EngineX-Iluvatar/enginex_bi_series-sherpa-onnx
Archived
15
0
Fork 0
Code Issues Pull Requests Actions Projects Releases Wiki Activity
This repository has been archived on 2025-08-26. You can view files and clone it, but cannot push or open issues or pull requests.
Files
f0960342ad484273e6de29cf1a5f1101c0141c42
enginex_bi_series-sherpa-onnx/.github/scripts/test-online-transducer.sh
Askars Salimbajevs f0960342ad Add LODR support to online and offline recognizers (#2026) 
This PR integrates LODR (Level-Ordered Deterministic Rescoring) support from Icefall into both online and offline recognizers, enabling LODR for LM shallow fusion and LM rescore.

- Extended OnlineLMConfig and OfflineLMConfig to include lodr_fst, lodr_scale, and lodr_backoff_id.
- Implemented LodrFst and LodrStateCost classes and wired them into RNN LM scoring in both online and offline code paths.
- Updated Python bindings, CLI entry points, examples, and CI test scripts to accept and exercise the new LODR options.
2025-07-09 16:23:46 +08:00

334 lines
9.1 KiB
Bash
Executable File
Raw Blame History

#!/usr/bin/env bash
set -e
log() {
# This function is from espnet
local fname=${BASH_SOURCE[1]##*/}
echo -e "$(date '+%Y-%m-%d %H:%M:%S') (${fname}:${BASH_LINENO[0]}:${FUNCNAME[1]}) $*"
}
export GIT_CLONE_PROTECTION_ACTIVE=false
echo "EXE is $EXE"
echo "PATH: $PATH"
which $EXE
log "------------------------------------------------------------"
log "Run NeMo transducer (English)"
log "------------------------------------------------------------"
repo_url=https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/sherpa-onnx-nemo-streaming-fast-conformer-transducer-en-80ms.tar.bz2
curl -SL -O $repo_url
tar xvf sherpa-onnx-nemo-streaming-fast-conformer-transducer-en-80ms.tar.bz2
rm sherpa-onnx-nemo-streaming-fast-conformer-transducer-en-80ms.tar.bz2
repo=sherpa-onnx-nemo-streaming-fast-conformer-transducer-en-80ms
log "Start testing ${repo_url}"
waves=(
$repo/test_wavs/0.wav
$repo/test_wavs/1.wav
$repo/test_wavs/8k.wav
)
for wave in ${waves[@]}; do
time $EXE \
--tokens=$repo/tokens.txt \
--encoder=$repo/encoder.onnx \
--decoder=$repo/decoder.onnx \
--joiner=$repo/joiner.onnx \
--num-threads=2 \
$wave
done
time $EXE \
--tokens=$repo/tokens.txt \
--encoder=$repo/encoder.onnx \
--decoder=$repo/decoder.onnx \
--joiner=$repo/joiner.onnx \
--num-threads=2 \
$repo/test_wavs/0.wav \
$repo/test_wavs/1.wav \
$repo/test_wavs/8k.wav
rm -rf $repo
log "------------------------------------------------------------"
log "Run LSTM transducer (English)"
log "------------------------------------------------------------"
repo_url=https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/sherpa-onnx-lstm-en-2023-02-17.tar.bz2
curl -SL -O $repo_url
tar xvf sherpa-onnx-lstm-en-2023-02-17.tar.bz2
rm sherpa-onnx-lstm-en-2023-02-17.tar.bz2
repo=sherpa-onnx-lstm-en-2023-02-17
log "Start testing ${repo_url}"
waves=(
$repo/test_wavs/0.wav
$repo/test_wavs/1.wav
$repo/test_wavs/8k.wav
)
for wave in ${waves[@]}; do
time $EXE \
--tokens=$repo/tokens.txt \
--encoder=$repo/encoder-epoch-99-avg-1.onnx \
--decoder=$repo/decoder-epoch-99-avg-1.onnx \
--joiner=$repo/joiner-epoch-99-avg-1.onnx \
--num-threads=2 \
$wave
done
for wave in ${waves[@]}; do
time $EXE \
--tokens=$repo/tokens.txt \
--encoder=$repo/encoder-epoch-99-avg-1.int8.onnx \
--decoder=$repo/decoder-epoch-99-avg-1.onnx \
