mirrors/icefall
1
0
Fork 0
mirror of https://github.com/k2-fsa/icefall.git synced 2025-08-26 10:16:14 +00:00
Code Issues Packages Projects Releases Wiki Activity

Merge remote-tracking branch 'k2-fsa/master' into zipformer-ctc-aed

Browse Source
This commit is contained in:
yaozengwei 2024-05-25 17:49:14 +08:00
commit 84dfb5765b
1001 changed files with 125807 additions and 8113 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
.flake8
View File

@ -15,7 +15,7 @@ per-file-ignores =
egs/librispeech/ASR/zipformer_mmi/*.py: E501, E203
egs/librispeech/ASR/zipformer/*.py: E501, E203
egs/librispeech/ASR/RESULTS.md: E999,
egs/ljspeech/TTS/vits/*.py: E501, E203
# invalid escape sequence (cause by tex formular), W605
icefall/utils.py: E501, W605

1
.github/scripts/.gitignore vendored Normal file
View File

@ -0,0 +1 @@
piper_phonemize.html

343
.github/scripts/aishell/ASR/run.sh vendored Executable file
View File

@ -0,0 +1,343 @@
#!/usr/bin/env bash
set -ex
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]}) $*"
}
cd egs/aishell/ASR
function download_test_dev_manifests() {
git lfs install
fbank_url=https://huggingface.co/csukuangfj/aishell-test-dev-manifests
log "Downloading pre-commputed fbank from $fbank_url"
git clone https://huggingface.co/csukuangfj/aishell-test-dev-manifests
ln -s $PWD/aishell-test-dev-manifests/data .
}
function test_transducer_stateless3_2022_06_20() {
repo_url=https://huggingface.co/csukuangfj/icefall-aishell-pruned-transducer-stateless3-2022-06-20
log "Downloading pre-trained model from $repo_url"
git clone $repo_url
repo=$(basename $repo_url)
log "Display test files"
tree $repo/
ls -lh $repo/test_wavs/*.wav
pushd $repo/exp
ln -s pretrained-epoch-29-avg-5-torch-1.10.0.pt pretrained.pt
popd
log "test greedy_search with pretrained.py"
for sym in 1 2 3; do
log "Greedy search with --max-sym-per-frame $sym"
./pruned_transducer_stateless3/pretrained.py \
--method greedy_search \
--max-sym-per-frame $sym \
--checkpoint $repo/exp/pretrained.pt \
--lang-dir $repo/data/lang_char \
$repo/test_wavs/BAC009S0764W0121.wav \
$repo/test_wavs/BAC009S0764W0122.wav \
$repo/test_wavs/BAC009S0764W0123.wav
done
log "test beam search with pretrained.py"
for method in modified_beam_search beam_search fast_beam_search; do
log "$method"
./pruned_transducer_stateless3/pretrained.py \
--method $method \
--beam-size 4 \
--checkpoint $repo/exp/pretrained.pt \
--lang-dir $repo/data/lang_char \
$repo/test_wavs/BAC009S0764W0121.wav \
$repo/test_wavs/BAC009S0764W0122.wav \
$repo/test_wavs/BAC009S0764W0123.wav
done
echo "GITHUB_EVENT_NAME: ${GITHUB_EVENT_NAME}"
echo "GITHUB_EVENT_LABEL_NAME: ${GITHUB_EVENT_LABEL_NAME}"
if [[ x"${GITHUB_EVENT_NAME}" == x"schedule" || x"${GITHUB_EVENT_LABEL_NAME}" == x"run-decode" ]]; then
mkdir -p pruned_transducer_stateless3/exp
ln -s $PWD/$repo/exp/pretrained.pt pruned_transducer_stateless3/exp/epoch-999.pt
ln -s $PWD/$repo/data/lang_char data/
ls -lh data
ls -lh pruned_transducer_stateless3/exp
log "Decoding test and dev"
# use a small value for decoding with CPU
max_duration=100
for method in greedy_search fast_beam_search modified_beam_search; do
log "Decoding with $method"
./pruned_transducer_stateless3/decode.py \
--decoding-method $method \
--epoch 999 \
--avg 1 \
--max-duration $max_duration \
--exp-dir pruned_transducer_stateless3/exp
done
rm pruned_transducer_stateless3/exp/*.pt
fi
rm -rf $repo
}
function test_zipformer_large_2023_10_24() {
log "CI testing large model"
repo_url=https://huggingface.co/zrjin/icefall-asr-aishell-zipformer-large-2023-10-24/
log "Downloading pre-trained model from $repo_url"
git clone $repo_url
repo=$(basename $repo_url)
log "Display test files"
tree $repo/
ls -lh $repo/test_wavs/*.wav
for method in modified_beam_search greedy_search fast_beam_search; do
log "$method"
./zipformer/pretrained.py \
--method $method \
--context-size 1 \
--checkpoint $repo/exp/pretrained.pt \
--tokens $repo/data/lang_char/tokens.txt \
--num-encoder-layers 2,2,4,5,4,2 \
--feedforward-dim 512,768,1536,2048,1536,768 \
--encoder-dim 192,256,512,768,512,256 \
--encoder-unmasked-dim 192,192,256,320,256,192 \
$repo/test_wavs/BAC009S0764W0121.wav \
$repo/test_wavs/BAC009S0764W0122.wav \
$repo/test_wavs/BAC009S0764W0123.wav
done
rm -rf $repo
}
function test_zipformer_2023_10_24() {
repo_url=https://huggingface.co/zrjin/icefall-asr-aishell-zipformer-2023-10-24/
log "Downloading pre-trained model from $repo_url"
git clone $repo_url
repo=$(basename $repo_url)
log "Display test files"
tree $repo/
ls -lh $repo/test_wavs/*.wav
for method in modified_beam_search greedy_search fast_beam_search; do
log "$method"
./zipformer/pretrained.py \
--method $method \
--context-size 1 \
--checkpoint $repo/exp/pretrained.pt \
--tokens $repo/data/lang_char/tokens.txt \
$repo/test_wavs/BAC009S0764W0121.wav \
$repo/test_wavs/BAC009S0764W0122.wav \
$repo/test_wavs/BAC009S0764W0123.wav
done
rm -rf $repo
}
function test_zipformer_small_2023_10_24() {
log "CI testing small model"
repo_url=https://huggingface.co/zrjin/icefall-asr-aishell-zipformer-small-2023-10-24/
log "Downloading pre-trained model from $repo_url"
git clone $repo_url
repo=$(basename $repo_url)
log "Display test files"
tree $repo/
ls -lh $repo/test_wavs/*.wav
for method in modified_beam_search greedy_search fast_beam_search; do
log "$method"
./zipformer/pretrained.py \
--method $method \
--context-size 1 \
--checkpoint $repo/exp/pretrained.pt \
--tokens $repo/data/lang_char/tokens.txt \
--num-encoder-layers 2,2,2,2,2,2 \
--feedforward-dim 512,768,768,768,768,768 \
--encoder-dim 192,256,256,256,256,256 \
--encoder-unmasked-dim 192,192,192,192,192,192 \
$repo/test_wavs/BAC009S0764W0121.wav \
$repo/test_wavs/BAC009S0764W0122.wav \
$repo/test_wavs/BAC009S0764W0123.wav
done
rm -rf $repo
}
function test_transducer_stateless_modified_2022_03_01() {
repo_url=https://huggingface.co/csukuangfj/icefall-aishell-transducer-stateless-modified-2022-03-01
log "Downloading pre-trained model from $repo_url"
git lfs install
git clone $repo_url
repo=$(basename $repo_url)
log "Display test files"
tree $repo/
ls -lh $repo/test_wavs/*.wav
for sym in 1 2 3; do
log "Greedy search with --max-sym-per-frame $sym"
./transducer_stateless_modified/pretrained.py \
--method greedy_search \
--max-sym-per-frame $sym \
--checkpoint $repo/exp/pretrained.pt \
--lang-dir $repo/data/lang_char \
$repo/test_wavs/BAC009S0764W0121.wav \
$repo/test_wavs/BAC009S0764W0122.wav \
$repo/test_wavs/BAC009S0764W0123.wav
done
for method in modified_beam_search beam_search; do
log "$method"
./transducer_stateless_modified/pretrained.py \
--method $method \
--beam-size 4 \
--checkpoint $repo/exp/pretrained.pt \
--lang-dir $repo/data/lang_char \
$repo/test_wavs/BAC009S0764W0121.wav \
$repo/test_wavs/BAC009S0764W0122.wav \
$repo/test_wavs/BAC009S0764W0123.wav
done
rm -rf $repo
}
function test_transducer_stateless_modified_2_2022_03_01() {
repo_url=https://huggingface.co/csukuangfj/icefall-aishell-transducer-stateless-modified-2-2022-03-01
log "Downloading pre-trained model from $repo_url"
git lfs install
git clone $repo_url
repo=$(basename $repo_url)
log "Display test files"
tree $repo/
ls -lh $repo/test_wavs/*.wav
for sym in 1 2 3; do
log "Greedy search with --max-sym-per-frame $sym"
./transducer_stateless_modified-2/pretrained.py \
--method greedy_search \
--max-sym-per-frame $sym \
--checkpoint $repo/exp/pretrained.pt \
--lang-dir $repo/data/lang_char \
$repo/test_wavs/BAC009S0764W0121.wav \
$repo/test_wavs/BAC009S0764W0122.wav \
$repo/test_wavs/BAC009S0764W0123.wav
done
for method in modified_beam_search beam_search; do
log "$method"
./transducer_stateless_modified-2/pretrained.py \
--method $method \
--beam-size 4 \
--checkpoint $repo/exp/pretrained.pt \
--lang-dir $repo/data/lang_char \
$repo/test_wavs/BAC009S0764W0121.wav \
$repo/test_wavs/BAC009S0764W0122.wav \
$repo/test_wavs/BAC009S0764W0123.wav
done
rm -rf $repo
}
function test_conformer_ctc() {
repo_url=https://huggingface.co/csukuangfj/icefall_asr_aishell_conformer_ctc
log "Downloading pre-trained model from $repo_url"
GIT_LFS_SKIP_SMUDGE=1 git clone $repo_url
repo=$(basename $repo_url)
pushd $repo
git lfs pull --include "exp/pretrained.pt"
git lfs pull --include "data/lang_char/H.fst"
git lfs pull --include "data/lang_char/HL.fst"
git lfs pull --include "data/lang_char/HLG.fst"
popd
log "Display test files"
tree $repo/
ls -lh $repo/test_wavs/*.wav
log "CTC decoding"
log "Exporting model with torchscript"
pushd $repo/exp
ln -s pretrained.pt epoch-99.pt
popd
./conformer_ctc/export.py \
--epoch 99 \
--avg 1 \
--exp-dir $repo/exp \
--tokens $repo/data/lang_char/tokens.txt \
--jit 1
ls -lh $repo/exp
ls -lh $repo/data/lang_char
log "Decoding with H on CPU with OpenFst"
./conformer_ctc/jit_pretrained_decode_with_H.py \
--nn-model $repo/exp/cpu_jit.pt \
--H $repo/data/lang_char/H.fst \
--tokens $repo/data/lang_char/tokens.txt \
$repo/test_wavs/0.wav \
$repo/test_wavs/1.wav \
$repo/test_wavs/2.wav
log "Decoding with HL on CPU with OpenFst"
./conformer_ctc/jit_pretrained_decode_with_HL.py \
--nn-model $repo/exp/cpu_jit.pt \
--HL $repo/data/lang_char/HL.fst \
--words $repo/data/lang_char/words.txt \
$repo/test_wavs/0.wav \
$repo/test_wavs/1.wav \
$repo/test_wavs/2.wav
log "Decoding with HLG on CPU with OpenFst"
./conformer_ctc/jit_pretrained_decode_with_HLG.py \
--nn-model $repo/exp/cpu_jit.pt \
--HLG $repo/data/lang_char/HLG.fst \
--words $repo/data/lang_char/words.txt \
$repo/test_wavs/0.wav \
$repo/test_wavs/1.wav \
$repo/test_wavs/2.wav
rm -rf $repo
}
download_test_dev_manifests
test_transducer_stateless3_2022_06_20
test_zipformer_large_2023_10_24
test_zipformer_2023_10_24
test_zipformer_small_2023_10_24
test_transducer_stateless_modified_2022_03_01
test_transducer_stateless_modified_2_2022_03_01
# test_conformer_ctc # fails for torch 1.13.x and torch 2.0.x

94
.github/scripts/audioset/AT/run.sh vendored Executable file
View File

@ -0,0 +1,94 @@
#!/usr/bin/env bash
set -ex
python3 -m pip install onnxoptimizer onnxsim
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]}) $*"
}
cd egs/audioset/AT
function test_pretrained() {
repo_url=https://huggingface.co/marcoyang/icefall-audio-tagging-audioset-zipformer-2024-03-12
repo=$(basename $repo_url)
GIT_LFS_SKIP_SMUDGE=1 git clone $repo_url
pushd $repo/exp
git lfs pull --include pretrained.pt
ln -s pretrained.pt epoch-99.pt
ls -lh
popd
log "test pretrained.pt"
python3 zipformer/pretrained.py \
--checkpoint $repo/exp/pretrained.pt \
--label-dict $repo/data/class_labels_indices.csv \
$repo/test_wavs/1.wav \
$repo/test_wavs/2.wav \
$repo/test_wavs/3.wav \
$repo/test_wavs/4.wav
log "test jit export"
ls -lh $repo/exp/
python3 zipformer/export.py \
--exp-dir $repo/exp \
--epoch 99 \
--avg 1 \
--use-averaged-model 0 \
--jit 1
ls -lh $repo/exp/
log "test jit models"
python3 zipformer/jit_pretrained.py \
--nn-model-filename $repo/exp/jit_script.pt \
--label-dict $repo/data/class_labels_indices.csv \
$repo/test_wavs/1.wav \
$repo/test_wavs/2.wav \
$repo/test_wavs/3.wav \
$repo/test_wavs/4.wav
log "test onnx export"
ls -lh $repo/exp/
python3 zipformer/export-onnx.py \
--exp-dir $repo/exp \
--epoch 99 \
--avg 1 \
--use-averaged-model 0
ls -lh $repo/exp/
pushd $repo/exp/
mv model-epoch-99-avg-1.onnx model.onnx
mv model-epoch-99-avg-1.int8.onnx model.int8.onnx
popd
ls -lh $repo/exp/
log "test onnx models"
for m in model.onnx model.int8.onnx; do
log "$m"
python3 zipformer/onnx_pretrained.py \
--model-filename $repo/exp/model.onnx \
--label-dict $repo/data/class_labels_indices.csv \
$repo/test_wavs/1.wav \
$repo/test_wavs/2.wav \
$repo/test_wavs/3.wav \
$repo/test_wavs/4.wav
done
log "prepare data for uploading to huggingface"
dst=/icefall/model-onnx
mkdir -p $dst
cp -v $repo/exp/*.onnx $dst/
cp -v $repo/data/* $dst/
cp -av $repo/test_wavs $dst
ls -lh $dst
ls -lh $dst/test_wavs
}
test_pretrained

71
.github/scripts/docker/Dockerfile vendored Normal file
View File

@ -0,0 +1,71 @@
ARG PYTHON_VERSION=3.8
FROM python:${PYTHON_VERSION}
ARG TORCHAUDIO_VERSION="0.13.0"
ARG TORCH_VERSION="1.13.0"
ARG K2_VERSION="1.24.4.dev20231220"
ARG KALDIFEAT_VERSION="1.25.3.dev20231221"
ARG _K2_VERSION="${K2_VERSION}+cpu.torch${TORCH_VERSION}"
ARG _KALDIFEAT_VERSION="${KALDIFEAT_VERSION}+cpu.torch${TORCH_VERSION}"
RUN apt-get update -y && \
apt-get install -qq -y \
cmake \
ffmpeg \
git \
git-lfs \
graphviz \
less \
tree \
vim \
&& \
apt-get clean && \
rm -rf /var/cache/apt/archives /var/lib/apt/lists
LABEL authors="Fangjun Kuang <csukuangfj@gmail.com>"
LABEL k2_version=${_K2_VERSION}
LABEL kaldifeat_version=${_KALDIFEAT_VERSION}
LABEL github_repo="https://github.com/k2-fsa/icefall"
# Install dependencies
RUN pip install --no-cache-dir \
torch==${TORCH_VERSION} torchaudio==${TORCHAUDIO_VERSION} -f https://download.pytorch.org/whl/cpu/torch_stable.html \
k2==${_K2_VERSION} -f https://k2-fsa.github.io/k2/cpu.html \
\
git+https://github.com/lhotse-speech/lhotse \
kaldifeat==${_KALDIFEAT_VERSION} -f https://csukuangfj.github.io/kaldifeat/cpu.html \
cython \
dill \
espnet_tts_frontend \
graphviz \
kaldi-decoder \
kaldi_native_io \
kaldialign \
kaldifst \
kaldilm \
matplotlib \
multi_quantization \
numba \
numpy \
onnxoptimizer \
onnxsim \
onnx \
onnxmltools \
onnxruntime \
piper_phonemize -f https://k2-fsa.github.io/icefall/piper_phonemize.html \
pypinyin==0.50.0 \
pytest \
sentencepiece>=0.1.96 \
six \
tensorboard \
typeguard
# RUN git clone https://github.com/k2-fsa/icefall /workspace/icefall && \
# cd /workspace/icefall && \
# pip install --no-cache-dir -r requirements.txt
#
# ENV PYTHONPATH /workspace/icefall:$PYTHONPATH
#
# WORKDIR /workspace/icefall

94
.github/scripts/docker/generate_build_matrix.py vendored Executable file
View File

@ -0,0 +1,94 @@
#!/usr/bin/env python3
# Copyright 2023 Xiaomi Corp. (authors: Fangjun Kuang)
import json
def version_gt(a, b):
a_major, a_minor = list(map(int, a.split(".")))[:2]
b_major, b_minor = list(map(int, b.split(".")))[:2]
if a_major > b_major:
return True
if a_major == b_major and a_minor > b_minor:
return True
return False
def version_ge(a, b):
a_major, a_minor = list(map(int, a.split(".")))[:2]
b_major, b_minor = list(map(int, b.split(".")))[:2]
if a_major > b_major:
return True
if a_major == b_major and a_minor >= b_minor:
return True
return False
def get_torchaudio_version(torch_version):
if torch_version == "1.13.0":
return "0.13.0"
elif torch_version == "1.13.1":
return "0.13.1"
elif torch_version == "2.0.0":
return "2.0.1"
elif torch_version == "2.0.1":
return "2.0.2"
else:
return torch_version
def get_matrix():
k2_version = "1.24.4.dev20240223"
kaldifeat_version = "1.25.4.dev20240223"
version = "20240401"
python_version = ["3.8", "3.9", "3.10", "3.11", "3.12"]
torch_version = []
torch_version += ["1.13.0", "1.13.1"]
torch_version += ["2.0.0", "2.0.1"]
torch_version += ["2.1.0", "2.1.1", "2.1.2"]
torch_version += ["2.2.0", "2.2.1", "2.2.2"]
matrix = []
for p in python_version:
for t in torch_version:
# torchaudio <= 1.13.x supports only python <= 3.10
if version_gt(p, "3.10") and not version_gt(t, "2.0"):
continue
# only torch>=2.2.0 supports python 3.12
if version_gt(p, "3.11") and not version_gt(t, "2.1"):
continue
k2_version_2 = k2_version
kaldifeat_version_2 = kaldifeat_version
if t == "2.2.2":
k2_version_2 = "1.24.4.dev20240328"
kaldifeat_version_2 = "1.25.4.dev20240329"
matrix.append(
{
"k2-version": k2_version_2,
"kaldifeat-version": kaldifeat_version_2,
"version": version,
"python-version": p,
"torch-version": t,
"torchaudio-version": get_torchaudio_version(t),
}
)
return matrix
def main():
matrix = get_matrix()
print(json.dumps({"include": matrix}))
if __name__ == "__main__":
main()

29
.github/scripts/generate-piper-phonemize-page.py vendored Executable file
View File

@ -0,0 +1,29 @@
#!/usr/bin/env python3
def main():
prefix = (
"https://github.com/csukuangfj/piper-phonemize/releases/download/2023.12.5/"
)
files = [
"piper_phonemize-1.2.0-cp310-cp310-macosx_10_14_x86_64.whl",
"piper_phonemize-1.2.0-cp310-cp310-manylinux_2_17_x86_64.manylinux2014_x86_64.whl",
"piper_phonemize-1.2.0-cp311-cp311-macosx_10_14_x86_64.whl",
"piper_phonemize-1.2.0-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl",
"piper_phonemize-1.2.0-cp312-cp312-macosx_10_14_x86_64.whl",
"piper_phonemize-1.2.0-cp312-cp312-manylinux_2_17_x86_64.manylinux2014_x86_64.whl",
"piper_phonemize-1.2.0-cp37-cp37m-macosx_10_14_x86_64.whl",
"piper_phonemize-1.2.0-cp37-cp37m-manylinux_2_17_x86_64.manylinux2014_x86_64.whl",
"piper_phonemize-1.2.0-cp38-cp38-macosx_10_14_x86_64.whl",
"piper_phonemize-1.2.0-cp38-cp38-manylinux_2_17_x86_64.manylinux2014_x86_64.whl",
"piper_phonemize-1.2.0-cp39-cp39-macosx_10_14_x86_64.whl",
"piper_phonemize-1.2.0-cp39-cp39-manylinux_2_17_x86_64.manylinux2014_x86_64.whl",
]
with open("piper_phonemize.html", "w") as f:
for file in files:
url = prefix + file
f.write(f'<a href="{url}">{file}</a><br/>\n')
if __name__ == "__main__":
main()

1644
.github/scripts/librispeech/ASR/run.sh vendored Executable file
View File

File diff suppressed because it is too large Load Diff

157
.github/scripts/ljspeech/TTS/run.sh vendored Executable file
View File

@ -0,0 +1,157 @@
#!/usr/bin/env bash
set -ex
python3 -m pip install piper_phonemize -f https://k2-fsa.github.io/icefall/piper_phonemize.html
python3 -m pip install espnet_tts_frontend
python3 -m pip install numba
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]}) $*"
}
cd egs/ljspeech/TTS
sed -i.bak s/600/8/g ./prepare.sh
sed -i.bak s/"first 100"/"first 3"/g ./prepare.sh
sed -i.bak s/500/5/g ./prepare.sh
git diff
function prepare_data() {
# We have created a subset of the data for testing
#
mkdir download
pushd download
wget -q https://huggingface.co/csukuangfj/ljspeech-subset-for-ci-test/resolve/main/LJSpeech-1.1.tar.bz2
tar xvf LJSpeech-1.1.tar.bz2
popd
./prepare.sh
tree .
}
function train() {
pushd ./vits
sed -i.bak s/200/3/g ./train.py
git diff .
popd
for t in low medium high; do
./vits/train.py \
--exp-dir vits/exp-$t \
--model-type $t \
--num-epochs 1 \
--save-every-n 1 \
--num-buckets 2 \
--tokens data/tokens.txt \
--max-duration 20
ls -lh vits/exp-$t
done
}
function infer() {
for t in low medium high; do
./vits/infer.py \
--num-buckets 2 \
--model-type $t \
--epoch 1 \
--exp-dir ./vits/exp-$t \
--tokens data/tokens.txt \
--max-duration 20
done
}
function export_onnx() {
for t in low medium high; do
./vits/export-onnx.py \
--model-type $t \
--epoch 1 \
--exp-dir ./vits/exp-$t \
--tokens data/tokens.txt
ls -lh vits/exp-$t/
done
}
function test_medium() {
git clone https://huggingface.co/csukuangfj/icefall-tts-ljspeech-vits-medium-2024-03-12
./vits/export-onnx.py \
--model-type medium \
--epoch 820 \
--exp-dir ./icefall-tts-ljspeech-vits-medium-2024-03-12/exp \
--tokens ./icefall-tts-ljspeech-vits-medium-2024-03-12/data/tokens.txt
ls -lh ./icefall-tts-ljspeech-vits-medium-2024-03-12/exp
./vits/test_onnx.py \
--model-filename ./icefall-tts-ljspeech-vits-medium-2024-03-12/exp/vits-epoch-820.onnx \
--tokens ./icefall-tts-ljspeech-vits-medium-2024-03-12/data/tokens.txt \
--output-filename /icefall/test-medium.wav
ls -lh /icefall/test-medium.wav
d=/icefall/vits-icefall-en_US-ljspeech-medium
mkdir $d
cp -v ./icefall-tts-ljspeech-vits-medium-2024-03-12/data/tokens.txt $d/
cp -v ./icefall-tts-ljspeech-vits-medium-2024-03-12/exp/vits-epoch-820.onnx $d/model.onnx
rm -rf icefall-tts-ljspeech-vits-medium-2024-03-12
pushd $d
wget -q https://github.com/k2-fsa/sherpa-onnx/releases/download/tts-models/espeak-ng-data.tar.bz2
tar xf espeak-ng-data.tar.bz2
rm espeak-ng-data.tar.bz2
cd ..
tar cjf vits-icefall-en_US-ljspeech-medium.tar.bz2 vits-icefall-en_US-ljspeech-medium
rm -rf vits-icefall-en_US-ljspeech-medium
ls -lh *.tar.bz2
popd
}
function test_low() {
git clone https://huggingface.co/csukuangfj/icefall-tts-ljspeech-vits-low-2024-03-12
./vits/export-onnx.py \
--model-type low \
--epoch 1600 \
--exp-dir ./icefall-tts-ljspeech-vits-low-2024-03-12/exp \
--tokens ./icefall-tts-ljspeech-vits-low-2024-03-12/data/tokens.txt
ls -lh ./icefall-tts-ljspeech-vits-low-2024-03-12/exp
./vits/test_onnx.py \
--model-filename ./icefall-tts-ljspeech-vits-low-2024-03-12/exp/vits-epoch-1600.onnx \
--tokens ./icefall-tts-ljspeech-vits-low-2024-03-12/data/tokens.txt \
--output-filename /icefall/test-low.wav
ls -lh /icefall/test-low.wav
d=/icefall/vits-icefall-en_US-ljspeech-low
mkdir $d
cp -v ./icefall-tts-ljspeech-vits-low-2024-03-12/data/tokens.txt $d/
cp -v ./icefall-tts-ljspeech-vits-low-2024-03-12/exp/vits-epoch-1600.onnx $d/model.onnx
rm -rf icefall-tts-ljspeech-vits-low-2024-03-12
pushd $d
wget -q https://github.com/k2-fsa/sherpa-onnx/releases/download/tts-models/espeak-ng-data.tar.bz2
tar xf espeak-ng-data.tar.bz2
rm espeak-ng-data.tar.bz2
cd ..
tar cjf vits-icefall-en_US-ljspeech-low.tar.bz2 vits-icefall-en_US-ljspeech-low
rm -rf vits-icefall-en_US-ljspeech-low
ls -lh *.tar.bz2
popd
}
prepare_data
train
infer
export_onnx
rm -rf vits/exp-{low,medium,high}
test_medium
test_low

