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

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

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

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

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