens=$dst/tokens.txt \
$dst/test_wavs/0.wav \
$dst/test_wavs/1.wav \
$dst/test_wavs/8k.wav
}
function export_medium() {
echo "----------medium subset----------"
src=icefall-asr-zipformer-libriheavy-20230926
dst=sherpa-onnx-zipformer-en-libriheavy-20230926-medium
mkdir $dst
cp -v $src/data/lang_bpe_500/bpe.model $dst/
cp -v $src/data/lang_bpe_500/tokens.txt $dst/
cp -v $src/exp_medium_subset/*.onnx $dst/
download_test_wavs $dst
ls -lh $dst
ls -lh $dst/test_wavs
sherpa-onnx-offline \
--encoder=$dst/encoder-epoch-60-avg-20.onnx \
--decoder=$dst/decoder-epoch-60-avg-20.onnx \
--joiner=$dst/joiner-epoch-60-avg-20.onnx \
--tokens=$dst/tokens.txt \
$dst/test_wavs/0.wav \
$dst/test_wavs/1.wav \
$dst/test_wavs/8k.wav
sherpa-onnx-offline \
--encoder=$dst/encoder-epoch-60-avg-20.int8.onnx \
--decoder=$dst/decoder-epoch-60-avg-20.onnx \
--joiner=$dst/joiner-epoch-60-avg-20.int8.onnx \
--tokens=$dst/tokens.txt \
$dst/test_wavs/0.wav \
$dst/test_wavs/1.wav \
$dst/test_wavs/8k.wav
}
function export_small() {
echo "----------small subset----------"
src=icefall-asr-zipformer-libriheavy-20230926
dst=sherpa-onnx-zipformer-en-libriheavy-20230926-small
mkdir $dst
cp -v $src/data/lang_bpe_500/bpe.model $dst/
cp -v $src/data/lang_bpe_500/tokens.txt $dst/
cp -v $src/exp_small_subset/*.onnx $dst/
download_test_wavs $dst
ls -lh $dst
ls -lh $dst/test_wavs
sherpa-onnx-offline \
--encoder=$dst/encoder-epoch-90-avg-20.onnx \
--decoder=$dst/decoder-epoch-90-avg-20.onnx \
--joiner=$dst/joiner-epoch-90-avg-20.onnx \
--tokens=$dst/tokens.txt \
$dst/test_wavs/0.wav \
$dst/test_wavs/1.wav \
$dst/test_wavs/8k.wav
sherpa-onnx-offline \
--encoder=$dst/encoder-epoch-90-avg-20.int8.onnx \
--decoder=$dst/decoder-epoch-90-avg-20.onnx \
--joiner=$dst/joiner-epoch-90-avg-20.int8.onnx \
--tokens=$dst/tokens.txt \
$dst/test_wavs/0.wav \
$dst/test_wavs/1.wav \
$dst/test_wavs/8k.wav
}
download_model
export_large
export_medium
export_small
rm -rf icefall-asr-zipformer-libriheavy-20230926