158
.github/scripts/multi-zh-hans.sh vendored Executable file
View File

@ -0,0 +1,158 @@
#!/usr/bin/env bash
set -ex
git config --global user.name "k2-fsa"
git config --global user.email "csukuangfj@gmail.com"
git config --global lfs.allowincompletepush true
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]}) $*"
}
log "pwd: $PWD"
cd egs/multi_zh-hans/ASR
repo_url=https://huggingface.co/zrjin/icefall-asr-multi-zh-hans-zipformer-ctc-streaming-2023-11-05
log "Downloading pre-trained model from $repo_url"
GIT_LFS_SKIP_SMUDGE=1 git clone $repo_url
repo=$(basename $repo_url)
pushd $repo
cd exp/
git lfs pull --include pretrained.pt
rm -fv epoch-20.pt
rm -fv *.onnx
ln -s pretrained.pt epoch-20.pt
cd ../data/lang_bpe_2000
ls -lh
git lfs pull --include L.pt L_disambig.pt Linv.pt bpe.model
git lfs pull --include "*.model"
ls -lh
popd
log "----------------------------------------"
log "Export streaming ONNX CTC models "
log "----------------------------------------"
./zipformer/export-onnx-streaming-ctc.py \
--exp-dir $repo/exp \
--tokens $repo/data/lang_bpe_2000/tokens.txt \
--causal 1 \
--avg 1 \
--epoch 20 \
--use-averaged-model 0 \
--chunk-size 16 \
--left-context-frames 128 \
--use-ctc 1
ls -lh $repo/exp/
log "------------------------------------------------------------"
log "Test exported streaming ONNX CTC models (greedy search) "
log "------------------------------------------------------------"
test_wavs=(
DEV_T0000000000.wav
DEV_T0000000001.wav
DEV_T0000000002.wav
TEST_MEETING_T0000000113.wav
TEST_MEETING_T0000000219.wav
TEST_MEETING_T0000000351.wav
)
for w in ${test_wavs[@]}; do
./zipformer/onnx_pretrained-streaming-ctc.py \
--model-filename $repo/exp/ctc-epoch-20-avg-1-chunk-16-left-128.int8.onnx \
--tokens $repo/data/lang_bpe_2000/tokens.txt \
$repo/test_wavs/$w
done
log "Upload onnx CTC models to huggingface"
url=https://huggingface.co/k2-fsa/sherpa-onnx-streaming-zipformer-ctc-multi-zh-hans-2023-12-13
GIT_LFS_SKIP_SMUDGE=1 git clone $url
dst=$(basename $url)
cp -v $repo/exp/ctc*.onnx $dst
cp -v $repo/data/lang_bpe_2000/tokens.txt $dst
cp -v $repo/data/lang_bpe_2000/bpe.model $dst
mkdir -p $dst/test_wavs
cp -v $repo/test_wavs/*.wav $dst/test_wavs
cd $dst
git lfs track "*.onnx" "bpe.model"
ls -lh
file bpe.model
git status
git add .
git commit -m "upload model" && git push https://k2-fsa:${HF_TOKEN}@huggingface.co/k2-fsa/$dst main || true
log "Upload models to https://github.com/k2-fsa/sherpa-onnx"
rm -rf .git
rm -fv .gitattributes
cd ..
tar cjfv $dst.tar.bz2 $dst
ls -lh *.tar.bz2
mv -v $dst.tar.bz2 ../../../
log "----------------------------------------"
log "Export streaming ONNX transducer models "
log "----------------------------------------"
./zipformer/export-onnx-streaming.py \
--exp-dir $repo/exp \
--tokens $repo/data/lang_bpe_2000/tokens.txt \
--causal 1 \
--avg 1 \
--epoch 20 \
--use-averaged-model 0 \
--chunk-size 16 \
--left-context-frames 128 \
--use-ctc 0
ls -lh $repo/exp
log "------------------------------------------------------------"
log "Test exported streaming ONNX transducer models (Python code)"
log "------------------------------------------------------------"
log "test fp32"
./zipformer/onnx_pretrained-streaming.py \
--encoder-model-filename $repo/exp/encoder-epoch-20-avg-1-chunk-16-left-128.onnx \
--decoder-model-filename $repo/exp/decoder-epoch-20-avg-1-chunk-16-left-128.onnx \
--joiner-model-filename $repo/exp/joiner-epoch-20-avg-1-chunk-16-left-128.onnx \
--tokens $repo/data/lang_bpe_2000/tokens.txt \
$repo/test_wavs/DEV_T0000000000.wav
log "test int8"
./zipformer/onnx_pretrained-streaming.py \
--encoder-model-filename $repo/exp/encoder-epoch-20-avg-1-chunk-16-left-128.int8.onnx \
--decoder-model-filename $repo/exp/decoder-epoch-20-avg-1-chunk-16-left-128.onnx \
--joiner-model-filename $repo/exp/joiner-epoch-20-avg-1-chunk-16-left-128.int8.onnx \
--tokens $repo/data/lang_bpe_2000/tokens.txt \
$repo/test_wavs/DEV_T0000000000.wav
log "Upload onnx transducer models to huggingface"
url=https://huggingface.co/k2-fsa/sherpa-onnx-streaming-zipformer-multi-zh-hans-2023-12-12
GIT_LFS_SKIP_SMUDGE=1 git clone $url
dst=$(basename $url)
cp -v $repo/exp/encoder*.onnx $dst
cp -v $repo/exp/decoder*.onnx $dst
cp -v $repo/exp/joiner*.onnx $dst
cp -v $repo/data/lang_bpe_2000/tokens.txt $dst
cp -v $repo/data/lang_bpe_2000/bpe.model $dst
mkdir -p $dst/test_wavs
cp -v $repo/test_wavs/*.wav $dst/test_wavs
cd $dst
git lfs track "*.onnx" bpe.model
git add .
git commit -m "upload model" && git push https://k2-fsa:${HF_TOKEN}@huggingface.co/k2-fsa/$dst main || true
log "Upload models to https://github.com/k2-fsa/sherpa-onnx"
rm -rf .git
rm -fv .gitattributes
cd ..
tar cjfv $dst.tar.bz2 $dst
ls -lh *.tar.bz2
mv -v $dst.tar.bz2 ../../../

87
.github/scripts/run-aishell-pruned-transducer-stateless3-2022-06-20.sh vendored
View File

@ -1,87 +0,0 @@
#!/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]}) $*"
}
cd egs/aishell/ASR
git lfs install
fbank_url=https://huggingface.co/csukuangfj/aishell-test-dev-manifests
log "Downloading pre-commputed fbank from $fbank_url"
git clone https://huggingface.co/csukuangfj/aishell-test-dev-manifests
ln -s $PWD/aishell-test-dev-manifests/data .
repo_url=https://huggingface.co/csukuangfj/icefall-aishell-pruned-transducer-stateless3-2022-06-20
log "Downloading pre-trained model from $repo_url"
git clone $repo_url
repo=$(basename $repo_url)
log "Display test files"
tree $repo/
ls -lh $repo/test_wavs/*.wav
pushd $repo/exp
ln -s pretrained-epoch-29-avg-5-torch-1.10.0.pt pretrained.pt
popd
for sym in 1 2 3; do
log "Greedy search with --max-sym-per-frame $sym"
./pruned_transducer_stateless3/pretrained.py \
--method greedy_search \
--max-sym-per-frame $sym \
--checkpoint $repo/exp/pretrained.pt \
--lang-dir $repo/data/lang_char \
$repo/test_wavs/BAC009S0764W0121.wav \
$repo/test_wavs/BAC009S0764W0122.wav \
$repo/test_wavs/BAC009S0764W0123.wav
done
for method in modified_beam_search beam_search fast_beam_search; do
log "$method"
./pruned_transducer_stateless3/pretrained.py \
--method $method \
--beam-size 4 \
--checkpoint $repo/exp/pretrained.pt \
--lang-dir $repo/data/lang_char \
$repo/test_wavs/BAC009S0764W0121.wav \
$repo/test_wavs/BAC009S0764W0122.wav \
$repo/test_wavs/BAC009S0764W0123.wav
done
echo "GITHUB_EVENT_NAME: ${GITHUB_EVENT_NAME}"
echo "GITHUB_EVENT_LABEL_NAME: ${GITHUB_EVENT_LABEL_NAME}"
if [[ x"${GITHUB_EVENT_NAME}" == x"schedule" || x"${GITHUB_EVENT_LABEL_NAME}" == x"run-decode" ]]; then
mkdir -p pruned_transducer_stateless3/exp
ln -s $PWD/$repo/exp/pretrained.pt pruned_transducer_stateless3/exp/epoch-999.pt
ln -s $PWD/$repo/data/lang_char data/
ls -lh data
ls -lh pruned_transducer_stateless3/exp
log "Decoding test and dev"
# use a small value for decoding with CPU
max_duration=100
for method in greedy_search fast_beam_search modified_beam_search; do
log "Decoding with $method"
./pruned_transducer_stateless3/decode.py \
--decoding-method $method \
--epoch 999 \
--avg 1 \
--max-duration $max_duration \
--exp-dir pruned_transducer_stateless3/exp
done
rm pruned_transducer_stateless3/exp/*.pt
fi

103
.github/scripts/run-aishell-zipformer-2023-10-24.sh vendored
View File

@ -1,103 +0,0 @@
#!/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]}) $*"
}
cd egs/aishell/ASR
git lfs install
fbank_url=https://huggingface.co/csukuangfj/aishell-test-dev-manifests
log "Downloading pre-commputed fbank from $fbank_url"
git clone https://huggingface.co/csukuangfj/aishell-test-dev-manifests
ln -s $PWD/aishell-test-dev-manifests/data .
log "======================="
log "CI testing large model"
repo_url=https://huggingface.co/zrjin/icefall-asr-aishell-zipformer-large-2023-10-24/
log "Downloading pre-trained model from $repo_url"
git clone $repo_url
repo=$(basename $repo_url)
log "Display test files"
tree $repo/
ls -lh $repo/test_wavs/*.wav
for method in modified_beam_search greedy_search fast_beam_search; do
log "$method"
./zipformer/pretrained.py \
--method $method \
--context-size 1 \
--checkpoint $repo/exp/pretrained.pt \
--tokens $repo/data/lang_char/tokens.txt \
--num-encoder-layers 2,2,4,5,4,2 \
--feedforward-dim 512,768,1536,2048,1536,768 \
--encoder-dim 192,256,512,768,512,256 \
--encoder-unmasked-dim 192,192,256,320,256,192 \
$repo/test_wavs/BAC009S0764W0121.wav \
$repo/test_wavs/BAC009S0764W0122.wav \
$repo/test_wavs/BAC009S0764W0123.wav
done
log "======================="
log "CI testing medium model"
repo_url=https://huggingface.co/zrjin/icefall-asr-aishell-zipformer-2023-10-24/
log "Downloading pre-trained model from $repo_url"
git clone $repo_url
repo=$(basename $repo_url)
log "Display test files"
tree $repo/
ls -lh $repo/test_wavs/*.wav
for method in modified_beam_search greedy_search fast_beam_search; do
log "$method"
./zipformer/pretrained.py \
--method $method \
--context-size 1 \
--checkpoint $repo/exp/pretrained.pt \
--tokens $repo/data/lang_char/tokens.txt \
$repo/test_wavs/BAC009S0764W0121.wav \
$repo/test_wavs/BAC009S0764W0122.wav \
$repo/test_wavs/BAC009S0764W0123.wav
done
log "======================="
log "CI testing small model"
repo_url=https://huggingface.co/zrjin/icefall-asr-aishell-zipformer-small-2023-10-24/
log "Downloading pre-trained model from $repo_url"
git clone $repo_url
repo=$(basename $repo_url)
log "Display test files"
tree $repo/
ls -lh $repo/test_wavs/*.wav
for method in modified_beam_search greedy_search fast_beam_search; do
log "$method"
./zipformer/pretrained.py \
--method $method \
--context-size 1 \
--checkpoint $repo/exp/pretrained.pt \
--tokens $repo/data/lang_char/tokens.txt \
--num-encoder-layers 2,2,2,2,2,2 \
--feedforward-dim 512,768,768,768,768,768 \
--encoder-dim 192,256,256,256,256,256 \
--encoder-unmasked-dim 192,192,192,192,192,192 \
$repo/test_wavs/BAC009S0764W0121.wav \
$repo/test_wavs/BAC009S0764W0122.wav \
$repo/test_wavs/BAC009S0764W0123.wav
done

74
.github/scripts/run-gigaspeech-zipformer-2023-10-17.sh vendored
View File

@ -26,16 +26,80 @@ git lfs pull --include "data/lang_bpe_500/bpe.model"
git lfs pull --include "data/lang_bpe_500/tokens.txt"
git lfs pull --include "exp/jit_script.pt"
git lfs pull --include "exp/pretrained.pt"
ln -s pretrained.pt epoch-99.pt
ls -lh *.pt
rm epoch-30.pt
ln -s pretrained.pt epoch-30.pt
rm *.onnx
ls -lh
popd
log "----------------------------------------"
log "Export ONNX transducer models "
log "----------------------------------------"
./zipformer/export-onnx.py \
--tokens $repo/data/lang_bpe_500/tokens.txt \
--use-averaged-model 0 \
--epoch 30 \
--avg 1 \
--exp-dir $repo/exp
ls -lh $repo/exp
log "------------------------------------------------------------"
log "Test exported ONNX transducer models (Python code) "
log "------------------------------------------------------------"
log "test fp32"
./zipformer/onnx_pretrained.py \
--encoder-model-filename $repo/exp/encoder-epoch-30-avg-1.onnx \
--decoder-model-filename $repo/exp/decoder-epoch-30-avg-1.onnx \
--joiner-model-filename $repo/exp/joiner-epoch-30-avg-1.onnx \
--tokens $repo/data/lang_bpe_500/tokens.txt \
$repo/test_wavs/1089-134686-0001.wav \
$repo/test_wavs/1221-135766-0001.wav \
$repo/test_wavs/1221-135766-0002.wav
log "test int8"
./zipformer/onnx_pretrained.py \
--encoder-model-filename $repo/exp/encoder-epoch-30-avg-1.int8.onnx \
--decoder-model-filename $repo/exp/decoder-epoch-30-avg-1.onnx \
--joiner-model-filename $repo/exp/joiner-epoch-30-avg-1.int8.onnx \
--tokens $repo/data/lang_bpe_500/tokens.txt \
$repo/test_wavs/1089-134686-0001.wav \
$repo/test_wavs/1221-135766-0001.wav \
$repo/test_wavs/1221-135766-0002.wav
log "Upload models to huggingface"
git config --global user.name "k2-fsa"
git config --global user.email "xxx@gmail.com"
url=https://huggingface.co/k2-fsa/sherpa-onnx-zipformer-gigaspeech-2023-12-12
GIT_LFS_SKIP_SMUDGE=1 git clone $url
dst=$(basename $url)
cp -v $repo/exp/*.onnx $dst
cp -v $repo/data/lang_bpe_500/tokens.txt $dst
cp -v $repo/data/lang_bpe_500/bpe.model $dst
mkdir -p $dst/test_wavs
cp -v $repo/test_wavs/*.wav $dst/test_wavs
cd $dst
git lfs track "*.onnx"
git add .
git commit -m "upload model" && git push https://k2-fsa:${HF_TOKEN}@huggingface.co/k2-fsa/$dst main || true
log "Upload models to https://github.com/k2-fsa/sherpa-onnx"
rm -rf .git
rm -fv .gitattributes
cd ..
tar cjfv $dst.tar.bz2 $dst
ls -lh
mv -v $dst.tar.bz2 ../../../
log "Export to torchscript model"
./zipformer/export.py \
--exp-dir $repo/exp \
--use-averaged-model false \
--tokens $repo/data/lang_bpe_500/tokens.txt \
--epoch 99 \
--epoch 30 \
--avg 1 \
--jit 1
@ -67,7 +131,7 @@ echo "GITHUB_EVENT_NAME: ${GITHUB_EVENT_NAME}"
echo "GITHUB_EVENT_LABEL_NAME: ${GITHUB_EVENT_LABEL_NAME}"
if [[ x"${GITHUB_EVENT_NAME}" == x"schedule" || x"${GITHUB_EVENT_LABEL_NAME}" == x"run-decode" ]]; then
mkdir -p zipformer/exp
ln -s $PWD/$repo/exp/pretrained.pt zipformer/exp/epoch-999.pt
ln -s $PWD/$repo/exp/pretrained.pt zipformer/exp/epoch-30.pt
ln -s $PWD/$repo/data/lang_bpe_500 data/
ls -lh data
@ -83,7 +147,7 @@ if [[ x"${GITHUB_EVENT_NAME}" == x"schedule" || x"${GITHUB_EVENT_LABEL_NAME}" ==
./zipformer/decode.py \
--decoding-method $method \
--epoch 999 \
--epoch 30 \
--avg 1 \
--use-averaged-model 0 \
--max-duration $max_duration \

122
.github/scripts/run-librispeech-conformer-ctc3-2022-11-28.sh vendored
View File

@ -1,122 +0,0 @@
#!/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]}) $*"
}
cd egs/librispeech/ASR
repo_url=https://huggingface.co/Zengwei/icefall-asr-librispeech-conformer-ctc3-2022-11-27
log "Downloading pre-trained model from $repo_url"
GIT_LFS_SKIP_SMUDGE=1 git clone $repo_url
repo=$(basename $repo_url)
log "Display test files"
tree $repo/
ls -lh $repo/test_wavs/*.wav
pushd $repo/exp
git lfs pull --include "data/lang_bpe_500/HLG.pt"
git lfs pull --include "data/lang_bpe_500/L.pt"
git lfs pull --include "data/lang_bpe_500/LG.pt"
git lfs pull --include "data/lang_bpe_500/Linv.pt"
git lfs pull --include "data/lang_bpe_500/bpe.model"
git lfs pull --include "data/lm/G_4_gram.pt"
git lfs pull --include "exp/jit_trace.pt"
git lfs pull --include "exp/pretrained.pt"
ln -s pretrained.pt epoch-99.pt
ls -lh *.pt
popd
log "Decode with models exported by torch.jit.trace()"
for m in ctc-decoding 1best; do
./conformer_ctc3/jit_pretrained.py \
--model-filename $repo/exp/jit_trace.pt \
--words-file $repo/data/lang_bpe_500/words.txt \
--HLG $repo/data/lang_bpe_500/HLG.pt \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
--G $repo/data/lm/G_4_gram.pt \
--method $m \
--sample-rate 16000 \
$repo/test_wavs/1089-134686-0001.wav \
$repo/test_wavs/1221-135766-0001.wav \
$repo/test_wavs/1221-135766-0002.wav
done
log "Export to torchscript model"
./conformer_ctc3/export.py \
--exp-dir $repo/exp \
--tokens $repo/data/lang_bpe_500/tokens.txt \
--jit-trace 1 \
--epoch 99 \
--avg 1 \
--use-averaged-model 0
ls -lh $repo/exp/*.pt
log "Decode with models exported by torch.jit.trace()"
for m in ctc-decoding 1best; do
./conformer_ctc3/jit_pretrained.py \
--model-filename $repo/exp/jit_trace.pt \
--words-file $repo/data/lang_bpe_500/words.txt \
--HLG $repo/data/lang_bpe_500/HLG.pt \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
--G $repo/data/lm/G_4_gram.pt \
--method $m \
--sample-rate 16000 \
$repo/test_wavs/1089-134686-0001.wav \
$repo/test_wavs/1221-135766-0001.wav \
$repo/test_wavs/1221-135766-0002.wav
done
for m in ctc-decoding 1best; do
./conformer_ctc3/pretrained.py \
--checkpoint $repo/exp/pretrained.pt \
--words-file $repo/data/lang_bpe_500/words.txt \
--HLG $repo/data/lang_bpe_500/HLG.pt \
--tokens $repo/data/lang_bpe_500/tokens.txt \
--G $repo/data/lm/G_4_gram.pt \
--method $m \
--sample-rate 16000 \
$repo/test_wavs/1089-134686-0001.wav \
$repo/test_wavs/1221-135766-0001.wav \
$repo/test_wavs/1221-135766-0002.wav
done
echo "GITHUB_EVENT_NAME: ${GITHUB_EVENT_NAME}"
echo "GITHUB_EVENT_LABEL_NAME: ${GITHUB_EVENT_LABEL_NAME}"
if [[ x"${GITHUB_EVENT_NAME}" == x"schedule" || x"${GITHUB_EVENT_LABEL_NAME}" == x"run-decode" ]]; then
mkdir -p conformer_ctc3/exp
ln -s $PWD/$repo/exp/pretrained.pt conformer_ctc3/exp/epoch-999.pt
ln -s $PWD/$repo/data/lang_bpe_500 data/
ls -lh data
ls -lh conformer_ctc3/exp
log "Decoding test-clean and test-other"
# use a small value for decoding with CPU
max_duration=100
for method in ctc-decoding 1best; do
log "Decoding with $method"
./conformer_ctc3/decode.py \
--epoch 999 \
--avg 1 \
--use-averaged-model 0 \
--exp-dir conformer_ctc3/exp/ \
--max-duration $max_duration \
--decoding-method $method \
--lm-dir data/lm
done
rm conformer_ctc3/exp/*.pt
fi

77
.github/scripts/run-librispeech-pruned-transducer-stateless-2022-03-12.sh vendored
View File

@ -1,77 +0,0 @@
#!/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]}) $*"
}
cd egs/librispeech/ASR
repo_url=https://huggingface.co/csukuangfj/icefall-asr-librispeech-pruned-transducer-stateless-2022-03-12
log "Downloading pre-trained model from $repo_url"
git lfs install
git clone $repo_url
repo=$(basename $repo_url)
log "Display test files"
tree $repo/
ls -lh $repo/test_wavs/*.wav
for sym in 1 2 3; do
log "Greedy search with --max-sym-per-frame $sym"
./pruned_transducer_stateless/pretrained.py \
--method greedy_search \
--max-sym-per-frame $sym \
--checkpoint $repo/exp/pretrained.pt \
--tokens $repo/data/lang_bpe_500/tokens.txt \
$repo/test_wavs/1089-134686-0001.wav \
$repo/test_wavs/1221-135766-0001.wav \
$repo/test_wavs/1221-135766-0002.wav
done
for method in fast_beam_search modified_beam_search beam_search; do
log "$method"
./pruned_transducer_stateless/pretrained.py \
--method $method \
--beam-size 4 \
--checkpoint $repo/exp/pretrained.pt \
--tokens $repo/data/lang_bpe_500/tokens.txt \
$repo/test_wavs/1089-134686-0001.wav \
$repo/test_wavs/1221-135766-0001.wav \
$repo/test_wavs/1221-135766-0002.wav
done
echo "GITHUB_EVENT_NAME: ${GITHUB_EVENT_NAME}"
echo "GITHUB_EVENT_LABEL_NAME: ${GITHUB_EVENT_LABEL_NAME}"
if [[ x"${GITHUB_EVENT_NAME}" == x"schedule" || x"${GITHUB_EVENT_LABEL_NAME}" == x"run-decode" ]]; then
mkdir -p pruned_transducer_stateless/exp
ln -s $PWD/$repo/exp/pretrained.pt pruned_transducer_stateless/exp/epoch-999.pt
ln -s $PWD/$repo/data/lang_bpe_500 data/
ls -lh data
ls -lh pruned_transducer_stateless/exp
log "Decoding test-clean and test-other"
# use a small value for decoding with CPU
max_duration=100
for method in greedy_search fast_beam_search modified_beam_search; do
log "Decoding with $method"
./pruned_transducer_stateless/decode.py \
--decoding-method $method \
--epoch 999 \
--avg 1 \
--max-duration $max_duration \
--exp-dir pruned_transducer_stateless/exp
done
rm pruned_transducer_stateless/exp/*.pt
fi

86
.github/scripts/run-librispeech-pruned-transducer-stateless2-2022-04-29.sh vendored
View File