--joiner=$repo/joiner-epoch-99-avg-1.int8.onnx \
--num-threads=2 \
$wave
done
rm -rf $repo
log "------------------------------------------------------------"
log "Run LSTM transducer (Chinese)"
log "------------------------------------------------------------"
repo_url=https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/sherpa-onnx-lstm-zh-2023-02-20.tar.bz2
curl -SL -O $repo_url
tar xvf sherpa-onnx-lstm-zh-2023-02-20.tar.bz2
rm sherpa-onnx-lstm-zh-2023-02-20.tar.bz2
repo=sherpa-onnx-lstm-zh-2023-02-20
log "Start testing ${repo_url}"
waves=(
$repo/test_wavs/0.wav
$repo/test_wavs/1.wav
$repo/test_wavs/8k.wav
)
for wave in ${waves[@]}; do
time $EXE \
--tokens=$repo/tokens.txt \
--encoder=$repo/encoder-epoch-11-avg-1.onnx \
--decoder=$repo/decoder-epoch-11-avg-1.onnx \
--joiner=$repo/joiner-epoch-11-avg-1.onnx \
--num-threads=2 \
$wave
done
for wave in ${waves[@]}; do
time $EXE \
--tokens=$repo/tokens.txt \
--encoder=$repo/encoder-epoch-11-avg-1.int8.onnx \
--decoder=$repo/decoder-epoch-11-avg-1.onnx \
--joiner=$repo/joiner-epoch-11-avg-1.int8.onnx \
--num-threads=2 \
$wave
done
rm -rf $repo
log "------------------------------------------------------------"
log "Run streaming Zipformer transducer (English)"
log "------------------------------------------------------------"
repo_url=https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/sherpa-onnx-streaming-zipformer-en-2023-02-21.tar.bz2
curl -SL -O $repo_url
tar xvf sherpa-onnx-streaming-zipformer-en-2023-02-21.tar.bz2
rm sherpa-onnx-streaming-zipformer-en-2023-02-21.tar.bz2
repo=sherpa-onnx-streaming-zipformer-en-2023-02-21
log "Start testing ${repo_url}"
waves=(
$repo/test_wavs/0.wav
$repo/test_wavs/1.wav
$repo/test_wavs/8k.wav
)
for wave in ${waves[@]}; do
time $EXE \
--tokens=$repo/tokens.txt \
--encoder=$repo/encoder-epoch-99-avg-1.onnx \
--decoder=$repo/decoder-epoch-99-avg-1.onnx \
--joiner=$repo/joiner-epoch-99-avg-1.onnx \
--num-threads=2 \
$wave
done
# test int8
#
for wave in ${waves[@]}; do
time $EXE \
--tokens=$repo/tokens.txt \
--encoder=$repo/encoder-epoch-99-avg-1.int8.onnx \
--decoder=$repo/decoder-epoch-99-avg-1.onnx \
--joiner=$repo/joiner-epoch-99-avg-1.int8.onnx \
--num-threads=2 \
$wave
done
lm_repo_url=https://huggingface.co/vsd-vector/icefall-librispeech-rnn-lm
log "Download pre-trained RNN-LM model from ${lm_repo_url}"
GIT_LFS_SKIP_SMUDGE=1 git clone $lm_repo_url
lm_repo=$(basename $lm_repo_url)
pushd $lm_repo
git lfs pull --include "with-state-epoch-99-avg-1.onnx"
popd
bigram_repo_url=https://huggingface.co/vsd-vector/librispeech_bigram_sherpa-onnx-zipformer-large-en-2023-06-26
log "Download bi-gram LM from ${bigram_repo_url}"
GIT_LFS_SKIP_SMUDGE=1 git clone $bigram_repo_url
bigramlm_repo=$(basename $bigram_repo_url)
pushd $bigramlm_repo
git lfs pull --include "2gram.fst"
popd
log "Start testing LODR"
waves=(
$repo/test_wavs/0.wav
$repo/test_wavs/1.wav
$repo/test_wavs/8k.wav
)
for wave in ${waves[@]}; do
time $EXE \
--tokens=$repo/tokens.txt \
--encoder=$repo/encoder-epoch-99-avg-1.onnx \
--decoder=$repo/decoder-epoch-99-avg-1.onnx \
--joiner=$repo/joiner-epoch-99-avg-1.onnx \
--num-threads=2 \
--decoding_method="modified_beam_search" \
--lm=$lm_repo/with-state-epoch-99-avg-1.onnx \
--lodr-fst=$bigramlm_repo/2gram.fst \
--lodr-scale=-0.5 \
$wave
done
for wave in ${waves[@]}; do
time $EXE \
--tokens=$repo/tokens.txt \