@ -1,86 +0,0 @@
#!/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]}) $*"
}
cd egs/librispeech/ASR
repo_url=https://huggingface.co/csukuangfj/icefall-asr-librispeech-pruned-transducer-stateless2-2022-04-29
log "Downloading pre-trained model from $repo_url"
GIT_LFS_SKIP_SMUDGE=1 git clone $repo_url
repo=$(basename $repo_url)
pushd $repo
git lfs pull --include "data/lang_bpe_500/bpe.model"
git lfs pull --include "exp/pretrained-epoch-38-avg-10.pt"
popd
log "Display test files"
tree $repo/
ls -lh $repo/test_wavs/*.wav
pushd $repo/exp
ln -s pretrained-epoch-38-avg-10.pt pretrained.pt
popd
for sym in 1 2 3; do
log "Greedy search with --max-sym-per-frame $sym"
./pruned_transducer_stateless2/pretrained.py \
--method greedy_search \
--max-sym-per-frame $sym \
--checkpoint $repo/exp/pretrained.pt \
--tokens $repo/data/lang_bpe_500/tokens.txt \
$repo/test_wavs/1089-134686-0001.wav \
$repo/test_wavs/1221-135766-0001.wav \
$repo/test_wavs/1221-135766-0002.wav
done
for method in modified_beam_search beam_search fast_beam_search; do
log "$method"
./pruned_transducer_stateless2/pretrained.py \
--method $method \
--beam-size 4 \
--checkpoint $repo/exp/pretrained.pt \
--tokens $repo/data/lang_bpe_500/tokens.txt \
$repo/test_wavs/1089-134686-0001.wav \
$repo/test_wavs/1221-135766-0001.wav \
$repo/test_wavs/1221-135766-0002.wav
done
echo "GITHUB_EVENT_NAME: ${GITHUB_EVENT_NAME}"
echo "GITHUB_EVENT_LABEL_NAME: ${GITHUB_EVENT_LABEL_NAME}"
if [[ x"${GITHUB_EVENT_NAME}" == x"schedule" || x"${GITHUB_EVENT_LABEL_NAME}" == x"run-decode" ]]; then
mkdir -p pruned_transducer_stateless2/exp
ln -s $PWD/$repo/exp/pretrained.pt pruned_transducer_stateless2/exp/epoch-999.pt
ln -s $PWD/$repo/data/lang_bpe_500 data/
ls -lh data
ls -lh pruned_transducer_stateless2/exp
log "Decoding test-clean and test-other"
# use a small value for decoding with CPU
max_duration=100
for method in greedy_search fast_beam_search modified_beam_search; do
log "Decoding with $method"
./pruned_transducer_stateless2/decode.py \
--decoding-method $method \
--epoch 999 \
--avg 1 \
--max-duration $max_duration \
--exp-dir pruned_transducer_stateless2/exp
done
rm pruned_transducer_stateless2/exp/*.pt
rm -r data/lang_bpe_500
fi

85
.github/scripts/run-librispeech-pruned-transducer-stateless3-2022-04-29.sh vendored
View File

@ -1,85 +0,0 @@
#!/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]}) $*"
}
cd egs/librispeech/ASR
repo_url=https://huggingface.co/csukuangfj/icefall-asr-librispeech-pruned-transducer-stateless3-2022-04-29
log "Downloading pre-trained model from $repo_url"
GIT_LFS_SKIP_SMUDGE=1 git clone $repo_url
repo=$(basename $repo_url)
pushd $repo
git lfs pull --include "data/lang_bpe_500/bpe.model"
git lfs pull --include "exp/pretrained-epoch-25-avg-6.pt"
popd
log "Display test files"
tree $repo/
ls -lh $repo/test_wavs/*.wav
pushd $repo/exp
ln -s pretrained-epoch-25-avg-6.pt pretrained.pt
popd
for sym in 1 2 3; do
log "Greedy search with --max-sym-per-frame $sym"
./pruned_transducer_stateless3/pretrained.py \
--method greedy_search \
--max-sym-per-frame $sym \
--checkpoint $repo/exp/pretrained.pt \
--tokens $repo/data/lang_bpe_500/tokens.txt \
$repo/test_wavs/1089-134686-0001.wav \
$repo/test_wavs/1221-135766-0001.wav \
$repo/test_wavs/1221-135766-0002.wav
done
for method in modified_beam_search beam_search fast_beam_search; do
log "$method"
./pruned_transducer_stateless3/pretrained.py \
--method $method \
--beam-size 4 \
--checkpoint $repo/exp/pretrained.pt \
--tokens $repo/data/lang_bpe_500/tokens.txt \
$repo/test_wavs/1089-134686-0001.wav \
$repo/test_wavs/1221-135766-0001.wav \
$repo/test_wavs/1221-135766-0002.wav
done
echo "GITHUB_EVENT_NAME: ${GITHUB_EVENT_NAME}"
echo "GITHUB_EVENT_LABEL_NAME: ${GITHUB_EVENT_LABEL_NAME}"
if [[ x"${GITHUB_EVENT_NAME}" == x"schedule" || x"${GITHUB_EVENT_LABEL_NAME}" == x"run-decode" ]]; then
mkdir -p pruned_transducer_stateless3/exp
ln -s $PWD/$repo/exp/pretrained.pt pruned_transducer_stateless3/exp/epoch-999.pt
ln -s $PWD/$repo/data/lang_bpe_500 data/
ls -lh data
ls -lh pruned_transducer_stateless3/exp
log "Decoding test-clean and test-other"
# use a small value for decoding with CPU
max_duration=100
for method in greedy_search fast_beam_search modified_beam_search; do
log "Decoding with $method"
./pruned_transducer_stateless3/decode.py \
--decoding-method $method \
--epoch 999 \
--avg 1 \
--max-duration $max_duration \
--exp-dir pruned_transducer_stateless3/exp
done
rm pruned_transducer_stateless3/exp/*.pt
rm -r data/lang_bpe_500
fi

123
.github/scripts/run-librispeech-pruned-transducer-stateless3-2022-05-13.sh vendored
View File

@ -1,123 +0,0 @@
#!/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]}) $*"
}
cd egs/librispeech/ASR
repo_url=https://huggingface.co/csukuangfj/icefall-asr-librispeech-pruned-transducer-stateless3-2022-05-13
log "Downloading pre-trained model from $repo_url"
git lfs install
git clone $repo_url
repo=$(basename $repo_url)
log "Display test files"
tree $repo/
ls -lh $repo/test_wavs/*.wav
pushd $repo/exp
ln -s pretrained-iter-1224000-avg-14.pt pretrained.pt
ln -s pretrained-iter-1224000-avg-14.pt epoch-99.pt
popd
log "Export to torchscript model"
./pruned_transducer_stateless3/export.py \
--exp-dir $repo/exp \
--tokens $repo/data/lang_bpe_500/tokens.txt \
--epoch 99 \
--avg 1 \
--jit 1
./pruned_transducer_stateless3/export.py \
--exp-dir $repo/exp \
--tokens $repo/data/lang_bpe_500/tokens.txt \
--epoch 99 \
--avg 1 \
--jit-trace 1
ls -lh $repo/exp/*.pt
log "Decode with models exported by torch.jit.trace()"
./pruned_transducer_stateless3/jit_pretrained.py \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
--encoder-model-filename $repo/exp/encoder_jit_trace.pt \
--decoder-model-filename $repo/exp/decoder_jit_trace.pt \
--joiner-model-filename $repo/exp/joiner_jit_trace.pt \
$repo/test_wavs/1089-134686-0001.wav \
$repo/test_wavs/1221-135766-0001.wav \
$repo/test_wavs/1221-135766-0002.wav
log "Decode with models exported by torch.jit.script()"
./pruned_transducer_stateless3/jit_pretrained.py \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
--encoder-model-filename $repo/exp/encoder_jit_script.pt \
--decoder-model-filename $repo/exp/decoder_jit_script.pt \
--joiner-model-filename $repo/exp/joiner_jit_script.pt \
$repo/test_wavs/1089-134686-0001.wav \
$repo/test_wavs/1221-135766-0001.wav \
$repo/test_wavs/1221-135766-0002.wav
for sym in 1 2 3; do
log "Greedy search with --max-sym-per-frame $sym"
./pruned_transducer_stateless3/pretrained.py \
--method greedy_search \
--max-sym-per-frame $sym \
--checkpoint $repo/exp/pretrained.pt \
--tokens $repo/data/lang_bpe_500/tokens.txt \
$repo/test_wavs/1089-134686-0001.wav \
$repo/test_wavs/1221-135766-0001.wav \
$repo/test_wavs/1221-135766-0002.wav
done
for method in modified_beam_search beam_search fast_beam_search; do
log "$method"
./pruned_transducer_stateless3/pretrained.py \
--method $method \
--beam-size 4 \
--checkpoint $repo/exp/pretrained.pt \
--tokens $repo/data/lang_bpe_500/tokens.txt \
$repo/test_wavs/1089-134686-0001.wav \
$repo/test_wavs/1221-135766-0001.wav \
$repo/test_wavs/1221-135766-0002.wav
done
echo "GITHUB_EVENT_NAME: ${GITHUB_EVENT_NAME}"
echo "GITHUB_EVENT_LABEL_NAME: ${GITHUB_EVENT_LABEL_NAME}"
if [[ x"${GITHUB_EVENT_NAME}" == x"schedule" || x"${GITHUB_EVENT_LABEL_NAME}" == x"run-decode" ]]; then
mkdir -p pruned_transducer_stateless3/exp
ln -s $PWD/$repo/exp/pretrained.pt pruned_transducer_stateless3/exp/epoch-999.pt
ln -s $PWD/$repo/data/lang_bpe_500 data/
ls -lh data
ls -lh pruned_transducer_stateless3/exp
log "Decoding test-clean and test-other"
# use a small value for decoding with CPU
max_duration=100
for method in greedy_search fast_beam_search modified_beam_search; do
log "Decoding with $method"
./pruned_transducer_stateless3/decode.py \
--decoding-method $method \
--epoch 999 \
--avg 1 \
--max-duration $max_duration \
--exp-dir pruned_transducer_stateless3/exp
done
rm pruned_transducer_stateless3/exp/*.pt
fi

100
.github/scripts/run-librispeech-pruned-transducer-stateless5-2022-05-13.sh vendored
View File

@ -1,100 +0,0 @@
#!/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]}) $*"
}
cd egs/librispeech/ASR
repo_url=https://huggingface.co/csukuangfj/icefall-asr-librispeech-pruned-transducer-stateless5-2022-05-13
log "Downloading pre-trained model from $repo_url"
git lfs install
git clone $repo_url
repo=$(basename $repo_url)
log "Display test files"
tree $repo/
ls -lh $repo/test_wavs/*.wav
pushd $repo/exp
ln -s pretrained-epoch-39-avg-7.pt pretrained.pt
popd
for sym in 1 2 3; do
log "Greedy search with --max-sym-per-frame $sym"
./pruned_transducer_stateless5/pretrained.py \
--method greedy_search \
--max-sym-per-frame $sym \
--checkpoint $repo/exp/pretrained.pt \
--tokens $repo/data/lang_bpe_500/tokens.txt \
--num-encoder-layers 18 \
--dim-feedforward 2048 \
--nhead 8 \
--encoder-dim 512 \
--decoder-dim 512 \
--joiner-dim 512 \
$repo/test_wavs/1089-134686-0001.wav \
$repo/test_wavs/1221-135766-0001.wav \
$repo/test_wavs/1221-135766-0002.wav
done
for method in modified_beam_search beam_search fast_beam_search; do
log "$method"
./pruned_transducer_stateless5/pretrained.py \
--method $method \
--beam-size 4 \
--checkpoint $repo/exp/pretrained.pt \
--tokens $repo/data/lang_bpe_500/tokens.txt \
$repo/test_wavs/1089-134686-0001.wav \
$repo/test_wavs/1221-135766-0001.wav \
$repo/test_wavs/1221-135766-0002.wav \
--num-encoder-layers 18 \
--dim-feedforward 2048 \
--nhead 8 \
--encoder-dim 512 \
--decoder-dim 512 \
--joiner-dim 512
done
echo "GITHUB_EVENT_NAME: ${GITHUB_EVENT_NAME}"
echo "GITHUB_EVENT_LABEL_NAME: ${GITHUB_EVENT_LABEL_NAME}"
if [[ x"${GITHUB_EVENT_NAME}" == x"schedule" || x"${GITHUB_EVENT_LABEL_NAME}" == x"run-decode" ]]; then
mkdir -p pruned_transducer_stateless5/exp
ln -s $PWD/$repo/exp/pretrained-epoch-39-avg-7.pt pruned_transducer_stateless5/exp/epoch-999.pt
ln -s $PWD/$repo/data/lang_bpe_500 data/
ls -lh data
ls -lh pruned_transducer_stateless5/exp
log "Decoding test-clean and test-other"
# use a small value for decoding with CPU
max_duration=100
for method in greedy_search fast_beam_search modified_beam_search; do
log "Decoding with $method"
./pruned_transducer_stateless5/decode.py \
--decoding-method $method \
--use-averaged-model 0 \
--epoch 999 \
--avg 1 \
--max-duration $max_duration \
--exp-dir pruned_transducer_stateless5/exp \
--num-encoder-layers 18 \
--dim-feedforward 2048 \
--nhead 8 \
--encoder-dim 512 \
--decoder-dim 512 \
--joiner-dim 512
done
rm pruned_transducer_stateless5/exp/*.pt
fi

106
.github/scripts/run-librispeech-pruned-transducer-stateless7-2022-11-11.sh vendored
View File

@ -1,106 +0,0 @@
#!/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]}) $*"
}
cd egs/librispeech/ASR
repo_url=https://huggingface.co/csukuangfj/icefall-asr-librispeech-pruned-transducer-stateless7-2022-11-11
log "Downloading pre-trained model from $repo_url"
git lfs install
GIT_LFS_SKIP_SMUDGE=1 git clone $repo_url
repo=$(basename $repo_url)
log "Display test files"
tree $repo/
ls -lh $repo/test_wavs/*.wav
pushd $repo/exp
git lfs pull --include "data/lang_bpe_500/bpe.model"
git lfs pull --include "exp/cpu_jit.pt"
git lfs pull --include "exp/pretrained.pt"
ln -s pretrained.pt epoch-99.pt
ls -lh *.pt
popd
log "Export to torchscript model"
./pruned_transducer_stateless7/export.py \
--exp-dir $repo/exp \
--use-averaged-model false \
--tokens $repo/data/lang_bpe_500/tokens.txt \
--epoch 99 \
--avg 1 \
--jit 1
ls -lh $repo/exp/*.pt
log "Decode with models exported by torch.jit.script()"
./pruned_transducer_stateless7/jit_pretrained.py \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
--nn-model-filename $repo/exp/cpu_jit.pt \
$repo/test_wavs/1089-134686-0001.wav \
$repo/test_wavs/1221-135766-0001.wav \
$repo/test_wavs/1221-135766-0002.wav
for sym in 1 2 3; do
log "Greedy search with --max-sym-per-frame $sym"
./pruned_transducer_stateless7/pretrained.py \
--method greedy_search \
--max-sym-per-frame $sym \
--checkpoint $repo/exp/pretrained.pt \
--tokens $repo/data/lang_bpe_500/tokens.txt \
$repo/test_wavs/1089-134686-0001.wav \
$repo/test_wavs/1221-135766-0001.wav \
$repo/test_wavs/1221-135766-0002.wav
done
for method in modified_beam_search beam_search fast_beam_search; do
log "$method"
./pruned_transducer_stateless7/pretrained.py \
--method $method \
--beam-size 4 \
--checkpoint $repo/exp/pretrained.pt \
--tokens $repo/data/lang_bpe_500/tokens.txt \
$repo/test_wavs/1089-134686-0001.wav \
$repo/test_wavs/1221-135766-0001.wav \
$repo/test_wavs/1221-135766-0002.wav
done
echo "GITHUB_EVENT_NAME: ${GITHUB_EVENT_NAME}"
echo "GITHUB_EVENT_LABEL_NAME: ${GITHUB_EVENT_LABEL_NAME}"
if [[ x"${GITHUB_EVENT_NAME}" == x"schedule" || x"${GITHUB_EVENT_LABEL_NAME}" == x"run-decode" ]]; then
mkdir -p pruned_transducer_stateless7/exp
ln -s $PWD/$repo/exp/pretrained.pt pruned_transducer_stateless7/exp/epoch-999.pt
ln -s $PWD/$repo/data/lang_bpe_500 data/
ls -lh data
ls -lh pruned_transducer_stateless7/exp
log "Decoding test-clean and test-other"
# use a small value for decoding with CPU
max_duration=100
for method in greedy_search fast_beam_search modified_beam_search; do
log "Decoding with $method"
./pruned_transducer_stateless7/decode.py \
--decoding-method $method \
--epoch 999 \
--avg 1 \
--use-averaged-model 0 \
--max-duration $max_duration \
--exp-dir pruned_transducer_stateless7/exp
done
rm pruned_transducer_stateless7/exp/*.pt
fi

150
.github/scripts/run-librispeech-pruned-transducer-stateless7-ctc-2022-12-01.sh vendored
View File

@ -1,150 +0,0 @@
#!/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]}) $*"
}
cd egs/librispeech/ASR
repo_url=https://huggingface.co/Zengwei/icefall-asr-librispeech-pruned-transducer-stateless7-ctc-2022-12-01
log "Downloading pre-trained model from $repo_url"
GIT_LFS_SKIP_SMUDGE=1 git clone $repo_url
repo=$(basename $repo_url)
log "Display test files"
tree $repo/
ls -lh $repo/test_wavs/*.wav
pushd $repo/exp
git lfs pull --include "data/lang_bpe_500/HLG.pt"
git lfs pull --include "data/lang_bpe_500/L.pt"
git lfs pull --include "data/lang_bpe_500/LG.pt"
git lfs pull --include "data/lang_bpe_500/Linv.pt"
git lfs pull --include "data/lang_bpe_500/bpe.model"
git lfs pull --include "data/lm/G_4_gram.pt"
git lfs pull --include "exp/cpu_jit.pt"
git lfs pull --include "exp/pretrained.pt"
ln -s pretrained.pt epoch-99.pt
ls -lh *.pt
popd
log "Export to torchscript model"
./pruned_transducer_stateless7_ctc/export.py \
--exp-dir $repo/exp \
--use-averaged-model false \
--tokens $repo/data/lang_bpe_500/tokens.txt \
--epoch 99 \
--avg 1 \
--jit 1
ls -lh $repo/exp/*.pt
log "Decode with models exported by torch.jit.script()"
./pruned_transducer_stateless7_ctc/jit_pretrained.py \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
--nn-model-filename $repo/exp/cpu_jit.pt \
$repo/test_wavs/1089-134686-0001.wav \
$repo/test_wavs/1221-135766-0001.wav \
$repo/test_wavs/1221-135766-0002.wav
for m in ctc-decoding 1best; do
./pruned_transducer_stateless7_ctc/jit_pretrained_ctc.py \
--model-filename $repo/exp/cpu_jit.pt \
--words-file $repo/data/lang_bpe_500/words.txt \
--HLG $repo/data/lang_bpe_500/HLG.pt \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
--G $repo/data/lm/G_4_gram.pt \
--method $m \
--sample-rate 16000 \
$repo/test_wavs/1089-134686-0001.wav \
$repo/test_wavs/1221-135766-0001.wav \
$repo/test_wavs/1221-135766-0002.wav
done
for sym in 1 2 3; do
log "Greedy search with --max-sym-per-frame $sym"
./pruned_transducer_stateless7_ctc/pretrained.py \
--method greedy_search \
--max-sym-per-frame $sym \
--checkpoint $repo/exp/pretrained.pt \
--tokens $repo/data/lang_bpe_500/tokens.txt \
$repo/test_wavs/1089-134686-0001.wav \
$repo/test_wavs/1221-135766-0001.wav \
$repo/test_wavs/1221-135766-0002.wav
done
for method in modified_beam_search beam_search fast_beam_search; do
log "$method"
./pruned_transducer_stateless7_ctc/pretrained.py \
--method $method \
--beam-size 4 \
--checkpoint $repo/exp/pretrained.pt \
--tokens $repo/data/lang_bpe_500/tokens.txt \
$repo/test_wavs/1089-134686-0001.wav \
$repo/test_wavs/1221-135766-0001.wav \
$repo/test_wavs/1221-135766-0002.wav
done
for m in ctc-decoding 1best; do
./pruned_transducer_stateless7_ctc/pretrained_ctc.py \
--checkpoint $repo/exp/pretrained.pt \
--words-file $repo/data/lang_bpe_500/words.txt \
--HLG $repo/data/lang_bpe_500/HLG.pt \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
--G $repo/data/lm/G_4_gram.pt \
--method $m \
--sample-rate 16000 \
$repo/test_wavs/1089-134686-0001.wav \
$repo/test_wavs/1221-135766-0001.wav \
$repo/test_wavs/1221-135766-0002.wav
done
echo "GITHUB_EVENT_NAME: ${GITHUB_EVENT_NAME}"
echo "GITHUB_EVENT_LABEL_NAME: ${GITHUB_EVENT_LABEL_NAME}"
if [[ x"${GITHUB_EVENT_NAME}" == x"schedule" || x"${GITHUB_EVENT_LABEL_NAME}" == x"run-decode" ]]; then
mkdir -p pruned_transducer_stateless7_ctc/exp
ln -s $PWD/$repo/exp/pretrained.pt pruned_transducer_stateless7_ctc/exp/epoch-999.pt
ln -s $PWD/$repo/data/lang_bpe_500 data/
ls -lh data
ls -lh pruned_transducer_stateless7_ctc/exp
log "Decoding test-clean and test-other"
# use a small value for decoding with CPU
max_duration=100
for method in greedy_search fast_beam_search modified_beam_search; do
log "Decoding with $method"
./pruned_transducer_stateless7_ctc/decode.py \
--decoding-method $method \
--epoch 999 \
--avg 1 \
--use-averaged-model 0 \
--max-duration $max_duration \
--exp-dir pruned_transducer_stateless7_ctc/exp
done
for m in ctc-decoding 1best; do
./pruned_transducer_stateless7_ctc/ctc_decode.py \
--epoch 999 \
--avg 1 \
--exp-dir ./pruned_transducer_stateless7_ctc/exp \
--max-duration $max_duration \
--use-averaged-model 0 \
--decoding-method $m \
--hlg-scale 0.6 \
--lm-dir data/lm
done
rm pruned_transducer_stateless7_ctc/exp/*.pt
fi

147
.github/scripts/run-librispeech-pruned-transducer-stateless7-ctc-bs-2023-01-29.sh vendored
View File

@ -1,147 +0,0 @@
#!/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]}) $*"
}
cd egs/librispeech/ASR
repo_url=https://huggingface.co/yfyeung/icefall-asr-librispeech-pruned_transducer_stateless7_ctc_bs-2023-01-29
log "Downloading pre-trained model from $repo_url"
GIT_LFS_SKIP_SMUDGE=1 git clone $repo_url
repo=$(basename $repo_url)
log "Display test files"
tree $repo/
ls -lh $repo/test_wavs/*.wav
pushd $repo/exp
git lfs pull --include "data/lang_bpe_500/L.pt"
git lfs pull --include "data/lang_bpe_500/LG.pt"
git lfs pull --include "data/lang_bpe_500/HLG.pt"
git lfs pull --include "data/lang_bpe_500/Linv.pt"
git lfs pull --include "data/lang_bpe_500/bpe.model"
git lfs pull --include "exp/cpu_jit.pt"
git lfs pull --include "exp/pretrained.pt"
ln -s pretrained.pt epoch-99.pt
ls -lh *.pt
popd
log "Export to torchscript model"
./pruned_transducer_stateless7_ctc_bs/export.py \
--exp-dir $repo/exp \
--use-averaged-model false \
--tokens $repo/data/lang_bpe_500/tokens.txt \
--epoch 99 \
--avg 1 \
--jit 1
ls -lh $repo/exp/*.pt
log "Decode with models exported by torch.jit.script()"
./pruned_transducer_stateless7_ctc_bs/jit_pretrained.py \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
--nn-model-filename $repo/exp/cpu_jit.pt \
$repo/test_wavs/1089-134686-0001.wav \
$repo/test_wavs/1221-135766-0001.wav \
$repo/test_wavs/1221-135766-0002.wav
for m in ctc-decoding 1best; do
./pruned_transducer_stateless7_ctc_bs/jit_pretrained_ctc.py \
--model-filename $repo/exp/cpu_jit.pt \
--words-file $repo/data/lang_bpe_500/words.txt \
--HLG $repo/data/lang_bpe_500/HLG.pt \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
--method $m \
--sample-rate 16000 \
$repo/test_wavs/1089-134686-0001.wav \
$repo/test_wavs/1221-135766-0001.wav \
$repo/test_wavs/1221-135766-0002.wav
done
for sym in 1 2 3; do
log "Greedy search with --max-sym-per-frame $sym"
./pruned_transducer_stateless7_ctc_bs/pretrained.py \
--method greedy_search \
--max-sym-per-frame $sym \
--checkpoint $repo/exp/pretrained.pt \
--tokens $repo/data/lang_bpe_500/tokens.txt \
$repo/test_wavs/1089-134686-0001.wav \
$repo/test_wavs/1221-135766-0001.wav \
$repo/test_wavs/1221-135766-0002.wav
done
for method in modified_beam_search beam_search fast_beam_search; do
log "$method"
./pruned_transducer_stateless7_ctc_bs/pretrained.py \
--method $method \
--beam-size 4 \
--checkpoint $repo/exp/pretrained.pt \
--tokens $repo/data/lang_bpe_500/tokens.txt \
$repo/test_wavs/1089-134686-0001.wav \
$repo/test_wavs/1221-135766-0001.wav \
$repo/test_wavs/1221-135766-0002.wav
done
for m in ctc-decoding 1best; do
./pruned_transducer_stateless7_ctc_bs/pretrained_ctc.py \
--checkpoint $repo/exp/pretrained.pt \
--words-file $repo/data/lang_bpe_500/words.txt \
--HLG $repo/data/lang_bpe_500/HLG.pt \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
--method $m \
--sample-rate 16000 \
$repo/test_wavs/1089-134686-0001.wav \
$repo/test_wavs/1221-135766-0001.wav \
$repo/test_wavs/1221-135766-0002.wav
done
echo "GITHUB_EVENT_NAME: ${GITHUB_EVENT_NAME}"
echo "GITHUB_EVENT_LABEL_NAME: ${GITHUB_EVENT_LABEL_NAME}"
if [[ x"${GITHUB_EVENT_NAME}" == x"schedule" || x"${GITHUB_EVENT_LABEL_NAME}" == x"run-decode" ]]; then
mkdir -p pruned_transducer_stateless7_ctc_bs/exp
ln -s $PWD/$repo/exp/pretrained.pt pruned_transducer_stateless7_ctc_bs/exp/epoch-999.pt
ln -s $PWD/$repo/data/lang_bpe_500 data/
ls -lh data
ls -lh pruned_transducer_stateless7_ctc_bs/exp
log "Decoding test-clean and test-other"
# use a small value for decoding with CPU
max_duration=100
for method in greedy_search fast_beam_search modified_beam_search; do
log "Decoding with $method"
./pruned_transducer_stateless7_ctc_bs/decode.py \
--decoding-method $method \
--epoch 999 \
--avg 1 \
--use-averaged-model 0 \
--max-duration $max_duration \
--exp-dir pruned_transducer_stateless7_ctc_bs/exp
done
for m in ctc-decoding 1best; do
./pruned_transducer_stateless7_ctc_bs/ctc_decode.py \
--epoch 999 \
--avg 1 \
--exp-dir ./pruned_transducer_stateless7_ctc_bs/exp \
--max-duration $max_duration \
--use-averaged-model 0 \
--decoding-method $m \
--hlg-scale 0.6
done
rm pruned_transducer_stateless7_ctc_bs/exp/*.pt
fi

148
.github/scripts/run-librispeech-pruned-transducer-stateless7-streaming-2022-12-29.sh vendored
View File

@ -1,148 +0,0 @@
#!/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]}) $*"
}
cd egs/librispeech/ASR
repo_url=https://huggingface.co/Zengwei/icefall-asr-librispeech-pruned-transducer-stateless7-streaming-2022-12-29
log "Downloading pre-trained model from $repo_url"
git lfs install
GIT_LFS_SKIP_SMUDGE=1 git clone $repo_url
repo=$(basename $repo_url)
log "Display test files"
tree $repo/
ls -lh $repo/test_wavs/*.wav
pushd $repo
git lfs pull --include "data/lang_bpe_500/bpe.model"
git lfs pull --include "exp/cpu_jit.pt"
git lfs pull --include "exp/pretrained.pt"
git lfs pull --include "exp/encoder_jit_trace.pt"
git lfs pull --include "exp/decoder_jit_trace.pt"
git lfs pull --include "exp/joiner_jit_trace.pt"
cd exp
ln -s pretrained.pt epoch-99.pt
ls -lh *.pt
popd
log "Export to torchscript model"
./pruned_transducer_stateless7_streaming/export.py \
--exp-dir $repo/exp \
--use-averaged-model false \
--tokens $repo/data/lang_bpe_500/tokens.txt \
--decode-chunk-len 32 \
--epoch 99 \
--avg 1 \
--jit 1
ls -lh $repo/exp/*.pt
log "Decode with models exported by torch.jit.script()"
./pruned_transducer_stateless7_streaming/jit_pretrained.py \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
--nn-model-filename $repo/exp/cpu_jit.pt \
--decode-chunk-len 32 \
$repo/test_wavs/1089-134686-0001.wav \
$repo/test_wavs/1221-135766-0001.wav \
$repo/test_wavs/1221-135766-0002.wav
log "Export to torchscript model by torch.jit.trace()"
./pruned_transducer_stateless7_streaming/jit_trace_export.py \
--exp-dir $repo/exp \
--use-averaged-model false \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
--decode-chunk-len 32 \
--epoch 99 \
--avg 1
log "Decode with models exported by torch.jit.trace()"
./pruned_transducer_stateless7_streaming/jit_trace_pretrained.py \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
--encoder-model-filename $repo/exp/encoder_jit_trace.pt \
--decoder-model-filename $repo/exp/decoder_jit_trace.pt \
--joiner-model-filename $repo/exp/joiner_jit_trace.pt \
--decode-chunk-len 32 \
$repo/test_wavs/1089-134686-0001.wav
for sym in 1 2 3; do
log "Greedy search with --max-sym-per-frame $sym"
./pruned_transducer_stateless7_streaming/pretrained.py \
--method greedy_search \
--max-sym-per-frame $sym \
--checkpoint $repo/exp/pretrained.pt \
--tokens $repo/data/lang_bpe_500/tokens.txt \
--decode-chunk-len 32 \
$repo/test_wavs/1089-134686-0001.wav \
$repo/test_wavs/1221-135766-0001.wav \
$repo/test_wavs/1221-135766-0002.wav
done
for method in modified_beam_search beam_search fast_beam_search; do
log "$method"
./pruned_transducer_stateless7_streaming/pretrained.py \
--method $method \
--beam-size 4 \
--checkpoint $repo/exp/pretrained.pt \
--tokens $repo/data/lang_bpe_500/tokens.txt \
--decode-chunk-len 32 \
$repo/test_wavs/1089-134686-0001.wav \
$repo/test_wavs/1221-135766-0001.wav \
$repo/test_wavs/1221-135766-0002.wav
done
echo "GITHUB_EVENT_NAME: ${GITHUB_EVENT_NAME}"
echo "GITHUB_EVENT_LABEL_NAME: ${GITHUB_EVENT_LABEL_NAME}"
if [[ x"${GITHUB_EVENT_NAME}" == x"schedule" || x"${GITHUB_EVENT_LABEL_NAME}" == x"run-decode" ]]; then
mkdir -p pruned_transducer_stateless7_streaming/exp
ln -s $PWD/$repo/exp/pretrained.pt pruned_transducer_stateless7_streaming/exp/epoch-999.pt
ln -s $PWD/$repo/data/lang_bpe_500 data/
ls -lh data
ls -lh pruned_transducer_stateless7_streaming/exp
log "Decoding test-clean and test-other"
# use a small value for decoding with CPU
max_duration=100
num_decode_stream=200
for method in greedy_search fast_beam_search modified_beam_search; do
log "decoding with $method"
./pruned_transducer_stateless7_streaming/decode.py \
--decoding-method $method \
--epoch 999 \
--avg 1 \
--use-averaged-model 0 \
--max-duration $max_duration \
--decode-chunk-len 32 \
--exp-dir pruned_transducer_stateless7_streaming/exp
done
for method in greedy_search fast_beam_search modified_beam_search; do
log "Decoding with $method"
./pruned_transducer_stateless7_streaming/streaming_decode.py \
--decoding-method $method \
--epoch 999 \
--avg 1 \
--use-averaged-model 0 \
--decode-chunk-len 32 \
--num-decode-streams $num_decode_stream
--exp-dir pruned_transducer_stateless7_streaming/exp
done
rm pruned_transducer_stateless7_streaming/exp/*.pt
fi

115
.github/scripts/run-librispeech-pruned-transducer-stateless8-2022-11-14.sh vendored
View File

@ -1,115 +0,0 @@
#!/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]}) $*"
}
cd egs/librispeech/ASR
repo_url=https://huggingface.co/csukuangfj/icefall-asr-librispeech-pruned-transducer-stateless8-2022-11-14
log "Downloading pre-trained model from $repo_url"
git lfs install
GIT_LFS_SKIP_SMUDGE=1 git clone $repo_url
repo=$(basename $repo_url)
log "Display test files"
tree $repo/
ls -lh $repo/test_wavs/*.wav
pushd $repo/exp
git lfs pull --include "data/lang_bpe_500/bpe.model"
git lfs pull --include "exp/cpu_jit.pt"
git lfs pull --include "exp/pretrained.pt"
ln -s pretrained.pt epoch-99.pt
ls -lh *.pt
popd
log "Decode with models exported by torch.jit.script()"
./pruned_transducer_stateless8/jit_pretrained.py \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
--nn-model-filename $repo/exp/cpu_jit.pt \
$repo/test_wavs/1089-134686-0001.wav \
$repo/test_wavs/1221-135766-0001.wav \
$repo/test_wavs/1221-135766-0002.wav
log "Export to torchscript model"
./pruned_transducer_stateless8/export.py \
--exp-dir $repo/exp \
--tokens $repo/data/lang_bpe_500/tokens.txt \
--use-averaged-model false \
--epoch 99 \
--avg 1 \
--jit 1
ls -lh $repo/exp/*.pt
log "Decode with models exported by torch.jit.script()"
./pruned_transducer_stateless8/jit_pretrained.py \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
--nn-model-filename $repo/exp/cpu_jit.pt \
$repo/test_wavs/1089-134686-0001.wav \
$repo/test_wavs/1221-135766-0001.wav \
$repo/test_wavs/1221-135766-0002.wav
for sym in 1 2 3; do
log "Greedy search with --max-sym-per-frame $sym"
./pruned_transducer_stateless8/pretrained.py \
--method greedy_search \
--max-sym-per-frame $sym \
--checkpoint $repo/exp/pretrained.pt \
--tokens $repo/data/lang_bpe_500/tokens.txt \
$repo/test_wavs/1089-134686-0001.wav \
$repo/test_wavs/1221-135766-0001.wav \
$repo/test_wavs/1221-135766-0002.wav
done
for method in modified_beam_search beam_search fast_beam_search; do
log "$method"
./pruned_transducer_stateless8/pretrained.py \
--method $method \
--beam-size 4 \
--checkpoint $repo/exp/pretrained.pt \
--tokens $repo/data/lang_bpe_500/tokens.txt \
$repo/test_wavs/1089-134686-0001.wav \
$repo/test_wavs/1221-135766-0001.wav \
$repo/test_wavs/1221-135766-0002.wav
done
echo "GITHUB_EVENT_NAME: ${GITHUB_EVENT_NAME}"
echo "GITHUB_EVENT_LABEL_NAME: ${GITHUB_EVENT_LABEL_NAME}"
if [[ x"${GITHUB_EVENT_NAME}" == x"schedule" || x"${GITHUB_EVENT_LABEL_NAME}" == x"run-decode" ]]; then
mkdir -p pruned_transducer_stateless8/exp
ln -s $PWD/$repo/exp/pretrained.pt pruned_transducer_stateless8/exp/epoch-999.pt
ln -s $PWD/$repo/data/lang_bpe_500 data/
ls -lh data
ls -lh pruned_transducer_stateless8/exp
log "Decoding test-clean and test-other"
# use a small value for decoding with CPU
max_duration=100
for method in greedy_search fast_beam_search modified_beam_search; do
log "Decoding with $method"
./pruned_transducer_stateless8/decode.py \
--decoding-method $method \
--epoch 999 \
--avg 1 \
--use-averaged-model 0 \
--max-duration $max_duration \
--exp-dir pruned_transducer_stateless8/exp
done
rm pruned_transducer_stateless8/exp/*.pt
fi

101
.github/scripts/run-librispeech-streaming-pruned-transducer-stateless2-2022-06-26.sh vendored
View File

@ -1,101 +0,0 @@
#!/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]}) $*"
}
cd egs/librispeech/ASR
repo_url=https://huggingface.co/pkufool/icefall_librispeech_streaming_pruned_transducer_stateless2_20220625
log "Downloading pre-trained model from $repo_url"
git lfs install
git clone $repo_url
repo=$(basename $repo_url)
log "Display test files"
tree $repo/
ls -lh $repo/test_wavs/*.wav
pushd $repo/exp
ln -s pretrained-epoch-24-avg-10.pt pretrained.pt
popd
for sym in 1 2 3; do
log "Greedy search with --max-sym-per-frame $sym"
./pruned_transducer_stateless2/pretrained.py \
--method greedy_search \
--max-sym-per-frame $sym \
--checkpoint $repo/exp/pretrained.pt \
--tokens $repo/data/lang_bpe_500/tokens.txt \
--simulate-streaming 1 \
--causal-convolution 1 \
$repo/test_wavs/1089-134686-0001.wav \
$repo/test_wavs/1221-135766-0001.wav \
$repo/test_wavs/1221-135766-0002.wav
done
for method in modified_beam_search beam_search fast_beam_search; do
log "$method"
./pruned_transducer_stateless2/pretrained.py \
--method $method \
--beam-size 4 \
--checkpoint $repo/exp/pretrained.pt \
--tokens $repo/data/lang_bpe_500/tokens.txt \
--simulate-streaming 1 \
--causal-convolution 1 \
$repo/test_wavs/1089-134686-0001.wav \
$repo/test_wavs/1221-135766-0001.wav \
$repo/test_wavs/1221-135766-0002.wav
done
echo "GITHUB_EVENT_NAME: ${GITHUB_EVENT_NAME}"
echo "GITHUB_EVENT_LABEL_NAME: ${GITHUB_EVENT_LABEL_NAME}"
if [[ x"${GITHUB_EVENT_NAME}" == x"schedule" || x"${GITHUB_EVENT_LABEL_NAME}" == x"run-decode" ]]; then
mkdir -p pruned_transducer_stateless2/exp
ln -s $PWD/$repo/exp/pretrained-epoch-24-avg-10.pt pruned_transducer_stateless2/exp/epoch-999.pt
ln -s $PWD/$repo/data/lang_bpe_500 data/
ls -lh data
ls -lh pruned_transducer_stateless2/exp
log "Decoding test-clean and test-other"
# use a small value for decoding with CPU
max_duration=100
for method in greedy_search fast_beam_search modified_beam_search; do
log "Simulate streaming decoding with $method"
./pruned_transducer_stateless2/decode.py \
--decoding-method $method \
--epoch 999 \
--avg 1 \
--max-duration $max_duration \
--exp-dir pruned_transducer_stateless2/exp \
--simulate-streaming 1 \
--causal-convolution 1
done
for method in greedy_search fast_beam_search modified_beam_search; do
log "Real streaming decoding with $method"
./pruned_transducer_stateless2/streaming_decode.py \
--decoding-method $method \
--epoch 999 \
--avg 1 \
--num-decode-streams 100 \
--exp-dir pruned_transducer_stateless2/exp \
--left-context 32 \
--decode-chunk-size 8 \
--right-context 0
done
rm pruned_transducer_stateless2/exp/*.pt
fi

116
.github/scripts/run-librispeech-streaming-zipformer-2023-05-18.sh vendored
View File

@ -1,116 +0,0 @@
#!/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]}) $*"
}
cd egs/librispeech/ASR
repo_url=https://huggingface.co/Zengwei/icefall-asr-librispeech-streaming-zipformer-2023-05-17
log "Downloading pre-trained model from $repo_url"
git lfs install
GIT_LFS_SKIP_SMUDGE=1 git clone $repo_url
repo=$(basename $repo_url)
log "Display test files"
tree $repo/
ls -lh $repo/test_wavs/*.wav
pushd $repo/exp
git lfs pull --include "data/lang_bpe_500/bpe.model"
git lfs pull --include "data/lang_bpe_500/tokens.txt"
git lfs pull --include "exp/jit_script_chunk_16_left_128.pt"
git lfs pull --include "exp/pretrained.pt"
ln -s pretrained.pt epoch-99.pt
ls -lh *.pt
popd
log "Export to torchscript model"
./zipformer/export.py \
--exp-dir $repo/exp \
--use-averaged-model false \
--tokens $repo/data/lang_bpe_500/tokens.txt \
--causal 1 \
--chunk-size 16 \
--left-context-frames 128 \
--epoch 99 \
--avg 1 \
--jit 1
ls -lh $repo/exp/*.pt
log "Decode with models exported by torch.jit.script()"
./zipformer/jit_pretrained_streaming.py \
--tokens $repo/data/lang_bpe_500/tokens.txt \
--nn-model-filename $repo/exp/jit_script_chunk_16_left_128.pt \
$repo/test_wavs/1089-134686-0001.wav
for method in greedy_search modified_beam_search fast_beam_search; do
log "$method"
./zipformer/pretrained.py \
--causal 1 \
--chunk-size 16 \
--left-context-frames 128 \
--method $method \
--beam-size 4 \
--checkpoint $repo/exp/pretrained.pt \
--tokens $repo/data/lang_bpe_500/tokens.txt \
$repo/test_wavs/1089-134686-0001.wav \
$repo/test_wavs/1221-135766-0001.wav \
$repo/test_wavs/1221-135766-0002.wav
done
echo "GITHUB_EVENT_NAME: ${GITHUB_EVENT_NAME}"
echo "GITHUB_EVENT_LABEL_NAME: ${GITHUB_EVENT_LABEL_NAME}"
if [[ x"${GITHUB_EVENT_NAME}" == x"schedule" || x"${GITHUB_EVENT_LABEL_NAME}" == x"run-decode" ]]; then
mkdir -p zipformer/exp
ln -s $PWD/$repo/exp/pretrained.pt zipformer/exp/epoch-999.pt
ln -s $PWD/$repo/data/lang_bpe_500 data/
ls -lh data
ls -lh zipformer/exp
log "Decoding test-clean and test-other"
# use a small value for decoding with CPU
max_duration=100
for method in greedy_search fast_beam_search modified_beam_search; do
log "Simulated streaming decoding with $method"
./zipformer/decode.py \
--causal 1 \
--chunk-size 16 \
--left-context-frames 128 \
--decoding-method $method \
--epoch 999 \
--avg 1 \
--use-averaged-model 0 \
--max-duration $max_duration \
--exp-dir zipformer/exp
done
for method in greedy_search fast_beam_search modified_beam_search; do
log "Chunk-wise streaming decoding with $method"
./zipformer/streaming_decode.py \
--causal 1 \
--chunk-size 16 \
--left-context-frames 128 \
--decoding-method $method \
--epoch 999 \
--avg 1 \
--use-averaged-model 0 \
--max-duration $max_duration \
--exp-dir zipformer/exp
done
rm zipformer/exp/*.pt
fi

77
.github/scripts/run-librispeech-transducer-stateless2-2022-04-19.sh vendored
View File

@ -1,77 +0,0 @@
#!/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]}) $*"
}
cd egs/librispeech/ASR
repo_url=https://huggingface.co/csukuangfj/icefall-asr-librispeech-transducer-stateless2-torchaudio-2022-04-19
log "Downloading pre-trained model from $repo_url"
git lfs install
git clone $repo_url
repo=$(basename $repo_url)
log "Display test files"
tree $repo/
ls -lh $repo/test_wavs/*.wav
for sym in 1 2 3; do
log "Greedy search with --max-sym-per-frame $sym"
./transducer_stateless2/pretrained.py \
--method greedy_search \
--max-sym-per-frame $sym \
--checkpoint $repo/exp/pretrained.pt \
--tokens $repo/data/lang_bpe_500/tokens.txt \
$repo/test_wavs/1089-134686-0001.wav \
$repo/test_wavs/1221-135766-0001.wav \
$repo/test_wavs/1221-135766-0002.wav
done
for method in fast_beam_search modified_beam_search beam_search; do
log "$method"
./transducer_stateless2/pretrained.py \
--method $method \
--beam-size 4 \
--checkpoint $repo/exp/pretrained.pt \
--tokens $repo/data/lang_bpe_500/tokens.txt \
$repo/test_wavs/1089-134686-0001.wav \
$repo/test_wavs/1221-135766-0001.wav \
$repo/test_wavs/1221-135766-0002.wav
done
echo "GITHUB_EVENT_NAME: ${GITHUB_EVENT_NAME}"
echo "GITHUB_EVENT_LABEL_NAME: ${GITHUB_EVENT_LABEL_NAME}"
if [[ x"${GITHUB_EVENT_NAME}" == x"schedule" || x"${GITHUB_EVENT_LABEL_NAME}" == x"run-decode" ]]; then
mkdir -p transducer_stateless2/exp
ln -s $PWD/$repo/exp/pretrained.pt transducer_stateless2/exp/epoch-999.pt
ln -s $PWD/$repo/data/lang_bpe_500 data/
ls -lh data
ls -lh transducer_stateless2/exp
log "Decoding test-clean and test-other"
# use a small value for decoding with CPU
max_duration=100
for method in greedy_search fast_beam_search modified_beam_search; do
log "Decoding with $method"
./transducer_stateless2/decode.py \
--decoding-method $method \
--epoch 999 \
--avg 1 \
--max-duration $max_duration \
--exp-dir transducer_stateless2/exp
done
rm transducer_stateless2/exp/*.pt
fi

94
.github/scripts/run-librispeech-zipformer-2023-05-18.sh vendored
View File

@ -1,94 +0,0 @@
#!/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]}) $*"
}
cd egs/librispeech/ASR
repo_url=https://huggingface.co/Zengwei/icefall-asr-librispeech-zipformer-2023-05-15
log "Downloading pre-trained model from $repo_url"
git lfs install
GIT_LFS_SKIP_SMUDGE=1 git clone $repo_url
repo=$(basename $repo_url)
log "Display test files"
tree $repo/
ls -lh $repo/test_wavs/*.wav
pushd $repo/exp
git lfs pull --include "data/lang_bpe_500/bpe.model"
git lfs pull --include "data/lang_bpe_500/tokens.txt"
git lfs pull --include "exp/jit_script.pt"
git lfs pull --include "exp/pretrained.pt"
ln -s pretrained.pt epoch-99.pt
ls -lh *.pt
popd
log "Export to torchscript model"
./zipformer/export.py \
--exp-dir $repo/exp \
--use-averaged-model false \
--tokens $repo/data/lang_bpe_500/tokens.txt \
--epoch 99 \
--avg 1 \
--jit 1
ls -lh $repo/exp/*.pt
log "Decode with models exported by torch.jit.script()"
./zipformer/jit_pretrained.py \
--tokens $repo/data/lang_bpe_500/tokens.txt \
--nn-model-filename $repo/exp/jit_script.pt \
$repo/test_wavs/1089-134686-0001.wav \
$repo/test_wavs/1221-135766-0001.wav \
$repo/test_wavs/1221-135766-0002.wav
for method in greedy_search modified_beam_search fast_beam_search; do
log "$method"
./zipformer/pretrained.py \
--method $method \
--beam-size 4 \
--checkpoint $repo/exp/pretrained.pt \
--tokens $repo/data/lang_bpe_500/tokens.txt \
$repo/test_wavs/1089-134686-0001.wav \
$repo/test_wavs/1221-135766-0001.wav \
$repo/test_wavs/1221-135766-0002.wav
done
echo "GITHUB_EVENT_NAME: ${GITHUB_EVENT_NAME}"
echo "GITHUB_EVENT_LABEL_NAME: ${GITHUB_EVENT_LABEL_NAME}"
if [[ x"${GITHUB_EVENT_NAME}" == x"schedule" || x"${GITHUB_EVENT_LABEL_NAME}" == x"run-decode" ]]; then
mkdir -p zipformer/exp
ln -s $PWD/$repo/exp/pretrained.pt zipformer/exp/epoch-999.pt
ln -s $PWD/$repo/data/lang_bpe_500 data/
ls -lh data
ls -lh zipformer/exp
log "Decoding test-clean and test-other"
# use a small value for decoding with CPU
max_duration=100
for method in greedy_search fast_beam_search modified_beam_search; do
log "Decoding with $method"
./zipformer/decode.py \
--decoding-method $method \
--epoch 999 \
--avg 1 \
--use-averaged-model 0 \
--max-duration $max_duration \
--exp-dir zipformer/exp
done
rm zipformer/exp/*.pt
fi

117
.github/scripts/run-librispeech-zipformer-ctc-2023-06-14.sh vendored
View File

@ -1,117 +0,0 @@
#!/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]}) $*"
}
cd egs/librispeech/ASR
repo_url=https://huggingface.co/Zengwei/icefall-asr-librispeech-zipformer-transducer-ctc-2023-06-13
log "Downloading pre-trained model from $repo_url"
git lfs install
GIT_LFS_SKIP_SMUDGE=1 git clone $repo_url
repo=$(basename $repo_url)
log "Display test files"
tree $repo/
ls -lh $repo/test_wavs/*.wav
pushd $repo/exp
git lfs pull --include "data/lang_bpe_500/bpe.model"
git lfs pull --include "data/lang_bpe_500/tokens.txt"
git lfs pull --include "data/lang_bpe_500/HLG.pt"
git lfs pull --include "data/lang_bpe_500/L.pt"
git lfs pull --include "data/lang_bpe_500/LG.pt"
git lfs pull --include "data/lang_bpe_500/Linv.pt"
git lfs pull --include "data/lm/G_4_gram.pt"
git lfs pull --include "exp/jit_script.pt"
git lfs pull --include "exp/pretrained.pt"
ln -s pretrained.pt epoch-99.pt
ls -lh *.pt
popd
log "Export to torchscript model"
./zipformer/export.py \
--exp-dir $repo/exp \
--use-transducer 1 \
--use-ctc 1 \
--use-averaged-model false \
--tokens $repo/data/lang_bpe_500/tokens.txt \
--epoch 99 \
--avg 1 \
--jit 1
ls -lh $repo/exp/*.pt
log "Decode with models exported by torch.jit.script()"
for method in ctc-decoding 1best; do
./zipformer/jit_pretrained_ctc.py \
--tokens $repo/data/lang_bpe_500/tokens.txt \
--model-filename $repo/exp/jit_script.pt \
--HLG $repo/data/lang_bpe_500/HLG.pt \
--words-file $repo/data/lang_bpe_500/words.txt \
--G $repo/data/lm/G_4_gram.pt \
--method $method \
--sample-rate 16000 \
$repo/test_wavs/1089-134686-0001.wav \
$repo/test_wavs/1221-135766-0001.wav \
$repo/test_wavs/1221-135766-0002.wav
done
for method in ctc-decoding 1best; do
log "$method"
./zipformer/pretrained_ctc.py \
--use-transducer 1 \
--use-ctc 1 \
--method $method \
--checkpoint $repo/exp/pretrained.pt \
--tokens $repo/data/lang_bpe_500/tokens.txt \
--HLG $repo/data/lang_bpe_500/HLG.pt \
--G $repo/data/lm/G_4_gram.pt \
--words-file $repo/data/lang_bpe_500/words.txt \
--sample-rate 16000 \
$repo/test_wavs/1089-134686-0001.wav \
$repo/test_wavs/1221-135766-0001.wav \
$repo/test_wavs/1221-135766-0002.wav
done
echo "GITHUB_EVENT_NAME: ${GITHUB_EVENT_NAME}"
echo "GITHUB_EVENT_LABEL_NAME: ${GITHUB_EVENT_LABEL_NAME}"
if [[ x"${GITHUB_EVENT_NAME}" == x"schedule" || x"${GITHUB_EVENT_LABEL_NAME}" == x"run-decode" ]]; then
mkdir -p zipformer/exp
ln -s $PWD/$repo/exp/pretrained.pt zipformer/exp/epoch-999.pt
ln -s $PWD/$repo/data/lang_bpe_500 data/
ls -lh data
ls -lh zipformer/exp
log "Decoding test-clean and test-other"
# use a small value for decoding with CPU
max_duration=100
for method in ctc-decoding 1best; do
log "Decoding with $method"
./zipformer/ctc_decode.py \
--use-transducer 1 \
--use-ctc 1 \
--decoding-method $method \
--nbest-scale 1.0 \
--hlg-scale 0.6 \
--epoch 999 \
--avg 1 \
--use-averaged-model 0 \
--max-duration $max_duration \
--exp-dir zipformer/exp
done
rm zipformer/exp/*.pt
fi