--encoder=$repo/encoder-epoch-99-avg-1.onnx \
--decoder=$repo/decoder-epoch-99-avg-1.onnx \
--joiner=$repo/joiner-epoch-99-avg-1.onnx \
--num-threads=2 \
--decoding_method="modified_beam_search" \
--lm=$lm_repo/with-state-epoch-99-avg-1.onnx \
--lodr-fst=$bigramlm_repo/2gram.fst \
--lodr-scale=-0.5 \
--lm-shallow-fusion=true \
$wave
done
rm -rf $repo $bigramlm_repo $lm_repo
log "------------------------------------------------------------"
log "Run streaming Zipformer transducer (Bilingual, Chinese + English)"
log "------------------------------------------------------------"
repo_url=https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/sherpa-onnx-streaming-zipformer-bilingual-zh-en-2023-02-20.tar.bz2
curl -SL -O $repo_url
tar xvf sherpa-onnx-streaming-zipformer-bilingual-zh-en-2023-02-20.tar.bz2
rm sherpa-onnx-streaming-zipformer-bilingual-zh-en-2023-02-20.tar.bz2
repo=sherpa-onnx-streaming-zipformer-bilingual-zh-en-2023-02-20
log "Start testing ${repo_url}"
waves=(
$repo/test_wavs/0.wav
$repo/test_wavs/1.wav
$repo/test_wavs/2.wav
$repo/test_wavs/3.wav
$repo/test_wavs/8k.wav
)
for wave in ${waves[@]}; do
time $EXE \
--tokens=$repo/tokens.txt \
--encoder=$repo/encoder-epoch-99-avg-1.onnx \
--decoder=$repo/decoder-epoch-99-avg-1.onnx \
--joiner=$repo/joiner-epoch-99-avg-1.onnx \
--num-threads=2 \
$wave
done
for wave in ${waves[@]}; do
time $EXE \
--tokens=$repo/tokens.txt \
--encoder=$repo/encoder-epoch-99-avg-1.int8.onnx \
--decoder=$repo/decoder-epoch-99-avg-1.onnx \
--joiner=$repo/joiner-epoch-99-avg-1.int8.onnx \
--num-threads=2 \
$wave
done
# Decode a URL
if [ $EXE == "sherpa-onnx-ffmpeg" ]; then
time $EXE \
$repo/tokens.txt \
$repo/encoder-epoch-99-avg-1.onnx \
$repo/decoder-epoch-99-avg-1.onnx \
$repo/joiner-epoch-99-avg-1.onnx \
https://huggingface.co/csukuangfj/sherpa-onnx-streaming-zipformer-bilingual-zh-en-2023-02-20/resolve/main/test_wavs/4.wav \
2
fi
if [ $EXE == "sherpa-onnx-ffmpeg" ]; then
time $EXE \
$repo/tokens.txt \
$repo/encoder-epoch-99-avg-1.int8.onnx \
$repo/decoder-epoch-99-avg-1.onnx \
$repo/joiner-epoch-99-avg-1.int8.onnx \
https://huggingface.co/csukuangfj/sherpa-onnx-streaming-zipformer-bilingual-zh-en-2023-02-20/resolve/main/test_wavs/4.wav \
2
fi
rm -rf $repo
log "------------------------------------------------------------"
log "Run streaming Conformer transducer (English)"
log "------------------------------------------------------------"
repo_url=https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/sherpa-onnx-streaming-conformer-en-2023-05-09.tar.bz2
curl -SL -O $repo_url
tar xvf sherpa-onnx-streaming-conformer-en-2023-05-09.tar.bz2
rm sherpa-onnx-streaming-conformer-en-2023-05-09.tar.bz2
repo=sherpa-onnx-streaming-conformer-en-2023-05-09
log "Start testing ${repo_url}"
waves=(
$repo/test_wavs/0.wav
$repo/test_wavs/1.wav
$repo/test_wavs/2.wav
)
for wave in ${waves[@]}; do
time $EXE \
--tokens=$repo/tokens.txt \
--encoder=$repo/encoder-epoch-99-avg-1.onnx \
--decoder=$repo/decoder-epoch-99-avg-1.onnx \
--joiner=$repo/joiner-epoch-99-avg-1.onnx \
--num-threads=2 \
$wave
done
for wave in ${waves[@]}; do
time $EXE \
--tokens=$repo/tokens.txt \
--encoder=$repo/encoder-epoch-99-avg-1.int8.onnx \
--decoder=$repo/decoder-epoch-99-avg-1.onnx \
--joiner=$repo/joiner-epoch-99-avg-1.int8.onnx \
--num-threads=2 \
$wave
done
rm -rf $repo
Reference in New Issue View Git Blame Copy Permalink