102
.github/scripts/run-librispeech-zipformer-mmi-2022-12-08.sh vendored
View File

@ -1,102 +0,0 @@
#!/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]}) $*"
}
cd egs/librispeech/ASR
repo_url=https://huggingface.co/Zengwei/icefall-asr-librispeech-zipformer-mmi-2022-12-08
log "Downloading pre-trained model from $repo_url"
GIT_LFS_SKIP_SMUDGE=1 git clone $repo_url
repo=$(basename $repo_url)
log "Display test files"
tree $repo/
ls -lh $repo/test_wavs/*.wav
pushd $repo/exp
git lfs pull --include "data/lang_bpe_500/3gram.pt"
git lfs pull --include "data/lang_bpe_500/4gram.pt"
git lfs pull --include "data/lang_bpe_500/L.pt"
git lfs pull --include "data/lang_bpe_500/LG.pt"
git lfs pull --include "data/lang_bpe_500/Linv.pt"
git lfs pull --include "data/lang_bpe_500/bpe.model"
git lfs pull --include "exp/cpu_jit.pt"
git lfs pull --include "exp/pretrained.pt"
ln -s pretrained.pt epoch-99.pt
ls -lh *.pt
popd
log "Export to torchscript model"
./zipformer_mmi/export.py \
--exp-dir $repo/exp \
--use-averaged-model false \
--tokens $repo/data/lang_bpe_500/tokens.txt \
--epoch 99 \
--avg 1 \
--jit 1
ls -lh $repo/exp/*.pt
log "Decode with models exported by torch.jit.script()"
./zipformer_mmi/jit_pretrained.py \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
--nn-model-filename $repo/exp/cpu_jit.pt \
--lang-dir $repo/data/lang_bpe_500 \
$repo/test_wavs/1089-134686-0001.wav \
$repo/test_wavs/1221-135766-0001.wav \
$repo/test_wavs/1221-135766-0002.wav
for method in 1best nbest nbest-rescoring-LG nbest-rescoring-3-gram nbest-rescoring-4-gram; do
log "$method"
./zipformer_mmi/pretrained.py \
--method $method \
--checkpoint $repo/exp/pretrained.pt \
--lang-dir $repo/data/lang_bpe_500 \
--tokens $repo/data/lang_bpe_500/tokens.txt \
$repo/test_wavs/1089-134686-0001.wav \
$repo/test_wavs/1221-135766-0001.wav \
$repo/test_wavs/1221-135766-0002.wav
done
echo "GITHUB_EVENT_NAME: ${GITHUB_EVENT_NAME}"
echo "GITHUB_EVENT_LABEL_NAME: ${GITHUB_EVENT_LABEL_NAME}"
if [[ x"${GITHUB_EVENT_NAME}" == x"schedule" || x"${GITHUB_EVENT_LABEL_NAME}" == x"run-decode" ]]; then
mkdir -p zipformer_mmi/exp
ln -s $PWD/$repo/exp/pretrained.pt zipformer_mmi/exp/epoch-999.pt
ln -s $PWD/$repo/data/lang_bpe_500 data/
ls -lh data
ls -lh zipformer_mmi/exp
log "Decoding test-clean and test-other"
# use a small value for decoding with CPU
max_duration=100
for method in 1best nbest nbest-rescoring-LG nbest-rescoring-3-gram nbest-rescoring-4-gram; do
log "Decoding with $method"
./zipformer_mmi/decode.py \
--decoding-method $method \
--epoch 999 \
--avg 1 \
--use-averaged-model 0 \
--nbest-scale 1.2 \
--hp-scale 1.0 \
--max-duration $max_duration \
--lang-dir $repo/data/lang_bpe_500 \
--exp-dir zipformer_mmi/exp
done
rm zipformer_mmi/exp/*.pt
fi

40
.github/scripts/run-multi-zh_hans-zipformer.sh → .github/scripts/run-multi-corpora-zipformer.sh vendored
View File

@ -51,6 +51,8 @@ for method in modified_beam_search fast_beam_search; do
$repo/test_wavs/DEV_T0000000002.wav
done
rm -rf $repo
log "==== Test icefall-asr-multi-zh-hans-zipformer-ctc-2023-10-24 ===="
repo_url=https://huggingface.co/zrjin/icefall-asr-multi-zh-hans-zipformer-ctc-2023-10-24/
@ -93,3 +95,41 @@ for method in modified_beam_search fast_beam_search; do
$repo/test_wavs/DEV_T0000000001.wav \
$repo/test_wavs/DEV_T0000000002.wav
done
rm -rf $repo
cd ../../../egs/multi_zh_en/ASR
log "==== Test icefall-asr-zipformer-multi-zh-en-2023-11-22 ===="
repo_url=https://huggingface.co/zrjin/icefall-asr-zipformer-multi-zh-en-2023-11-22/
log "Downloading pre-trained model from $repo_url"
git lfs install
git clone $repo_url
repo=$(basename $repo_url)
log "Display test files"
tree $repo/
ls -lh $repo/test_wavs/*.wav
./zipformer/pretrained.py \
--checkpoint $repo/exp/pretrained.pt \
--bpe-model $repo/data/lang_bbpe_2000/bbpe.model \
--method greedy_search \
$repo/test_wavs/_1634_210_2577_1_1525157964032_3712259_29.wav \
$repo/test_wavs/_1634_210_2577_1_1525157964032_3712259_55.wav \
$repo/test_wavs/_1634_210_2577_1_1525157964032_3712259_75.wav
for method in modified_beam_search fast_beam_search; do
log "$method"
./zipformer/pretrained.py \
--method $method \
--beam-size 4 \
--checkpoint $repo/exp/pretrained.pt \
--bpe-model $repo/data/lang_bbpe_2000/bbpe.model \
$repo/test_wavs/_1634_210_2577_1_1525157964032_3712259_29.wav \
$repo/test_wavs/_1634_210_2577_1_1525157964032_3712259_55.wav \
$repo/test_wavs/_1634_210_2577_1_1525157964032_3712259_75.wav
done
rm -rf $repo

240
.github/scripts/run-pre-trained-ctc.sh vendored
View File

@ -1,240 +0,0 @@
#!/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]}) $*"
}
pushd egs/librispeech/ASR
repo_url=https://huggingface.co/csukuangfj/sherpa-onnx-zipformer-ctc-en-2023-10-02
log "Downloading pre-trained model from $repo_url"
git lfs install
git clone $repo_url
repo=$(basename $repo_url)
log "Display test files"
tree $repo/
ls -lh $repo/test_wavs/*.wav
log "CTC greedy search"
./zipformer/onnx_pretrained_ctc.py \
--nn-model $repo/model.onnx \
--tokens $repo/tokens.txt \
$repo/test_wavs/0.wav \
$repo/test_wavs/1.wav \
$repo/test_wavs/2.wav
log "CTC H decoding"
./zipformer/onnx_pretrained_ctc_H.py \
--nn-model $repo/model.onnx \
--tokens $repo/tokens.txt \
--H $repo/H.fst \
$repo/test_wavs/0.wav \
$repo/test_wavs/1.wav \
$repo/test_wavs/2.wav
log "CTC HL decoding"
./zipformer/onnx_pretrained_ctc_HL.py \
--nn-model $repo/model.onnx \
--words $repo/words.txt \
--HL $repo/HL.fst \
$repo/test_wavs/0.wav \
$repo/test_wavs/1.wav \
$repo/test_wavs/2.wav
log "CTC HLG decoding"
./zipformer/onnx_pretrained_ctc_HLG.py \
--nn-model $repo/model.onnx \
--words $repo/words.txt \
--HLG $repo/HLG.fst \
$repo/test_wavs/0.wav \
$repo/test_wavs/1.wav \
$repo/test_wavs/2.wav
rm -rf $repo
repo_url=https://huggingface.co/csukuangfj/icefall-asr-librispeech-conformer-ctc-jit-bpe-500-2021-11-09
log "Downloading pre-trained model from $repo_url"
GIT_LFS_SKIP_SMUDGE=1 git clone $repo_url
repo=$(basename $repo_url)
pushd $repo
git lfs pull --include "exp/pretrained.pt"
git lfs pull --include "data/lang_bpe_500/HLG.pt"
git lfs pull --include "data/lang_bpe_500/L.pt"
git lfs pull --include "data/lang_bpe_500/L_disambig.pt"
git lfs pull --include "data/lang_bpe_500/Linv.pt"
git lfs pull --include "data/lang_bpe_500/bpe.model"
git lfs pull --include "data/lang_bpe_500/lexicon.txt"
git lfs pull --include "data/lang_bpe_500/lexicon_disambig.txt"
git lfs pull --include "data/lang_bpe_500/tokens.txt"
git lfs pull --include "data/lang_bpe_500/words.txt"
git lfs pull --include "data/lm/G_3_gram.fst.txt"
popd
log "Display test files"
tree $repo/
ls -lh $repo/test_wavs/*.wav
log "CTC decoding"
./conformer_ctc/pretrained.py \
--method ctc-decoding \
--num-classes 500 \
--checkpoint $repo/exp/pretrained.pt \
--tokens $repo/data/lang_bpe_500/tokens.txt \
$repo/test_wavs/1089-134686-0001.wav \
$repo/test_wavs/1221-135766-0001.wav \
$repo/test_wavs/1221-135766-0002.wav
log "HLG decoding"
./conformer_ctc/pretrained.py \
--method 1best \
--num-classes 500 \
--checkpoint $repo/exp/pretrained.pt \
--tokens $repo/data/lang_bpe_500/tokens.txt \
--words-file $repo/data/lang_bpe_500/words.txt \
--HLG $repo/data/lang_bpe_500/HLG.pt \
$repo/test_wavs/1089-134686-0001.wav \
$repo/test_wavs/1221-135766-0001.wav \
$repo/test_wavs/1221-135766-0002.wav
log "CTC decoding on CPU with kaldi decoders using OpenFst"
log "Exporting model with torchscript"
pushd $repo/exp
ln -s pretrained.pt epoch-99.pt
popd
./conformer_ctc/export.py \
--epoch 99 \
--avg 1 \
--exp-dir $repo/exp \
--tokens $repo/data/lang_bpe_500/tokens.txt \
--jit 1
ls -lh $repo/exp
log "Generating H.fst, HL.fst"
./local/prepare_lang_fst.py --lang-dir $repo/data/lang_bpe_500 --ngram-G $repo/data/lm/G_3_gram.fst.txt
ls -lh $repo/data/lang_bpe_500
log "Decoding with H on CPU with OpenFst"
./conformer_ctc/jit_pretrained_decode_with_H.py \
--nn-model $repo/exp/cpu_jit.pt \
--H $repo/data/lang_bpe_500/H.fst \
--tokens $repo/data/lang_bpe_500/tokens.txt \
$repo/test_wavs/1089-134686-0001.wav \
$repo/test_wavs/1221-135766-0001.wav \
$repo/test_wavs/1221-135766-0002.wav
log "Decoding with HL on CPU with OpenFst"
./conformer_ctc/jit_pretrained_decode_with_HL.py \
--nn-model $repo/exp/cpu_jit.pt \
--HL $repo/data/lang_bpe_500/HL.fst \
--words $repo/data/lang_bpe_500/words.txt \
$repo/test_wavs/1089-134686-0001.wav \
$repo/test_wavs/1221-135766-0001.wav \
$repo/test_wavs/1221-135766-0002.wav
log "Decoding with HLG on CPU with OpenFst"
./conformer_ctc/jit_pretrained_decode_with_HLG.py \
--nn-model $repo/exp/cpu_jit.pt \
--HLG $repo/data/lang_bpe_500/HLG.fst \
--words $repo/data/lang_bpe_500/words.txt \
$repo/test_wavs/1089-134686-0001.wav \
$repo/test_wavs/1221-135766-0001.wav \
$repo/test_wavs/1221-135766-0002.wav
rm -rf $repo
popd
log "Test aishell"
pushd egs/aishell/ASR
repo_url=https://huggingface.co/csukuangfj/icefall_asr_aishell_conformer_ctc
log "Downloading pre-trained model from $repo_url"
GIT_LFS_SKIP_SMUDGE=1 git clone $repo_url
repo=$(basename $repo_url)
pushd $repo
git lfs pull --include "exp/pretrained.pt"
git lfs pull --include "data/lang_char/H.fst"
git lfs pull --include "data/lang_char/HL.fst"
git lfs pull --include "data/lang_char/HLG.fst"
popd
log "Display test files"
tree $repo/
ls -lh $repo/test_wavs/*.wav
log "CTC decoding"
log "Exporting model with torchscript"
pushd $repo/exp
ln -s pretrained.pt epoch-99.pt
popd
./conformer_ctc/export.py \
--epoch 99 \
--avg 1 \
--exp-dir $repo/exp \
--tokens $repo/data/lang_char/tokens.txt \
--jit 1
ls -lh $repo/exp
ls -lh $repo/data/lang_char
log "Decoding with H on CPU with OpenFst"
./conformer_ctc/jit_pretrained_decode_with_H.py \
--nn-model $repo/exp/cpu_jit.pt \
--H $repo/data/lang_char/H.fst \
--tokens $repo/data/lang_char/tokens.txt \
$repo/test_wavs/0.wav \
$repo/test_wavs/1.wav \
$repo/test_wavs/2.wav
log "Decoding with HL on CPU with OpenFst"
./conformer_ctc/jit_pretrained_decode_with_HL.py \
--nn-model $repo/exp/cpu_jit.pt \
--HL $repo/data/lang_char/HL.fst \
--words $repo/data/lang_char/words.txt \
$repo/test_wavs/0.wav \
$repo/test_wavs/1.wav \
$repo/test_wavs/2.wav
log "Decoding with HLG on CPU with OpenFst"
./conformer_ctc/jit_pretrained_decode_with_HLG.py \
--nn-model $repo/exp/cpu_jit.pt \
--HLG $repo/data/lang_char/HLG.fst \
--words $repo/data/lang_char/words.txt \
$repo/test_wavs/0.wav \
$repo/test_wavs/1.wav \
$repo/test_wavs/2.wav
rm -rf $repo

77
.github/scripts/run-pre-trained-transducer-stateless-librispeech-100h.sh vendored
View File

@ -1,77 +0,0 @@
#!/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]}) $*"
}
cd egs/librispeech/ASR
repo_url=https://huggingface.co/csukuangfj/icefall-asr-librispeech-100h-transducer-stateless-multi-datasets-bpe-500-2022-02-21
log "Downloading pre-trained model from $repo_url"
git lfs install
git clone $repo_url
repo=$(basename $repo_url)
log "Display test files"
tree $repo/
ls -lh $repo/test_wavs/*.wav
for sym in 1 2 3; do
log "Greedy search with --max-sym-per-frame $sym"
./transducer_stateless_multi_datasets/pretrained.py \
--method greedy_search \
--max-sym-per-frame $sym \
--checkpoint $repo/exp/pretrained.pt \
--tokens $repo/data/lang_bpe_500/tokens.txt \
$repo/test_wavs/1089-134686-0001.wav \
$repo/test_wavs/1221-135766-0001.wav \
$repo/test_wavs/1221-135766-0002.wav
done
for method in modified_beam_search beam_search fast_beam_search; do
log "$method"
./transducer_stateless_multi_datasets/pretrained.py \
--method $method \
--beam-size 4 \
--checkpoint $repo/exp/pretrained.pt \
--tokens $repo/data/lang_bpe_500/tokens.txt \
$repo/test_wavs/1089-134686-0001.wav \
$repo/test_wavs/1221-135766-0001.wav \
$repo/test_wavs/1221-135766-0002.wav
done
echo "GITHUB_EVENT_NAME: ${GITHUB_EVENT_NAME}"
echo "GITHUB_EVENT_LABEL_NAME: ${GITHUB_EVENT_LABEL_NAME}"
if [[ x"${GITHUB_EVENT_NAME}" == x"schedule" || x"${GITHUB_EVENT_LABEL_NAME}" == x"run-decode" ]]; then
mkdir -p transducer_stateless_multi_datasets/exp
ln -s $PWD/$repo/exp/pretrained.pt transducer_stateless_multi_datasets/exp/epoch-999.pt
ln -s $PWD/$repo/data/lang_bpe_500 data/
ls -lh data
ls -lh transducer_stateless_multi_datasets/exp
log "Decoding test-clean and test-other"
# use a small value for decoding with CPU
max_duration=100
for method in greedy_search fast_beam_search modified_beam_search; do
log "Decoding with $method"
./transducer_stateless_multi_datasets/decode.py \
--decoding-method $method \
--epoch 999 \
--avg 1 \
--max-duration $max_duration \
--exp-dir transducer_stateless_multi_datasets/exp
done
rm transducer_stateless_multi_datasets/exp/*.pt
fi

77
.github/scripts/run-pre-trained-transducer-stateless-librispeech-960h.sh vendored
View File

@ -1,77 +0,0 @@
#!/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]}) $*"
}
cd egs/librispeech/ASR
repo_url=https://huggingface.co/csukuangfj/icefall-asr-librispeech-transducer-stateless-multi-datasets-bpe-500-2022-03-01
log "Downloading pre-trained model from $repo_url"
git lfs install
git clone $repo_url
repo=$(basename $repo_url)
log "Display test files"
tree $repo/
ls -lh $repo/test_wavs/*.wav
for sym in 1 2 3; do
log "Greedy search with --max-sym-per-frame $sym"
./transducer_stateless_multi_datasets/pretrained.py \
--method greedy_search \
--max-sym-per-frame $sym \
--checkpoint $repo/exp/pretrained.pt \
--tokens $repo/data/lang_bpe_500/tokens.txt \
$repo/test_wavs/1089-134686-0001.wav \
$repo/test_wavs/1221-135766-0001.wav \
$repo/test_wavs/1221-135766-0002.wav
done
for method in modified_beam_search beam_search fast_beam_search; do
log "$method"
./transducer_stateless_multi_datasets/pretrained.py \
--method $method \
--beam-size 4 \
--checkpoint $repo/exp/pretrained.pt \
--tokens $repo/data/lang_bpe_500/tokens.txt \
$repo/test_wavs/1089-134686-0001.wav \
$repo/test_wavs/1221-135766-0001.wav \
$repo/test_wavs/1221-135766-0002.wav
done
echo "GITHUB_EVENT_NAME: ${GITHUB_EVENT_NAME}"
echo "GITHUB_EVENT_LABEL_NAME: ${GITHUB_EVENT_LABEL_NAME}"
if [[ x"${GITHUB_EVENT_NAME}" == x"schedule" || x"${GITHUB_EVENT_LABEL_NAME}" == x"run-decode" ]]; then
mkdir -p transducer_stateless_multi_datasets/exp
ln -s $PWD/$repo/exp/pretrained.pt transducer_stateless_multi_datasets/exp/epoch-999.pt
ln -s $PWD/$repo/data/lang_bpe_500 data/
ls -lh data
ls -lh transducer_stateless_multi_datasets/exp
log "Decoding test-clean and test-other"
# use a small value for decoding with CPU
max_duration=100
for method in greedy_search fast_beam_search modified_beam_search; do
log "Decoding with $method"
./transducer_stateless_multi_datasets/decode.py \
--decoding-method $method \
--epoch 999 \
--avg 1 \
--max-duration $max_duration \
--exp-dir transducer_stateless_multi_datasets/exp
done
rm transducer_stateless_multi_datasets/exp/*.pt
fi

48
.github/scripts/run-pre-trained-transducer-stateless-modified-2-aishell.sh vendored
View File

@ -1,48 +0,0 @@
#!/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]}) $*"
}
cd egs/aishell/ASR
repo_url=https://huggingface.co/csukuangfj/icefall-aishell-transducer-stateless-modified-2-2022-03-01
log "Downloading pre-trained model from $repo_url"
git lfs install
git clone $repo_url
repo=$(basename $repo_url)
log "Display test files"
tree $repo/
ls -lh $repo/test_wavs/*.wav
for sym in 1 2 3; do
log "Greedy search with --max-sym-per-frame $sym"
./transducer_stateless_modified-2/pretrained.py \
--method greedy_search \
--max-sym-per-frame $sym \
--checkpoint $repo/exp/pretrained.pt \
--lang-dir $repo/data/lang_char \
$repo/test_wavs/BAC009S0764W0121.wav \
$repo/test_wavs/BAC009S0764W0122.wav \
$repo/test_wavs/BAC009S0764W0123.wav
done
for method in modified_beam_search beam_search; do
log "$method"
./transducer_stateless_modified-2/pretrained.py \
--method $method \
--beam-size 4 \
--checkpoint $repo/exp/pretrained.pt \
--lang-dir $repo/data/lang_char \
$repo/test_wavs/BAC009S0764W0121.wav \
$repo/test_wavs/BAC009S0764W0122.wav \
$repo/test_wavs/BAC009S0764W0123.wav
done

48
.github/scripts/run-pre-trained-transducer-stateless-modified-aishell.sh vendored
View File

@ -1,48 +0,0 @@
#!/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]}) $*"
}
cd egs/aishell/ASR
repo_url=https://huggingface.co/csukuangfj/icefall-aishell-transducer-stateless-modified-2022-03-01
log "Downloading pre-trained model from $repo_url"
git lfs install
git clone $repo_url
repo=$(basename $repo_url)
log "Display test files"
tree $repo/
ls -lh $repo/test_wavs/*.wav
for sym in 1 2 3; do
log "Greedy search with --max-sym-per-frame $sym"
./transducer_stateless_modified/pretrained.py \
--method greedy_search \
--max-sym-per-frame $sym \
--checkpoint $repo/exp/pretrained.pt \
--lang-dir $repo/data/lang_char \
$repo/test_wavs/BAC009S0764W0121.wav \
$repo/test_wavs/BAC009S0764W0122.wav \
$repo/test_wavs/BAC009S0764W0123.wav
done
for method in modified_beam_search beam_search; do
log "$method"
./transducer_stateless_modified/pretrained.py \
--method $method \
--beam-size 4 \
--checkpoint $repo/exp/pretrained.pt \
--lang-dir $repo/data/lang_char \
$repo/test_wavs/BAC009S0764W0121.wav \
$repo/test_wavs/BAC009S0764W0122.wav \
$repo/test_wavs/BAC009S0764W0123.wav
done

77
.github/scripts/run-pre-trained-transducer-stateless.sh vendored
View File

@ -1,77 +0,0 @@
#!/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]}) $*"
}
cd egs/librispeech/ASR
repo_url=https://huggingface.co/csukuangfj/icefall-asr-librispeech-transducer-stateless-bpe-500-2022-02-07
log "Downloading pre-trained model from $repo_url"
git lfs install
git clone $repo_url
repo=$(basename $repo_url)
log "Display test files"
tree $repo/
ls -lh $repo/test_wavs/*.wav
for sym in 1 2 3; do
log "Greedy search with --max-sym-per-frame $sym"
./transducer_stateless/pretrained.py \
--method greedy_search \
--max-sym-per-frame $sym \
--checkpoint $repo/exp/pretrained.pt \
--tokens $repo/data/lang_bpe_500/tokens.txt \
$repo/test_wavs/1089-134686-0001.wav \
$repo/test_wavs/1221-135766-0001.wav \
$repo/test_wavs/1221-135766-0002.wav
done
for method in fast_beam_search modified_beam_search beam_search; do
log "$method"
./transducer_stateless/pretrained.py \
--method $method \
--beam-size 4 \
--checkpoint $repo/exp/pretrained.pt \
--tokens $repo/data/lang_bpe_500/tokens.txt \
$repo/test_wavs/1089-134686-0001.wav \
$repo/test_wavs/1221-135766-0001.wav \
$repo/test_wavs/1221-135766-0002.wav
done
echo "GITHUB_EVENT_NAME: ${GITHUB_EVENT_NAME}"
echo "GITHUB_EVENT_LABEL_NAME: ${GITHUB_EVENT_LABEL_NAME}"
if [[ x"${GITHUB_EVENT_NAME}" == x"schedule" || x"${GITHUB_EVENT_LABEL_NAME}" == x"run-decode" ]]; then
mkdir -p transducer_stateless/exp
ln -s $PWD/$repo/exp/pretrained.pt transducer_stateless/exp/epoch-999.pt
ln -s $PWD/$repo/data/lang_bpe_500 data/
ls -lh data
ls -lh transducer_stateless/exp
log "Decoding test-clean and test-other"
# use a small value for decoding with CPU
max_duration=100
for method in greedy_search fast_beam_search modified_beam_search; do
log "Decoding with $method"
./transducer_stateless/decode.py \
--decoding-method $method \
--epoch 999 \
--avg 1 \
--max-duration $max_duration \
--exp-dir transducer_stateless/exp
done
rm transducer_stateless/exp/*.pt
fi

33
.github/scripts/run-pre-trained-transducer.sh vendored
View File

@ -1,33 +0,0 @@
#!/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]}) $*"
}
cd egs/librispeech/ASR
repo_url=https://huggingface.co/csukuangfj/icefall-asr-librispeech-transducer-bpe-500-2021-12-23
log "Downloading pre-trained model from $repo_url"
git lfs install
git clone $repo_url
repo=$(basename $repo_url)
log "Display test files"
tree $repo/
ls -lh $repo/test_wavs/*.wav
log "Beam search decoding"
./transducer/pretrained.py \
--method beam_search \
--beam-size 4 \
--checkpoint $repo/exp/pretrained.pt \
--tokens $repo/data/lang_bpe_500/tokens.txt \
$repo/test_wavs/1089-134686-0001.wav \
$repo/test_wavs/1221-135766-0001.wav \
$repo/test_wavs/1221-135766-0002.wav

6
.github/scripts/run-wenetspeech-pruned-transducer-stateless2.sh vendored
View File

@ -30,7 +30,7 @@ log "Test exporting to ONNX format"
./pruned_transducer_stateless2/export-onnx.py \
--exp-dir $repo/exp \
--lang-dir $repo/data/lang_char \
--tokens $repo/data/lang_char/tokens.txt \
--epoch 99 \
--avg 1
@ -38,14 +38,14 @@ log "Export to torchscript model"
./pruned_transducer_stateless2/export.py \
--exp-dir $repo/exp \
--lang-dir $repo/data/lang_char \
--tokens $repo/data/lang_char/tokens.txt \
--epoch 99 \
--avg 1 \
--jit 1
./pruned_transducer_stateless2/export.py \
--exp-dir $repo/exp \
--lang-dir $repo/data/lang_char \
--tokens $repo/data/lang_char/tokens.txt \
--epoch 99 \
--avg 1 \
--jit-trace 1

86
.github/scripts/yesno/ASR/run.sh vendored Executable file
View File

@ -0,0 +1,86 @@
#!/usr/bin/env bash
set -ex
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]}) $*"
}
cd egs/yesno/ASR
log "data preparation"
./prepare.sh
log "training"
python3 ./tdnn/train.py
log "decoding"
python3 ./tdnn/decode.py
log "export to pretrained.pt"
python3 ./tdnn/export.py --epoch 14 --avg 2
python3 ./tdnn/pretrained.py \
--checkpoint ./tdnn/exp/pretrained.pt \
--HLG ./data/lang_phone/HLG.pt \
--words-file ./data/lang_phone/words.txt \
download/waves_yesno/0_0_0_1_0_0_0_1.wav \
download/waves_yesno/0_0_1_0_0_0_1_0.wav
log "Test exporting to torchscript"
python3 ./tdnn/export.py --epoch 14 --avg 2 --jit 1
python3 ./tdnn/jit_pretrained.py \
--nn-model ./tdnn/exp/cpu_jit.pt \
--HLG ./data/lang_phone/HLG.pt \
--words-file ./data/lang_phone/words.txt \
download/waves_yesno/0_0_0_1_0_0_0_1.wav \
download/waves_yesno/0_0_1_0_0_0_1_0.wav
log "Test exporting to onnx"
python3 ./tdnn/export_onnx.py --epoch 14 --avg 2
log "Test float32 model"
python3 ./tdnn/onnx_pretrained.py \
--nn-model ./tdnn/exp/model-epoch-14-avg-2.onnx \
--HLG ./data/lang_phone/HLG.pt \
--words-file ./data/lang_phone/words.txt \
download/waves_yesno/0_0_0_1_0_0_0_1.wav \
download/waves_yesno/0_0_1_0_0_0_1_0.wav
log "Test int8 model"
python3 ./tdnn/onnx_pretrained.py \
--nn-model ./tdnn/exp/model-epoch-14-avg-2.int8.onnx \
--HLG ./data/lang_phone/HLG.pt \
--words-file ./data/lang_phone/words.txt \
download/waves_yesno/0_0_0_1_0_0_0_1.wav \
download/waves_yesno/0_0_1_0_0_0_1_0.wav
log "Test decoding with H"
python3 ./tdnn/export.py --epoch 14 --avg 2 --jit 1
python3 ./tdnn/jit_pretrained_decode_with_H.py \
--nn-model ./tdnn/exp/cpu_jit.pt \
--H ./data/lang_phone/H.fst \
--tokens ./data/lang_phone/tokens.txt \
./download/waves_yesno/0_0_0_1_0_0_0_1.wav \
./download/waves_yesno/0_0_1_0_0_0_1_0.wav \
./download/waves_yesno/0_0_1_0_0_1_1_1.wav
log "Test decoding with HL"
python3 ./tdnn/export.py --epoch 14 --avg 2 --jit 1
python3 ./tdnn/jit_pretrained_decode_with_HL.py \
--nn-model ./tdnn/exp/cpu_jit.pt \
--HL ./data/lang_phone/HL.fst \
--words ./data/lang_phone/words.txt \
./download/waves_yesno/0_0_0_1_0_0_0_1.wav \
./download/waves_yesno/0_0_1_0_0_0_1_0.wav \
./download/waves_yesno/0_0_1_0_0_1_1_1.wav
log "Show generated files"
ls -lh tdnn/exp
ls -lh data/lang_phone

72
.github/workflows/aishell.yml vendored Normal file
View File

@ -0,0 +1,72 @@
name: aishell
on:
push:
branches:
- master
pull_request:
branches:
- master
workflow_dispatch:
concurrency:
group: aishell-${{ github.ref }}
cancel-in-progress: true
jobs:
generate_build_matrix:
if: (github.repository_owner == 'csukuangfj' || github.repository_owner == 'k2-fsa') && (github.event.label.name == 'ready' || github.event_name == 'push' || github.event_name == 'aishell')
# see https://github.com/pytorch/pytorch/pull/50633
runs-on: ubuntu-latest
outputs:
matrix: ${{ steps.set-matrix.outputs.matrix }}
steps:
- uses: actions/checkout@v4
with:
fetch-depth: 0
- name: Generating build matrix
id: set-matrix
run: |
# outputting for debugging purposes
python ./.github/scripts/docker/generate_build_matrix.py
MATRIX=$(python ./.github/scripts/docker/generate_build_matrix.py)
echo "::set-output name=matrix::${MATRIX}"
aishell:
needs: generate_build_matrix
name: py${{ matrix.python-version }} torch${{ matrix.torch-version }} v${{ matrix.version }}
runs-on: ubuntu-latest
strategy:
fail-fast: false
matrix:
${{ fromJson(needs.generate_build_matrix.outputs.matrix) }}
steps:
- uses: actions/checkout@v4
with:
fetch-depth: 0
- name: Free space
shell: bash
run: |
df -h
rm -rf /opt/hostedtoolcache
df -h
echo "pwd: $PWD"
echo "github.workspace ${{ github.workspace }}"
- name: Run aishell tests
uses: addnab/docker-run-action@v3
with:
image: ghcr.io/${{ github.repository_owner }}/icefall:cpu-py${{ matrix.python-version }}-torch${{ matrix.torch-version }}-v${{ matrix.version }}
options: |
--volume ${{ github.workspace }}/:/icefall
shell: bash
run: |
export PYTHONPATH=/icefall:$PYTHONPATH
cd /icefall
git config --global --add safe.directory /icefall
.github/scripts/aishell/ASR/run.sh

137
.github/workflows/audioset.yml vendored Normal file
View File

@ -0,0 +1,137 @@
name: audioset
on:
push:
branches:
- master
pull_request:
branches:
- master
workflow_dispatch:
concurrency:
group: audioset-${{ github.ref }}
cancel-in-progress: true
jobs:
generate_build_matrix:
if: github.repository_owner == 'csukuangfj' || github.repository_owner == 'k2-fsa'
# see https://github.com/pytorch/pytorch/pull/50633
runs-on: ubuntu-latest
outputs:
matrix: ${{ steps.set-matrix.outputs.matrix }}
steps:
- uses: actions/checkout@v4
with:
fetch-depth: 0
- name: Generating build matrix
id: set-matrix
run: |
# outputting for debugging purposes
python ./.github/scripts/docker/generate_build_matrix.py
MATRIX=$(python ./.github/scripts/docker/generate_build_matrix.py)
echo "::set-output name=matrix::${MATRIX}"
audioset:
needs: generate_build_matrix
name: py${{ matrix.python-version }} torch${{ matrix.torch-version }} v${{ matrix.version }}
runs-on: ubuntu-latest
strategy:
fail-fast: false
matrix:
${{ fromJson(needs.generate_build_matrix.outputs.matrix) }}
steps:
- uses: actions/checkout@v4
with:
fetch-depth: 0
- name: Free space
shell: bash
run: |
ls -lh
df -h
rm -rf /opt/hostedtoolcache
df -h
echo "pwd: $PWD"
echo "github.workspace ${{ github.workspace }}"
- name: Run tests
uses: addnab/docker-run-action@v3
with:
image: ghcr.io/${{ github.repository_owner }}/icefall:cpu-py${{ matrix.python-version }}-torch${{ matrix.torch-version }}-v${{ matrix.version }}
options: |
--volume ${{ github.workspace }}/:/icefall
shell: bash
run: |
export PYTHONPATH=/icefall:$PYTHONPATH
cd /icefall
git config --global --add safe.directory /icefall
.github/scripts/audioset/AT/run.sh
- name: Show model files
shell: bash
run: |
sudo chown -R runner ./model-onnx
ls -lh ./model-onnx
chmod -x ./model-onnx/class_labels_indices.csv
echo "----------"
ls -lh ./model-onnx/*
- name: Upload model to huggingface
if: matrix.python-version == '3.9' && matrix.torch-version == '2.2.0' && github.event_name == 'push'
env:
HF_TOKEN: ${{ secrets.HF_TOKEN }}
uses: nick-fields/retry@v3
with:
max_attempts: 20
timeout_seconds: 200
shell: bash
command: |
git config --global user.email "csukuangfj@gmail.com"
git config --global user.name "Fangjun Kuang"
rm -rf huggingface
export GIT_LFS_SKIP_SMUDGE=1
git clone https://huggingface.co/k2-fsa/sherpa-onnx-zipformer-audio-tagging-2024-04-09 huggingface
cd huggingface
git fetch
git pull
git merge -m "merge remote" --ff origin main
cp ../model-onnx/*.onnx ./
cp ../model-onnx/*.csv ./
cp -a ../model-onnx/test_wavs ./
ls -lh
git add .
git status
git commit -m "update models"
git status
git push https://csukuangfj:$HF_TOKEN@huggingface.co/k2-fsa/sherpa-onnx-zipformer-audio-tagging-2024-04-09 main || true
rm -rf huggingface
- name: Prepare for release
if: matrix.python-version == '3.9' && matrix.torch-version == '2.2.0' && github.event_name == 'push'
shell: bash
run: |
d=sherpa-onnx-zipformer-audio-tagging-2024-04-09
mv ./model-onnx $d
tar cjvf ${d}.tar.bz2 $d
ls -lh
- name: Release exported onnx models
if: matrix.python-version == '3.9' && matrix.torch-version == '2.2.0' && github.event_name == 'push'
uses: svenstaro/upload-release-action@v2
with:
file_glob: true
overwrite: true
file: sherpa-onnx-*.tar.bz2
repo_name: k2-fsa/sherpa-onnx
repo_token: ${{ secrets.UPLOAD_GH_SHERPA_ONNX_TOKEN }}
tag: audio-tagging-models

81
.github/workflows/build-cpu-docker.yml vendored Normal file
View File

@ -0,0 +1,81 @@
name: build-cpu-docker
on:
workflow_dispatch:
concurrency:
group: build-cpu-docker-${{ github.ref }}
cancel-in-progress: true
jobs:
generate_build_matrix:
if: github.repository_owner == 'csukuangfj' || github.repository_owner == 'k2-fsa'
# see https://github.com/pytorch/pytorch/pull/50633
runs-on: ubuntu-latest
outputs:
matrix: ${{ steps.set-matrix.outputs.matrix }}
steps:
- uses: actions/checkout@v4
with:
fetch-depth: 0
- name: Generating build matrix
id: set-matrix
run: |
# outputting for debugging purposes
python ./.github/scripts/docker/generate_build_matrix.py
MATRIX=$(python ./.github/scripts/docker/generate_build_matrix.py)
echo "::set-output name=matrix::${MATRIX}"
build-cpu-docker:
needs: generate_build_matrix
name: py${{ matrix.python-version }} torch${{ matrix.torch-version }} v${{ matrix.version }}
runs-on: ubuntu-latest
strategy:
fail-fast: false
matrix:
${{ fromJson(needs.generate_build_matrix.outputs.matrix) }}
steps:
# refer to https://github.com/actions/checkout
- uses: actions/checkout@v4
with:
fetch-depth: 0
- name: Free space
shell: bash
run: |
df -h
rm -rf /opt/hostedtoolcache
df -h
- name: 'Login to GitHub Container Registry'
uses: docker/login-action@v2
with:
registry: ghcr.io
username: ${{ github.actor }}
password: ${{ secrets.GITHUB_TOKEN }}
- name: Build docker Image
shell: bash
run: |
cd .github/scripts/docker
torch_version=${{ matrix.torch-version }}
torchaudio_version=${{ matrix.torchaudio-version }}
echo "torch_version: $torch_version"
echo "torchaudio_version: $torchaudio_version"
version=${{ matrix.version }}
tag=ghcr.io/${{ github.repository_owner }}/icefall:cpu-py${{ matrix.python-version }}-torch${{ matrix.torch-version }}-v$version
echo "tag: $tag"
docker build \
-t $tag \
--build-arg PYTHON_VERSION=${{ matrix.python-version }} \
--build-arg TORCH_VERSION=$torch_version \
--build-arg TORCHAUDIO_VERSION=$torchaudio_version \
--build-arg K2_VERSION=${{ matrix.k2-version }} \
--build-arg KALDIFEAT_VERSION=${{ matrix.kaldifeat-version }} \
.
docker image ls
docker push $tag

3
.github/workflows/build-doc.yml vendored
View File

@ -56,11 +56,14 @@ jobs:
- name: Build doc
shell: bash
run: |
.github/scripts/generate-piper-phonemize-page.py
cd docs
python3 -m pip install -r ./requirements.txt
make html
touch build/html/.nojekyll
cp -v ../piper_phonemize.html ./build/html/
- name: Deploy
uses: peaceiris/actions-gh-pages@v3
with:

2
.github/workflows/build-docker-image.yml vendored
View File

@ -16,7 +16,7 @@ jobs:
fail-fast: false
matrix:
os: [ubuntu-latest]
image: ["torch2.1.0-cuda12.1", "torch2.1.0-cuda11.8", "torch2.0.0-cuda11.7", "torch1.13.0-cuda11.6", "torch1.12.1-cuda11.3", "torch1.9.0-cuda10.2"]
image: ["torch2.2.2-cuda12.1", "torch2.2.2-cuda11.8", "torch2.2.1-cuda12.1", "torch2.2.1-cuda11.8", "torch2.2.0-cuda12.1", "torch2.2.0-cuda11.8", "torch2.1.0-cuda12.1", "torch2.1.0-cuda11.8", "torch2.0.0-cuda11.7", "torch1.13.0-cuda11.6", "torch1.12.1-cuda11.3", "torch1.9.0-cuda10.2"]
steps:
# refer to https://github.com/actions/checkout

71
.github/workflows/librispeech.yml vendored Normal file
View File

@ -0,0 +1,71 @@
name: librispeech
on:
push:
branches:
- master
pull_request:
branches:
- master
workflow_dispatch:
concurrency:
group: librispeech-${{ github.ref }}
cancel-in-progress: true
jobs:
generate_build_matrix:
if: github.repository_owner == 'csukuangfj' || github.repository_owner == 'k2-fsa'
# see https://github.com/pytorch/pytorch/pull/50633
runs-on: ubuntu-latest
outputs:
matrix: ${{ steps.set-matrix.outputs.matrix }}
steps:
- uses: actions/checkout@v4
with:
fetch-depth: 0
- name: Generating build matrix
id: set-matrix
run: |
# outputting for debugging purposes
python ./.github/scripts/docker/generate_build_matrix.py
MATRIX=$(python ./.github/scripts/docker/generate_build_matrix.py)
echo "::set-output name=matrix::${MATRIX}"
librispeech:
needs: generate_build_matrix
name: py${{ matrix.python-version }} torch${{ matrix.torch-version }} v${{ matrix.version }}
runs-on: ubuntu-latest
strategy:
fail-fast: false
matrix:
${{ fromJson(needs.generate_build_matrix.outputs.matrix) }}
steps:
# refer to https://github.com/actions/checkout
- uses: actions/checkout@v4
with:
fetch-depth: 0
- name: Free space
shell: bash
run: |
df -h
rm -rf /opt/hostedtoolcache
df -h
echo "pwd: $PWD"
echo "github.workspace ${{ github.workspace }}"
- name: Test zipformer/train.py with LibriSpeech
uses: addnab/docker-run-action@v3
with:
image: ghcr.io/${{ github.repository_owner }}/icefall:cpu-py${{ matrix.python-version }}-torch${{ matrix.torch-version }}-v${{ matrix.version }}
options: |
--volume ${{ github.workspace }}/:/icefall
shell: bash
run: |
export PYTHONPATH=/icefall:$PYTHONPATH
cd /icefall
git config --global --add safe.directory /icefall
.github/scripts/librispeech/ASR/run.sh

102
.github/workflows/ljspeech.yml vendored Normal file
View File

@ -0,0 +1,102 @@
name: ljspeech
on:
push:
branches:
- master
pull_request:
branches:
- master
workflow_dispatch:
concurrency:
group: ljspeech-${{ github.ref }}
cancel-in-progress: true
jobs:
generate_build_matrix:
if: github.repository_owner == 'csukuangfj' || github.repository_owner == 'k2-fsa'
# see https://github.com/pytorch/pytorch/pull/50633
runs-on: ubuntu-latest
outputs:
matrix: ${{ steps.set-matrix.outputs.matrix }}
steps:
- uses: actions/checkout@v4
with:
fetch-depth: 0
- name: Generating build matrix
id: set-matrix
run: |
# outputting for debugging purposes
python ./.github/scripts/docker/generate_build_matrix.py
MATRIX=$(python ./.github/scripts/docker/generate_build_matrix.py)
echo "::set-output name=matrix::${MATRIX}"
ljspeech:
needs: generate_build_matrix
name: py${{ matrix.python-version }} torch${{ matrix.torch-version }} v${{ matrix.version }}
runs-on: ubuntu-latest
strategy:
fail-fast: false
matrix:
${{ fromJson(needs.generate_build_matrix.outputs.matrix) }}
steps:
- uses: actions/checkout@v4
with:
fetch-depth: 0
- name: Free space
shell: bash
run: |
ls -lh
df -h
rm -rf /opt/hostedtoolcache
df -h
echo "pwd: $PWD"
echo "github.workspace ${{ github.workspace }}"
- name: Run tests
uses: addnab/docker-run-action@v3
with:
image: ghcr.io/${{ github.repository_owner }}/icefall:cpu-py${{ matrix.python-version }}-torch${{ matrix.torch-version }}-v${{ matrix.version }}
options: |
--volume ${{ github.workspace }}/:/icefall
shell: bash
run: |
export PYTHONPATH=/icefall:$PYTHONPATH
cd /icefall
git config --global --add safe.directory /icefall
.github/scripts/ljspeech/TTS/run.sh
- name: display files
shell: bash
run: |
ls -lh
- uses: actions/upload-artifact@v4
if: matrix.python-version == '3.9' && matrix.torch-version == '2.2.0'
with:
name: generated-test-files-${{ matrix.python-version }}-${{ matrix.torch-version }}
path: ./*.wav
- uses: actions/upload-artifact@v4
if: matrix.python-version == '3.9' && matrix.torch-version == '2.2.0'
with:
name: generated-models-py${{ matrix.python-version }}-torch${{ matrix.torch-version }}
path: ./*.wav
- name: Release exported onnx models
if: matrix.python-version == '3.9' && matrix.torch-version == '2.2.0' && github.event_name == 'push'
uses: svenstaro/upload-release-action@v2
with:
file_glob: true
overwrite: true
file: vits-icefall-*.tar.bz2
repo_name: k2-fsa/sherpa-onnx
repo_token: ${{ secrets.UPLOAD_GH_SHERPA_ONNX_TOKEN }}
tag: tts-models

55
.github/workflows/run-pretrained-transducer.yml → .github/workflows/multi-zh-hans.yml vendored
View File

@ -1,45 +1,30 @@
# Copyright 2021 Fangjun Kuang (csukuangfj@gmail.com)
# See ../../LICENSE for clarification regarding multiple authors
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
name: run-pre-trained-transducer
name: run-multi-zh-hans
on:
push:
branches:
- master
pull_request:
types: [labeled]
workflow_dispatch:
concurrency:
group: run_pre_trained_transducer-${{ github.ref }}
group: run-multi-zh-hans-${{ github.ref }}
cancel-in-progress: true
permissions:
contents: write
jobs:
run_pre_trained_transducer:
if: github.event.label.name == 'ready' || github.event_name == 'push'
multi-zh-hans:
runs-on: ${{ matrix.os }}
strategy:
fail-fast: false
matrix:
os: [ubuntu-latest]
python-version: [3.8]
fail-fast: false
steps:
- uses: actions/checkout@v2
- uses: actions/checkout@v4
with:
fetch-depth: 0
@ -68,13 +53,27 @@ jobs:
if: steps.my-cache.outputs.cache-hit != 'true'
shell: bash
run: |
make -j2 _kaldifeat
.github/scripts/install-kaldifeat.sh
- name: Inference with pre-trained model
- name: export-model
shell: bash
env:
HF_TOKEN: ${{ secrets.HF_TOKEN }}
run: |
sudo apt-get -qq install git-lfs tree
export PYTHONPATH=$PWD:$PYTHONPATH
export PYTHONPATH=~/tmp/kaldifeat/kaldifeat/python:$PYTHONPATH
export PYTHONPATH=~/tmp/kaldifeat/build/lib:$PYTHONPATH
.github/scripts/run-pre-trained-transducer.sh
.github/scripts/multi-zh-hans.sh
ls -lh
- name: upload model to https://github.com/k2-fsa/sherpa-onnx
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

123
.github/workflows/run-aishell-2022-06-20.yml vendored
View File

@ -1,123 +0,0 @@
# Copyright 2022 Fangjun Kuang (csukuangfj@gmail.com)
# See ../../LICENSE for clarification regarding multiple authors
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
name: run-aishell-2022-06-20
# pruned RNN-T + reworked model with random combiner
# https://huggingface.co/csukuangfj/icefall-aishell-pruned-transducer-stateless3-2022-06-20
on:
push:
branches:
- master
pull_request:
types: [labeled]
schedule:
# minute (0-59)
# hour (0-23)
# day of the month (1-31)
# month (1-12)
# day of the week (0-6)
# nightly build at 15:50 UTC time every day
- cron: "50 15 * * *"
concurrency:
group: run_aishell_2022_06_20-${{ github.ref }}
cancel-in-progress: true
jobs:
run_aishell_2022_06_20:
if: github.event.label.name == 'ready' || github.event.label.name == 'run-decode' || github.event_name == 'push' || github.event_name == 'schedule'
runs-on: ${{ matrix.os }}
strategy:
matrix:
os: [ubuntu-latest]
python-version: [3.8]
fail-fast: false
steps:
- uses: actions/checkout@v2
with:
fetch-depth: 0
- name: Setup Python ${{ matrix.python-version }}
uses: actions/setup-python@v2
with:
python-version: ${{ matrix.python-version }}
cache: 'pip'
cache-dependency-path: '**/requirements-ci.txt'
- name: Install Python dependencies
run: |
grep -v '^#' ./requirements-ci.txt | xargs -n 1 -L 1 pip install
pip uninstall -y protobuf
pip install --no-binary protobuf protobuf==3.20.*
- name: Cache kaldifeat
id: my-cache
uses: actions/cache@v2
with:
path: |
~/tmp/kaldifeat
key: cache-tmp-${{ matrix.python-version }}-2023-05-22
- name: Install kaldifeat
if: steps.my-cache.outputs.cache-hit != 'true'
shell: bash
run: |
.github/scripts/install-kaldifeat.sh
- name: Inference with pre-trained model
shell: bash
env:
GITHUB_EVENT_NAME: ${{ github.event_name }}
GITHUB_EVENT_LABEL_NAME: ${{ github.event.label.name }}
run: |
sudo apt-get -qq install git-lfs tree
export PYTHONPATH=$PWD:$PYTHONPATH
export PYTHONPATH=~/tmp/kaldifeat/kaldifeat/python:$PYTHONPATH
export PYTHONPATH=~/tmp/kaldifeat/build/lib:$PYTHONPATH
.github/scripts/run-aishell-pruned-transducer-stateless3-2022-06-20.sh
- name: Display decoding results for aishell pruned_transducer_stateless3
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
shell: bash
run: |
cd egs/aishell/ASR/
tree ./pruned_transducer_stateless3/exp
cd pruned_transducer_stateless3
echo "results for pruned_transducer_stateless3"
echo "===greedy search==="
find exp/greedy_search -name "log-*" -exec grep -n --color "best for test" {} + | sort -n -k2
find exp/greedy_search -name "log-*" -exec grep -n --color "best for dev" {} + | sort -n -k2
echo "===fast_beam_search==="
find exp/fast_beam_search -name "log-*" -exec grep -n --color "best for test" {} + | sort -n -k2
find exp/fast_beam_search -name "log-*" -exec grep -n --color "best for dev" {} + | sort -n -k2
echo "===modified beam search==="
find exp/modified_beam_search -name "log-*" -exec grep -n --color "best for test" {} + | sort -n -k2
find exp/modified_beam_search -name "log-*" -exec grep -n --color "best for dev" {} + | sort -n -k2
- name: Upload decoding results for aishell pruned_transducer_stateless3
uses: actions/upload-artifact@v2
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
with:
name: aishell-torch-${{ matrix.torch }}-python-${{ matrix.python-version }}-ubuntu-latest-cpu-pruned_transducer_stateless3-2022-06-20
path: egs/aishell/ASR/pruned_transducer_stateless3/exp/

95
.github/workflows/run-aishell-zipformer-2023-10-24.yml vendored
View File

@ -1,95 +0,0 @@
# Copyright 2023 Zengrui Jin (Xiaomi Corp.)
# See ../../LICENSE for clarification regarding multiple authors
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
name: run-aishell-zipformer-2023-10-24
on:
push:
branches:
- master
pull_request:
types: [labeled]
schedule:
# minute (0-59)
# hour (0-23)
# day of the month (1-31)
# month (1-12)
# day of the week (0-6)
# nightly build at 15:50 UTC time every day
- cron: "50 15 * * *"
concurrency:
group: run_aishell_zipformer_2023_10_24-${{ github.ref }}
cancel-in-progress: true
jobs:
run_aishell_zipformer_2023_10_24:
if: github.event.label.name == 'ready' || github.event.label.name == 'zipformer' || github.event.label.name == 'run-decode' || github.event_name == 'push' || github.event_name == 'schedule'
runs-on: ${{ matrix.os }}
strategy:
matrix:
os: [ubuntu-latest]
python-version: [3.8]
fail-fast: false
steps:
- uses: actions/checkout@v2
with:
fetch-depth: 0
- name: Setup Python ${{ matrix.python-version }}
uses: actions/setup-python@v2
with:
python-version: ${{ matrix.python-version }}
cache: 'pip'
cache-dependency-path: '**/requirements-ci.txt'
- name: Install Python dependencies
run: |
grep -v '^#' ./requirements-ci.txt | xargs -n 1 -L 1 pip install
pip uninstall -y protobuf
pip install --no-binary protobuf protobuf==3.20.*
- name: Cache kaldifeat
id: my-cache
uses: actions/cache@v2
with:
path: |
~/tmp/kaldifeat
key: cache-tmp-${{ matrix.python-version }}-2023-05-22
- name: Install kaldifeat
if: steps.my-cache.outputs.cache-hit != 'true'
shell: bash
run: |
.github/scripts/install-kaldifeat.sh
- name: Inference with pre-trained model
shell: bash
env:
GITHUB_EVENT_NAME: ${{ github.event_name }}
GITHUB_EVENT_LABEL_NAME: ${{ github.event.label.name }}
run: |
sudo apt-get -qq install git-lfs tree
export PYTHONPATH=$PWD:$PYTHONPATH
export PYTHONPATH=~/tmp/kaldifeat/kaldifeat/python:$PYTHONPATH
export PYTHONPATH=~/tmp/kaldifeat/build/lib:$PYTHONPATH
.github/scripts/run-aishell-zipformer-2023-10-24.sh

9
.github/workflows/run-docker-image.yml vendored
View File

@ -14,13 +14,20 @@ jobs:
fail-fast: false
matrix:
os: [ubuntu-latest]
image: ["torch2.1.0-cuda12.1", "torch2.1.0-cuda11.8", "torch2.0.0-cuda11.7", "torch1.13.0-cuda11.6", "torch1.12.1-cuda11.3", "torch1.9.0-cuda10.2"]
image: ["torch2.2.2-cuda12.1", "torch2.2.2-cuda11.8", "torch2.2.1-cuda12.1", "torch2.2.1-cuda11.8", "torch2.2.0-cuda12.1", "torch2.2.0-cuda11.8", "torch2.1.0-cuda12.1", "torch2.1.0-cuda11.8", "torch2.0.0-cuda11.7", "torch1.13.0-cuda11.6", "torch1.12.1-cuda11.3", "torch1.9.0-cuda10.2"]
steps:
# refer to https://github.com/actions/checkout
- uses: actions/checkout@v2
with:
fetch-depth: 0
- name: Free space
shell: bash
run: |
df -h
rm -rf /opt/hostedtoolcache
df -h
- name: Run the build process with Docker
uses: addnab/docker-run-action@v3
with:

14
.github/workflows/run-gigaspeech-zipformer-2023-10-17.yml vendored
View File

@ -21,6 +21,7 @@ on:
push:
branches:
- master
pull_request:
types: [labeled]
@ -33,6 +34,8 @@ on:
# nightly build at 15:50 UTC time every day
- cron: "50 15 * * *"
workflow_dispatch:
concurrency:
group: run_gigaspeech_2023_10_17_zipformer-${{ github.ref }}
cancel-in-progress: true
@ -85,6 +88,7 @@ jobs:
env:
GITHUB_EVENT_NAME: ${{ github.event_name }}
GITHUB_EVENT_LABEL_NAME: ${{ github.event.label.name }}
HF_TOKEN: ${{ secrets.HF_TOKEN }}
run: |
mkdir -p egs/gigaspeech/ASR/data
ln -sfv ~/tmp/fbank-libri egs/gigaspeech/ASR/data/fbank
@ -97,6 +101,16 @@ jobs:
.github/scripts/run-gigaspeech-zipformer-2023-10-17.sh
- name: upload model to https://github.com/k2-fsa/sherpa-onnx
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
- name: Display decoding results for gigaspeech zipformer
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
shell: bash

159
.github/workflows/run-librispeech-2022-03-12.yml vendored
View File

@ -1,159 +0,0 @@
# Copyright 2021 Fangjun Kuang (csukuangfj@gmail.com)
# See ../../LICENSE for clarification regarding multiple authors
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
name: run-librispeech-2022-03-12
# stateless transducer + k2 pruned rnnt-loss
on:
push:
branches:
- master
pull_request:
types: [labeled]
schedule:
# minute (0-59)
# hour (0-23)
# day of the month (1-31)
# month (1-12)
# day of the week (0-6)
# nightly build at 15:50 UTC time every day
- cron: "50 15 * * *"
concurrency:
group: run_librispeech_2022_03_12-${{ github.ref }}
cancel-in-progress: true
jobs:
run_librispeech_2022_03_12:
if: github.event.label.name == 'ready' || github.event.label.name == 'run-decode' || github.event_name == 'push' || github.event_name == 'schedule'
runs-on: ${{ matrix.os }}
strategy:
matrix:
os: [ubuntu-latest]
python-version: [3.8]
fail-fast: false
steps:
- uses: actions/checkout@v2
with:
fetch-depth: 0
- name: Setup Python ${{ matrix.python-version }}
uses: actions/setup-python@v2
with:
python-version: ${{ matrix.python-version }}
cache: 'pip'
cache-dependency-path: '**/requirements-ci.txt'
- name: Install Python dependencies
run: |
grep -v '^#' ./requirements-ci.txt | xargs -n 1 -L 1 pip install
pip uninstall -y protobuf
pip install --no-binary protobuf protobuf==3.20.*
- name: Cache kaldifeat
id: my-cache
uses: actions/cache@v2
with:
path: |
~/tmp/kaldifeat
key: cache-tmp-${{ matrix.python-version }}-2023-05-22
- name: Install kaldifeat
if: steps.my-cache.outputs.cache-hit != 'true'
shell: bash
run: |
.github/scripts/install-kaldifeat.sh
- name: Cache LibriSpeech test-clean and test-other datasets
id: libri-test-clean-and-test-other-data
uses: actions/cache@v2
with:
path: |
~/tmp/download
key: cache-libri-test-clean-and-test-other
- name: Download LibriSpeech test-clean and test-other
if: steps.libri-test-clean-and-test-other-data.outputs.cache-hit != 'true'
shell: bash
run: |
.github/scripts/download-librispeech-test-clean-and-test-other-dataset.sh
- name: Prepare manifests for LibriSpeech test-clean and test-other
shell: bash
run: |
.github/scripts/prepare-librispeech-test-clean-and-test-other-manifests.sh
- name: Cache LibriSpeech test-clean and test-other fbank features
id: libri-test-clean-and-test-other-fbank
uses: actions/cache@v2
with:
path: |
~/tmp/fbank-libri
key: cache-libri-fbank-test-clean-and-test-other-v2
- name: Compute fbank for LibriSpeech test-clean and test-other
if: steps.libri-test-clean-and-test-other-fbank.outputs.cache-hit != 'true'
shell: bash
run: |
.github/scripts/compute-fbank-librispeech-test-clean-and-test-other.sh
- name: Inference with pre-trained model
shell: bash
env:
GITHUB_EVENT_NAME: ${{ github.event_name }}
GITHUB_EVENT_LABEL_NAME: ${{ github.event.label.name }}
run: |
mkdir -p egs/librispeech/ASR/data
ln -sfv ~/tmp/fbank-libri egs/librispeech/ASR/data/fbank
ls -lh egs/librispeech/ASR/data/*
sudo apt-get -qq install git-lfs tree
export PYTHONPATH=$PWD:$PYTHONPATH
export PYTHONPATH=~/tmp/kaldifeat/kaldifeat/python:$PYTHONPATH
export PYTHONPATH=~/tmp/kaldifeat/build/lib:$PYTHONPATH
.github/scripts/run-librispeech-pruned-transducer-stateless-2022-03-12.sh
- name: Display decoding results for pruned_transducer_stateless
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
shell: bash
run: |
cd egs/librispeech/ASR/
tree ./pruned_transducer_stateless/exp
cd pruned_transducer_stateless
echo "results for pruned_transducer_stateless"
echo "===greedy search==="
find exp/greedy_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/greedy_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
echo "===fast_beam_search==="
find exp/fast_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/fast_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
echo "===modified beam search==="
find exp/modified_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/modified_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
- name: Upload decoding results for pruned_transducer_stateless
uses: actions/upload-artifact@v2
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
with:
name: torch-${{ matrix.torch }}-python-${{ matrix.python-version }}-ubuntu-latest-cpu-pruned_transducer_stateless-2022-03-12
path: egs/librispeech/ASR/pruned_transducer_stateless/exp/

185
.github/workflows/run-librispeech-2022-04-29.yml vendored
View File

@ -1,185 +0,0 @@
# Copyright 2021 Fangjun Kuang (csukuangfj@gmail.com)
# See ../../LICENSE for clarification regarding multiple authors
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
name: run-librispeech-2022-04-29
# stateless pruned transducer (reworked model) + giga speech
on:
push:
branches:
- master
pull_request:
types: [labeled]
schedule:
# minute (0-59)
# hour (0-23)
# day of the month (1-31)
# month (1-12)
# day of the week (0-6)
# nightly build at 15:50 UTC time every day
- cron: "50 15 * * *"
concurrency:
group: run_librispeech_2022_04_29-${{ github.ref }}
cancel-in-progress: true
jobs:
run_librispeech_2022_04_29:
if: github.event.label.name == 'ready' || github.event.label.name == 'run-decode' || github.event_name == 'push' || github.event_name == 'schedule'
runs-on: ${{ matrix.os }}
strategy:
matrix:
os: [ubuntu-latest]
python-version: [3.8]
fail-fast: false
steps:
- uses: actions/checkout@v2
with:
fetch-depth: 0
- name: Setup Python ${{ matrix.python-version }}
uses: actions/setup-python@v2
with:
python-version: ${{ matrix.python-version }}
cache: 'pip'
cache-dependency-path: '**/requirements-ci.txt'
- name: Install Python dependencies
run: |
grep -v '^#' ./requirements-ci.txt | xargs -n 1 -L 1 pip install
pip uninstall -y protobuf
pip install --no-binary protobuf protobuf==3.20.*
- name: Cache kaldifeat
id: my-cache
uses: actions/cache@v2
with:
path: |
~/tmp/kaldifeat
key: cache-tmp-${{ matrix.python-version }}-2023-05-22
- name: Install kaldifeat
if: steps.my-cache.outputs.cache-hit != 'true'
shell: bash
run: |
.github/scripts/install-kaldifeat.sh
- name: Cache LibriSpeech test-clean and test-other datasets
id: libri-test-clean-and-test-other-data
uses: actions/cache@v2
with:
path: |
~/tmp/download
key: cache-libri-test-clean-and-test-other
- name: Download LibriSpeech test-clean and test-other
if: steps.libri-test-clean-and-test-other-data.outputs.cache-hit != 'true'
shell: bash
run: |
.github/scripts/download-librispeech-test-clean-and-test-other-dataset.sh
- name: Prepare manifests for LibriSpeech test-clean and test-other
shell: bash
run: |
.github/scripts/prepare-librispeech-test-clean-and-test-other-manifests.sh
- name: Cache LibriSpeech test-clean and test-other fbank features
id: libri-test-clean-and-test-other-fbank
uses: actions/cache@v2
with:
path: |
~/tmp/fbank-libri
key: cache-libri-fbank-test-clean-and-test-other-v2
- name: Compute fbank for LibriSpeech test-clean and test-other
if: steps.libri-test-clean-and-test-other-fbank.outputs.cache-hit != 'true'
shell: bash
run: |
.github/scripts/compute-fbank-librispeech-test-clean-and-test-other.sh
- name: Inference with pre-trained model
shell: bash
env:
GITHUB_EVENT_NAME: ${{ github.event_name }}
GITHUB_EVENT_LABEL_NAME: ${{ github.event.label.name }}
run: |
mkdir -p egs/librispeech/ASR/data
ln -sfv ~/tmp/fbank-libri egs/librispeech/ASR/data/fbank
ls -lh egs/librispeech/ASR/data/*
sudo apt-get -qq install git-lfs tree
export PYTHONPATH=$PWD:$PYTHONPATH
export PYTHONPATH=~/tmp/kaldifeat/kaldifeat/python:$PYTHONPATH
export PYTHONPATH=~/tmp/kaldifeat/build/lib:$PYTHONPATH
.github/scripts/run-librispeech-pruned-transducer-stateless2-2022-04-29.sh
.github/scripts/run-librispeech-pruned-transducer-stateless3-2022-04-29.sh
- name: Display decoding results for pruned_transducer_stateless2
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
shell: bash
run: |
cd egs/librispeech/ASR
tree pruned_transducer_stateless2/exp
cd pruned_transducer_stateless2/exp
echo "===greedy search==="
find greedy_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find greedy_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
echo "===fast_beam_search==="
find fast_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find fast_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
echo "===modified beam search==="
find modified_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find modified_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
- name: Display decoding results for pruned_transducer_stateless3
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
shell: bash
run: |
cd egs/librispeech/ASR
tree pruned_transducer_stateless3/exp
cd pruned_transducer_stateless3/exp
echo "===greedy search==="
find greedy_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find greedy_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
echo "===fast_beam_search==="
find fast_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find fast_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
echo "===modified beam search==="
find modified_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find modified_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
- name: Upload decoding results for pruned_transducer_stateless2
uses: actions/upload-artifact@v2
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
with:
name: torch-${{ matrix.torch }}-python-${{ matrix.python-version }}-ubuntu-latest-cpu-pruned_transducer_stateless2-2022-04-29
path: egs/librispeech/ASR/pruned_transducer_stateless2/exp/
- name: Upload decoding results for pruned_transducer_stateless3
uses: actions/upload-artifact@v2
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
with:
name: torch-${{ matrix.torch }}-python-${{ matrix.python-version }}-ubuntu-latest-cpu-pruned_transducer_stateless3-2022-04-29
path: egs/librispeech/ASR/pruned_transducer_stateless3/exp/

159
.github/workflows/run-librispeech-2022-05-13.yml vendored
View File

@ -1,159 +0,0 @@
# Copyright 2022 Fangjun Kuang (csukuangfj@gmail.com)
# See ../../LICENSE for clarification regarding multiple authors
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
name: run-librispeech-2022-05-13
# stateless transducer + k2 pruned rnnt-loss + deeper model
on:
push:
branches:
- master
pull_request:
types: [labeled]
schedule:
# minute (0-59)
# hour (0-23)
# day of the month (1-31)
# month (1-12)
# day of the week (0-6)
# nightly build at 15:50 UTC time every day
- cron: "50 15 * * *"
concurrency:
group: run_librispeech_2022_05_13-${{ github.ref }}
cancel-in-progress: true
jobs:
run_librispeech_2022_05_13:
if: github.event.label.name == 'ready' || github.event.label.name == 'run-decode' || github.event_name == 'push' || github.event_name == 'schedule'
runs-on: ${{ matrix.os }}
strategy:
matrix:
os: [ubuntu-latest]
python-version: [3.8]
fail-fast: false
steps:
- uses: actions/checkout@v2
with:
fetch-depth: 0
- name: Setup Python ${{ matrix.python-version }}
uses: actions/setup-python@v2
with:
python-version: ${{ matrix.python-version }}
cache: 'pip'
cache-dependency-path: '**/requirements-ci.txt'
- name: Install Python dependencies
run: |
grep -v '^#' ./requirements-ci.txt | xargs -n 1 -L 1 pip install
pip uninstall -y protobuf
pip install --no-binary protobuf protobuf==3.20.*
- name: Cache kaldifeat
id: my-cache
uses: actions/cache@v2
with:
path: |
~/tmp/kaldifeat
key: cache-tmp-${{ matrix.python-version }}-2023-05-22
- name: Install kaldifeat
if: steps.my-cache.outputs.cache-hit != 'true'
shell: bash
run: |
.github/scripts/install-kaldifeat.sh
- name: Cache LibriSpeech test-clean and test-other datasets
id: libri-test-clean-and-test-other-data
uses: actions/cache@v2
with:
path: |
~/tmp/download
key: cache-libri-test-clean-and-test-other
- name: Download LibriSpeech test-clean and test-other
if: steps.libri-test-clean-and-test-other-data.outputs.cache-hit != 'true'
shell: bash
run: |
.github/scripts/download-librispeech-test-clean-and-test-other-dataset.sh
- name: Prepare manifests for LibriSpeech test-clean and test-other
shell: bash
run: |
.github/scripts/prepare-librispeech-test-clean-and-test-other-manifests.sh
- name: Cache LibriSpeech test-clean and test-other fbank features
id: libri-test-clean-and-test-other-fbank
uses: actions/cache@v2
with:
path: |
~/tmp/fbank-libri
key: cache-libri-fbank-test-clean-and-test-other-v2
- name: Compute fbank for LibriSpeech test-clean and test-other
if: steps.libri-test-clean-and-test-other-fbank.outputs.cache-hit != 'true'
shell: bash
run: |
.github/scripts/compute-fbank-librispeech-test-clean-and-test-other.sh
- name: Inference with pre-trained model
shell: bash
env:
GITHUB_EVENT_NAME: ${{ github.event_name }}
GITHUB_EVENT_LABEL_NAME: ${{ github.event.label.name }}
run: |
mkdir -p egs/librispeech/ASR/data
ln -sfv ~/tmp/fbank-libri egs/librispeech/ASR/data/fbank
ls -lh egs/librispeech/ASR/data/*
sudo apt-get -qq install git-lfs tree
export PYTHONPATH=$PWD:$PYTHONPATH
export PYTHONPATH=~/tmp/kaldifeat/kaldifeat/python:$PYTHONPATH
export PYTHONPATH=~/tmp/kaldifeat/build/lib:$PYTHONPATH
.github/scripts/run-librispeech-pruned-transducer-stateless5-2022-05-13.sh
- name: Display decoding results for librispeech pruned_transducer_stateless5
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
shell: bash
run: |
cd egs/librispeech/ASR/
tree ./pruned_transducer_stateless5/exp
cd pruned_transducer_stateless5
echo "results for pruned_transducer_stateless5"
echo "===greedy search==="
find exp/greedy_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/greedy_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
echo "===fast_beam_search==="
find exp/fast_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/fast_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
echo "===modified beam search==="
find exp/modified_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/modified_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
- name: Upload decoding results for librispeech pruned_transducer_stateless5
uses: actions/upload-artifact@v2
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
with:
name: torch-${{ matrix.torch }}-python-${{ matrix.python-version }}-ubuntu-latest-cpu-pruned_transducer_stateless5-2022-05-13
path: egs/librispeech/ASR/pruned_transducer_stateless5/exp/

159
.github/workflows/run-librispeech-2022-11-11-stateless7.yml vendored
View File

@ -1,159 +0,0 @@
# Copyright 2022 Fangjun Kuang (csukuangfj@gmail.com)
# See ../../LICENSE for clarification regarding multiple authors
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
name: run-librispeech-2022-11-11-stateless7
# zipformer
on:
push:
branches:
- master
pull_request:
types: [labeled]
schedule:
# minute (0-59)
# hour (0-23)
# day of the month (1-31)
# month (1-12)
# day of the week (0-6)
# nightly build at 15:50 UTC time every day
- cron: "50 15 * * *"
concurrency:
group: run_librispeech_2022_11_11_zipformer-${{ github.ref }}
cancel-in-progress: true
jobs:
run_librispeech_2022_11_11_zipformer:
if: github.event.label.name == 'ready' || github.event.label.name == 'run-decode' || github.event_name == 'push' || github.event_name == 'schedule'
runs-on: ${{ matrix.os }}
strategy:
matrix:
os: [ubuntu-latest]
python-version: [3.8]
fail-fast: false
steps:
- uses: actions/checkout@v2
with:
fetch-depth: 0
- name: Setup Python ${{ matrix.python-version }}
uses: actions/setup-python@v2
with:
python-version: ${{ matrix.python-version }}
cache: 'pip'
cache-dependency-path: '**/requirements-ci.txt'
- name: Install Python dependencies
run: |
grep -v '^#' ./requirements-ci.txt | xargs -n 1 -L 1 pip install
pip uninstall -y protobuf
pip install --no-binary protobuf protobuf==3.20.*
- name: Cache kaldifeat
id: my-cache
uses: actions/cache@v2
with:
path: |
~/tmp/kaldifeat
key: cache-tmp-${{ matrix.python-version }}-2023-05-22
- name: Install kaldifeat
if: steps.my-cache.outputs.cache-hit != 'true'
shell: bash
run: |
.github/scripts/install-kaldifeat.sh
- name: Cache LibriSpeech test-clean and test-other datasets
id: libri-test-clean-and-test-other-data
uses: actions/cache@v2
with:
path: |
~/tmp/download
key: cache-libri-test-clean-and-test-other
- name: Download LibriSpeech test-clean and test-other
if: steps.libri-test-clean-and-test-other-data.outputs.cache-hit != 'true'
shell: bash
run: |
.github/scripts/download-librispeech-test-clean-and-test-other-dataset.sh
- name: Prepare manifests for LibriSpeech test-clean and test-other
shell: bash
run: |
.github/scripts/prepare-librispeech-test-clean-and-test-other-manifests.sh
- name: Cache LibriSpeech test-clean and test-other fbank features
id: libri-test-clean-and-test-other-fbank
uses: actions/cache@v2
with:
path: |
~/tmp/fbank-libri
key: cache-libri-fbank-test-clean-and-test-other-v2
- name: Compute fbank for LibriSpeech test-clean and test-other
if: steps.libri-test-clean-and-test-other-fbank.outputs.cache-hit != 'true'
shell: bash
run: |
.github/scripts/compute-fbank-librispeech-test-clean-and-test-other.sh
- name: Inference with pre-trained model
shell: bash
env:
GITHUB_EVENT_NAME: ${{ github.event_name }}
GITHUB_EVENT_LABEL_NAME: ${{ github.event.label.name }}
run: |
mkdir -p egs/librispeech/ASR/data
ln -sfv ~/tmp/fbank-libri egs/librispeech/ASR/data/fbank
ls -lh egs/librispeech/ASR/data/*
sudo apt-get -qq install git-lfs tree
export PYTHONPATH=$PWD:$PYTHONPATH
export PYTHONPATH=~/tmp/kaldifeat/kaldifeat/python:$PYTHONPATH
export PYTHONPATH=~/tmp/kaldifeat/build/lib:$PYTHONPATH
.github/scripts/run-librispeech-pruned-transducer-stateless7-2022-11-11.sh
- name: Display decoding results for librispeech pruned_transducer_stateless7
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
shell: bash
run: |
cd egs/librispeech/ASR/
tree ./pruned_transducer_stateless7/exp
cd pruned_transducer_stateless7
echo "results for pruned_transducer_stateless7"
echo "===greedy search==="
find exp/greedy_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/greedy_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
echo "===fast_beam_search==="
find exp/fast_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/fast_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
echo "===modified beam search==="
find exp/modified_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/modified_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
- name: Upload decoding results for librispeech pruned_transducer_stateless7
uses: actions/upload-artifact@v2
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
with:
name: torch-${{ matrix.torch }}-python-${{ matrix.python-version }}-ubuntu-latest-cpu-pruned_transducer_stateless7-2022-11-11
path: egs/librispeech/ASR/pruned_transducer_stateless7/exp/

159
.github/workflows/run-librispeech-2022-11-14-stateless8.yml vendored
View File

@ -1,159 +0,0 @@
# Copyright 2022 Fangjun Kuang (csukuangfj@gmail.com)
# See ../../LICENSE for clarification regarding multiple authors
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
name: run-librispeech-2022-11-14-stateless8
# zipformer
on:
push:
branches:
- master
pull_request:
types: [labeled]
schedule:
# minute (0-59)
# hour (0-23)
# day of the month (1-31)
# month (1-12)
# day of the week (0-6)
# nightly build at 15:50 UTC time every day
- cron: "50 15 * * *"
concurrency:
group: run_librispeech_2022_11_14_zipformer_stateless8-${{ github.ref }}
cancel-in-progress: true
jobs:
run_librispeech_2022_11_14_zipformer_stateless8:
if: github.event.label.name == 'ready' || github.event.label.name == 'run-decode' || github.event_name == 'push' || github.event_name == 'schedule'
runs-on: ${{ matrix.os }}
strategy:
matrix:
os: [ubuntu-latest]
python-version: [3.8]
fail-fast: false
steps:
- uses: actions/checkout@v2
with:
fetch-depth: 0
- name: Setup Python ${{ matrix.python-version }}
uses: actions/setup-python@v2
with:
python-version: ${{ matrix.python-version }}
cache: 'pip'
cache-dependency-path: '**/requirements-ci.txt'
- name: Install Python dependencies
run: |
grep -v '^#' ./requirements-ci.txt | xargs -n 1 -L 1 pip install
pip uninstall -y protobuf
pip install --no-binary protobuf protobuf==3.20.*
- name: Cache kaldifeat
id: my-cache
uses: actions/cache@v2
with:
path: |
~/tmp/kaldifeat
key: cache-tmp-${{ matrix.python-version }}-2023-05-22
- name: Install kaldifeat
if: steps.my-cache.outputs.cache-hit != 'true'
shell: bash
run: |
.github/scripts/install-kaldifeat.sh
- name: Cache LibriSpeech test-clean and test-other datasets
id: libri-test-clean-and-test-other-data
uses: actions/cache@v2
with:
path: |
~/tmp/download
key: cache-libri-test-clean-and-test-other
- name: Download LibriSpeech test-clean and test-other
if: steps.libri-test-clean-and-test-other-data.outputs.cache-hit != 'true'
shell: bash
run: |
.github/scripts/download-librispeech-test-clean-and-test-other-dataset.sh
- name: Prepare manifests for LibriSpeech test-clean and test-other
shell: bash
run: |
.github/scripts/prepare-librispeech-test-clean-and-test-other-manifests.sh
- name: Cache LibriSpeech test-clean and test-other fbank features
id: libri-test-clean-and-test-other-fbank
uses: actions/cache@v2
with:
path: |
~/tmp/fbank-libri
key: cache-libri-fbank-test-clean-and-test-other-v2
- name: Compute fbank for LibriSpeech test-clean and test-other
if: steps.libri-test-clean-and-test-other-fbank.outputs.cache-hit != 'true'
shell: bash
run: |
.github/scripts/compute-fbank-librispeech-test-clean-and-test-other.sh
- name: Inference with pre-trained model
shell: bash
env:
GITHUB_EVENT_NAME: ${{ github.event_name }}
GITHUB_EVENT_LABEL_NAME: ${{ github.event.label.name }}
run: |
mkdir -p egs/librispeech/ASR/data
ln -sfv ~/tmp/fbank-libri egs/librispeech/ASR/data/fbank
ls -lh egs/librispeech/ASR/data/*
sudo apt-get -qq install git-lfs tree
export PYTHONPATH=$PWD:$PYTHONPATH
export PYTHONPATH=~/tmp/kaldifeat/kaldifeat/python:$PYTHONPATH
export PYTHONPATH=~/tmp/kaldifeat/build/lib:$PYTHONPATH
.github/scripts/run-librispeech-pruned-transducer-stateless8-2022-11-14.sh
- name: Display decoding results for librispeech pruned_transducer_stateless8
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
shell: bash
run: |
cd egs/librispeech/ASR/
tree ./pruned_transducer_stateless8/exp
cd pruned_transducer_stateless8
echo "results for pruned_transducer_stateless8"
echo "===greedy search==="
find exp/greedy_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/greedy_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
echo "===fast_beam_search==="
find exp/fast_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/fast_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
echo "===modified beam search==="
find exp/modified_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/modified_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
- name: Upload decoding results for librispeech pruned_transducer_stateless8
uses: actions/upload-artifact@v2
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
with:
name: torch-${{ matrix.torch }}-python-${{ matrix.python-version }}-ubuntu-latest-cpu-pruned_transducer_stateless8-2022-11-14
path: egs/librispeech/ASR/pruned_transducer_stateless8/exp/

163
.github/workflows/run-librispeech-2022-12-01-stateless7-ctc.yml vendored
View File

@ -1,163 +0,0 @@
# Copyright 2022 Fangjun Kuang (csukuangfj@gmail.com)
# See ../../LICENSE for clarification regarding multiple authors
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
name: run-librispeech-2022-12-01-stateless7-ctc
# zipformer
on:
push:
branches:
- master
pull_request:
types: [labeled]
schedule:
# minute (0-59)
# hour (0-23)
# day of the month (1-31)
# month (1-12)
# day of the week (0-6)
# nightly build at 15:50 UTC time every day
- cron: "50 15 * * *"
jobs:
run_librispeech_2022_11_11_zipformer:
if: github.event.label.name == 'ready' || github.event.label.name == 'run-decode' || github.event_name == 'push' || github.event_name == 'schedule'
runs-on: ${{ matrix.os }}
strategy:
matrix:
os: [ubuntu-latest]
python-version: [3.8]
fail-fast: false
steps:
- uses: actions/checkout@v2
with:
fetch-depth: 0
- name: Setup Python ${{ matrix.python-version }}
uses: actions/setup-python@v2
with:
python-version: ${{ matrix.python-version }}
cache: 'pip'
cache-dependency-path: '**/requirements-ci.txt'
- name: Install Python dependencies
run: |
grep -v '^#' ./requirements-ci.txt | xargs -n 1 -L 1 pip install
pip uninstall -y protobuf
pip install --no-binary protobuf protobuf==3.20.*
- name: Cache kaldifeat
id: my-cache
uses: actions/cache@v2
with:
path: |
~/tmp/kaldifeat
key: cache-tmp-${{ matrix.python-version }}-2023-05-22
- name: Install kaldifeat
if: steps.my-cache.outputs.cache-hit != 'true'
shell: bash
run: |
.github/scripts/install-kaldifeat.sh
- name: Cache LibriSpeech test-clean and test-other datasets
id: libri-test-clean-and-test-other-data
uses: actions/cache@v2
with:
path: |
~/tmp/download
key: cache-libri-test-clean-and-test-other
- name: Download LibriSpeech test-clean and test-other
if: steps.libri-test-clean-and-test-other-data.outputs.cache-hit != 'true'
shell: bash
run: |
.github/scripts/download-librispeech-test-clean-and-test-other-dataset.sh
- name: Prepare manifests for LibriSpeech test-clean and test-other
shell: bash
run: |
.github/scripts/prepare-librispeech-test-clean-and-test-other-manifests.sh
- name: Cache LibriSpeech test-clean and test-other fbank features
id: libri-test-clean-and-test-other-fbank
uses: actions/cache@v2
with:
path: |
~/tmp/fbank-libri
key: cache-libri-fbank-test-clean-and-test-other-v2
- name: Compute fbank for LibriSpeech test-clean and test-other
if: steps.libri-test-clean-and-test-other-fbank.outputs.cache-hit != 'true'
shell: bash
run: |
.github/scripts/compute-fbank-librispeech-test-clean-and-test-other.sh
- name: Inference with pre-trained model
shell: bash
env:
GITHUB_EVENT_NAME: ${{ github.event_name }}
GITHUB_EVENT_LABEL_NAME: ${{ github.event.label.name }}
run: |
mkdir -p egs/librispeech/ASR/data
ln -sfv ~/tmp/fbank-libri egs/librispeech/ASR/data/fbank
ls -lh egs/librispeech/ASR/data/*
sudo apt-get -qq install git-lfs tree
export PYTHONPATH=$PWD:$PYTHONPATH
export PYTHONPATH=~/tmp/kaldifeat/kaldifeat/python:$PYTHONPATH
export PYTHONPATH=~/tmp/kaldifeat/build/lib:$PYTHONPATH
.github/scripts/run-librispeech-pruned-transducer-stateless7-ctc-2022-12-01.sh
- name: Display decoding results for librispeech pruned_transducer_stateless7_ctc
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
shell: bash
run: |
cd egs/librispeech/ASR/
tree ./pruned_transducer_stateless7_ctc/exp
cd pruned_transducer_stateless7_ctc
echo "results for pruned_transducer_stateless7_ctc"
echo "===greedy search==="
find exp/greedy_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/greedy_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
echo "===fast_beam_search==="
find exp/fast_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/fast_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
echo "===modified beam search==="
find exp/modified_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/modified_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
echo "===ctc decoding==="
find exp/ctc-decoding -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/ctc-decoding -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
echo "===1best==="
find exp/1best -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/1best -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
- name: Upload decoding results for librispeech pruned_transducer_stateless7_ctc
uses: actions/upload-artifact@v2
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
with:
name: torch-${{ matrix.torch }}-python-${{ matrix.python-version }}-ubuntu-latest-cpu-pruned_transducer_stateless7-ctc-2022-12-01
path: egs/librispeech/ASR/pruned_transducer_stateless7_ctc/exp/

167
.github/workflows/run-librispeech-2022-12-08-zipformer-mmi.yml vendored
View File

@ -1,167 +0,0 @@
# Copyright 2022 Zengwei Yao
# See ../../LICENSE for clarification regarding multiple authors
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
name: run-librispeech-2022-12-08-zipformer-mmi
# zipformer
on:
push:
branches:
- master
pull_request:
types: [labeled]
schedule:
# minute (0-59)
# hour (0-23)
# day of the month (1-31)
# month (1-12)
# day of the week (0-6)
# nightly build at 15:50 UTC time every day
- cron: "50 15 * * *"
concurrency:
group: run_librispeech_2022_12_08_zipformer-${{ github.ref }}
cancel-in-progress: true
jobs:
run_librispeech_2022_12_08_zipformer:
if: github.event.label.name == 'ready' || github.event.label.name == 'run-decode' || github.event_name == 'push' || github.event_name == 'schedule'
runs-on: ${{ matrix.os }}
strategy:
matrix:
os: [ubuntu-latest]
python-version: [3.8]
fail-fast: false
steps:
- uses: actions/checkout@v2
with:
fetch-depth: 0
- name: Setup Python ${{ matrix.python-version }}
uses: actions/setup-python@v2
with:
python-version: ${{ matrix.python-version }}
cache: 'pip'
cache-dependency-path: '**/requirements-ci.txt'
- name: Install Python dependencies
run: |
grep -v '^#' ./requirements-ci.txt | xargs -n 1 -L 1 pip install
pip uninstall -y protobuf
pip install --no-binary protobuf protobuf==3.20.*
- name: Cache kaldifeat
id: my-cache
uses: actions/cache@v2
with:
path: |
~/tmp/kaldifeat
key: cache-tmp-${{ matrix.python-version }}-2023-05-22
- name: Install kaldifeat
if: steps.my-cache.outputs.cache-hit != 'true'
shell: bash
run: |
.github/scripts/install-kaldifeat.sh
- name: Cache LibriSpeech test-clean and test-other datasets
id: libri-test-clean-and-test-other-data
uses: actions/cache@v2
with:
path: |
~/tmp/download
key: cache-libri-test-clean-and-test-other
- name: Download LibriSpeech test-clean and test-other
if: steps.libri-test-clean-and-test-other-data.outputs.cache-hit != 'true'
shell: bash
run: |
.github/scripts/download-librispeech-test-clean-and-test-other-dataset.sh
- name: Prepare manifests for LibriSpeech test-clean and test-other
shell: bash
run: |
.github/scripts/prepare-librispeech-test-clean-and-test-other-manifests.sh
- name: Cache LibriSpeech test-clean and test-other fbank features
id: libri-test-clean-and-test-other-fbank
uses: actions/cache@v2
with:
path: |
~/tmp/fbank-libri
key: cache-libri-fbank-test-clean-and-test-other-v2
- name: Compute fbank for LibriSpeech test-clean and test-other
if: steps.libri-test-clean-and-test-other-fbank.outputs.cache-hit != 'true'
shell: bash
run: |
.github/scripts/compute-fbank-librispeech-test-clean-and-test-other.sh
- name: Inference with pre-trained model
shell: bash
env:
GITHUB_EVENT_NAME: ${{ github.event_name }}
GITHUB_EVENT_LABEL_NAME: ${{ github.event.label.name }}
run: |
mkdir -p egs/librispeech/ASR/data
ln -sfv ~/tmp/fbank-libri egs/librispeech/ASR/data/fbank
ls -lh egs/librispeech/ASR/data/*
sudo apt-get -qq install git-lfs tree
export PYTHONPATH=$PWD:$PYTHONPATH
export PYTHONPATH=~/tmp/kaldifeat/kaldifeat/python:$PYTHONPATH
export PYTHONPATH=~/tmp/kaldifeat/build/lib:$PYTHONPATH
.github/scripts/run-librispeech-zipformer-mmi-2022-12-08.sh
- name: Display decoding results for librispeech zipformer-mmi
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
shell: bash
run: |
cd egs/librispeech/ASR/
tree ./zipformer-mmi/exp
cd zipformer-mmi
echo "results for zipformer-mmi"
echo "===1best==="
find exp/1best -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/1best -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
echo "===nbest==="
find exp/nbest -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/nbest -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
echo "===nbest-rescoring-LG==="
find exp/nbest-rescoring-LG -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/nbest-rescoring-LG -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
echo "===nbest-rescoring-3-gram==="
find exp/nbest-rescoring-3-gram -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/nbest-rescoring-3-gram -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
echo "===nbest-rescoring-4-gram==="
find exp/nbest-rescoring-4-gram -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/nbest-rescoring-4-gram -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
- name: Upload decoding results for librispeech zipformer-mmi
uses: actions/upload-artifact@v2
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
with:
name: torch-${{ matrix.torch }}-python-${{ matrix.python-version }}-ubuntu-latest-cpu-zipformer_mmi-2022-12-08
path: egs/librispeech/ASR/zipformer_mmi/exp/

172
.github/workflows/run-librispeech-2022-12-29-stateless7-streaming.yml vendored
View File

@ -1,172 +0,0 @@
# Copyright 2022 Fangjun Kuang (csukuangfj@gmail.com)
# See ../../LICENSE for clarification regarding multiple authors
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
name: run-librispeech-2022-12-29-stateless7-streaming
# zipformer
on:
push:
branches:
- master
pull_request:
types: [labeled]
schedule:
# minute (0-59)
# hour (0-23)
# day of the month (1-31)
# month (1-12)
# day of the week (0-6)
# nightly build at 15:50 UTC time every day
- cron: "50 15 * * *"
concurrency:
group: run_librispeech_2022_12_29_zipformer_streaming-${{ github.ref }}
cancel-in-progress: true
jobs:
run_librispeech_2022_12_29_zipformer_streaming:
if: github.event.label.name == 'ready' || github.event.label.name == 'run-decode' || github.event.label.name == 'streaming-zipformer' || github.event_name == 'push' || github.event_name == 'schedule'
runs-on: ${{ matrix.os }}
strategy:
matrix:
os: [ubuntu-latest]
python-version: [3.8]
fail-fast: false
steps:
- uses: actions/checkout@v2
with:
fetch-depth: 0
- name: Setup Python ${{ matrix.python-version }}
uses: actions/setup-python@v2
with:
python-version: ${{ matrix.python-version }}
cache: 'pip'
cache-dependency-path: '**/requirements-ci.txt'
- name: Install Python dependencies
run: |
grep -v '^#' ./requirements-ci.txt | xargs -n 1 -L 1 pip install
pip uninstall -y protobuf
pip install --no-binary protobuf protobuf==3.20.*
- name: Cache kaldifeat
id: my-cache
uses: actions/cache@v2
with:
path: |
~/tmp/kaldifeat
key: cache-tmp-${{ matrix.python-version }}-2023-05-22
- name: Install kaldifeat
if: steps.my-cache.outputs.cache-hit != 'true'
shell: bash
run: |
.github/scripts/install-kaldifeat.sh
- name: Cache LibriSpeech test-clean and test-other datasets
id: libri-test-clean-and-test-other-data
uses: actions/cache@v2
with:
path: |
~/tmp/download
key: cache-libri-test-clean-and-test-other
- name: Download LibriSpeech test-clean and test-other
if: steps.libri-test-clean-and-test-other-data.outputs.cache-hit != 'true'
shell: bash
run: |
.github/scripts/download-librispeech-test-clean-and-test-other-dataset.sh
- name: Prepare manifests for LibriSpeech test-clean and test-other
shell: bash
run: |
.github/scripts/prepare-librispeech-test-clean-and-test-other-manifests.sh
- name: Cache LibriSpeech test-clean and test-other fbank features
id: libri-test-clean-and-test-other-fbank
uses: actions/cache@v2
with:
path: |
~/tmp/fbank-libri
key: cache-libri-fbank-test-clean-and-test-other-v2
- name: Compute fbank for LibriSpeech test-clean and test-other
if: steps.libri-test-clean-and-test-other-fbank.outputs.cache-hit != 'true'
shell: bash
run: |
.github/scripts/compute-fbank-librispeech-test-clean-and-test-other.sh
- name: Inference with pre-trained model
shell: bash
env:
GITHUB_EVENT_NAME: ${{ github.event_name }}
GITHUB_EVENT_LABEL_NAME: ${{ github.event.label.name }}
run: |
mkdir -p egs/librispeech/ASR/data
ln -sfv ~/tmp/fbank-libri egs/librispeech/ASR/data/fbank
ls -lh egs/librispeech/ASR/data/*
sudo apt-get -qq install git-lfs tree
export PYTHONPATH=$PWD:$PYTHONPATH
export PYTHONPATH=~/tmp/kaldifeat/kaldifeat/python:$PYTHONPATH
export PYTHONPATH=~/tmp/kaldifeat/build/lib:$PYTHONPATH
.github/scripts/run-librispeech-pruned-transducer-stateless7-streaming-2022-12-29.sh
- name: Display decoding results for librispeech pruned_transducer_stateless7_streaming
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
shell: bash
run: |
cd egs/librispeech/ASR/
tree ./pruned_transducer_stateless7_streaming/exp
cd pruned_transducer_stateless7_streaming
echo "results for pruned_transducer_stateless7_streaming"
echo "===greedy search==="
find exp/greedy_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/greedy_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
echo "===fast_beam_search==="
find exp/fast_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/fast_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
echo "===modified beam search==="
find exp/modified_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/modified_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
echo "===streaming greedy search==="
find exp/streaming/greedy_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/streaming/greedy_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
echo "===streaming fast_beam_search==="
find exp/streaming/fast_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/streaming/fast_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
echo "===streaming modified beam search==="
find exp/streaming/modified_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/streaming/modified_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
- name: Upload decoding results for librispeech pruned_transducer_stateless7_streaming
uses: actions/upload-artifact@v2
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
with:
name: torch-${{ matrix.torch }}-python-${{ matrix.python-version }}-ubuntu-latest-cpu-pruned_transducer_stateless7-streaming-2022-12-29
path: egs/librispeech/ASR/pruned_transducer_stateless7_streaming/exp/

163
.github/workflows/run-librispeech-2023-01-29-stateless7-ctc-bs.yml vendored
View File

@ -1,163 +0,0 @@
# Copyright 2022 Fangjun Kuang (csukuangfj@gmail.com)
# See ../../LICENSE for clarification regarding multiple authors
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
name: run-librispeech-2023-01-29-stateless7-ctc-bs
# zipformer
on:
push:
branches:
- master
pull_request:
types: [labeled]
schedule:
# minute (0-59)
# hour (0-23)
# day of the month (1-31)
# month (1-12)
# day of the week (0-6)
# nightly build at 15:50 UTC time every day
- cron: "50 15 * * *"
jobs:
run_librispeech_2023_01_29_zipformer_ctc_bs:
if: github.event.label.name == 'run-decode' || github.event.label.name == 'blank-skip' || github.event_name == 'push' || github.event_name == 'schedule'
runs-on: ${{ matrix.os }}
strategy:
matrix:
os: [ubuntu-latest]
python-version: [3.8]
fail-fast: false
steps:
- uses: actions/checkout@v2
with:
fetch-depth: 0
- name: Setup Python ${{ matrix.python-version }}
uses: actions/setup-python@v2
with:
python-version: ${{ matrix.python-version }}
cache: 'pip'
cache-dependency-path: '**/requirements-ci.txt'
- name: Install Python dependencies
run: |
grep -v '^#' ./requirements-ci.txt | xargs -n 1 -L 1 pip install
pip uninstall -y protobuf
pip install --no-binary protobuf protobuf==3.20.*
- name: Cache kaldifeat
id: my-cache
uses: actions/cache@v2
with:
path: |
~/tmp/kaldifeat
key: cache-tmp-${{ matrix.python-version }}-2023-05-22
- name: Install kaldifeat
if: steps.my-cache.outputs.cache-hit != 'true'
shell: bash
run: |
.github/scripts/install-kaldifeat.sh
- name: Cache LibriSpeech test-clean and test-other datasets
id: libri-test-clean-and-test-other-data
uses: actions/cache@v2
with:
path: |
~/tmp/download
key: cache-libri-test-clean-and-test-other
- name: Download LibriSpeech test-clean and test-other
if: steps.libri-test-clean-and-test-other-data.outputs.cache-hit != 'true'
shell: bash
run: |
.github/scripts/download-librispeech-test-clean-and-test-other-dataset.sh
- name: Prepare manifests for LibriSpeech test-clean and test-other
shell: bash
run: |
.github/scripts/prepare-librispeech-test-clean-and-test-other-manifests.sh
- name: Cache LibriSpeech test-clean and test-other fbank features
id: libri-test-clean-and-test-other-fbank
uses: actions/cache@v2
with:
path: |
~/tmp/fbank-libri
key: cache-libri-fbank-test-clean-and-test-other-v2
- name: Compute fbank for LibriSpeech test-clean and test-other
if: steps.libri-test-clean-and-test-other-fbank.outputs.cache-hit != 'true'
shell: bash
run: |
.github/scripts/compute-fbank-librispeech-test-clean-and-test-other.sh
- name: Inference with pre-trained model
shell: bash
env:
GITHUB_EVENT_NAME: ${{ github.event_name }}
GITHUB_EVENT_LABEL_NAME: ${{ github.event.label.name }}
run: |
mkdir -p egs/librispeech/ASR/data
ln -sfv ~/tmp/fbank-libri egs/librispeech/ASR/data/fbank
ls -lh egs/librispeech/ASR/data/*
sudo apt-get -qq install git-lfs tree
export PYTHONPATH=$PWD:$PYTHONPATH
export PYTHONPATH=~/tmp/kaldifeat/kaldifeat/python:$PYTHONPATH
export PYTHONPATH=~/tmp/kaldifeat/build/lib:$PYTHONPATH
.github/scripts/run-librispeech-pruned-transducer-stateless7-ctc-bs-2023-01-29.sh
- name: Display decoding results for librispeech pruned_transducer_stateless7_ctc_bs
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
shell: bash
run: |
cd egs/librispeech/ASR/
tree ./pruned_transducer_stateless7_ctc_bs/exp
cd pruned_transducer_stateless7_ctc_bs
echo "results for pruned_transducer_stateless7_ctc_bs"
echo "===greedy search==="
find exp/greedy_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/greedy_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
echo "===fast_beam_search==="
find exp/fast_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/fast_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
echo "===modified beam search==="
find exp/modified_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/modified_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
echo "===ctc decoding==="
find exp/ctc-decoding -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/ctc-decoding -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
echo "===1best==="
find exp/1best -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/1best -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
- name: Upload decoding results for librispeech pruned_transducer_stateless7_ctc_bs
uses: actions/upload-artifact@v2
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
with:
name: torch-${{ matrix.torch }}-python-${{ matrix.python-version }}-ubuntu-latest-cpu-pruned_transducer_stateless7-ctc-bs-2023-01-29
path: egs/librispeech/ASR/pruned_transducer_stateless7_ctc_bs/exp/

155
.github/workflows/run-librispeech-conformer-ctc3-2022-11-28.yml vendored
View File

@ -1,155 +0,0 @@
# Copyright 2022 Fangjun Kuang (csukuangfj@gmail.com)
# See ../../LICENSE for clarification regarding multiple authors
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
name: run-librispeech-conformer-ctc3-2022-11-28
# zipformer
on:
push:
branches:
- master
pull_request:
types: [labeled]
schedule:
# minute (0-59)
# hour (0-23)
# day of the month (1-31)
# month (1-12)
# day of the week (0-6)
# nightly build at 15:50 UTC time every day
- cron: "50 15 * * *"
concurrency:
group: run_librispeech_2022_11_28_conformer_ctc3-${{ github.ref }}
cancel-in-progress: true
jobs:
run_librispeech_2022_11_28_conformer_ctc3:
if: github.event.label.name == 'ready' || github.event.label.name == 'run-decode' || github.event_name == 'push' || github.event_name == 'schedule'
runs-on: ${{ matrix.os }}
strategy:
matrix:
os: [ubuntu-latest]
python-version: [3.8]
fail-fast: false
steps:
- uses: actions/checkout@v2
with:
fetch-depth: 0
- name: Setup Python ${{ matrix.python-version }}
uses: actions/setup-python@v2
with:
python-version: ${{ matrix.python-version }}
cache: 'pip'
cache-dependency-path: '**/requirements-ci.txt'
- name: Install Python dependencies
run: |
grep -v '^#' ./requirements-ci.txt | xargs -n 1 -L 1 pip install
pip uninstall -y protobuf
pip install --no-binary protobuf protobuf==3.20.*
- name: Cache kaldifeat
id: my-cache
uses: actions/cache@v2
with:
path: |
~/tmp/kaldifeat
key: cache-tmp-${{ matrix.python-version }}-2023-05-22
- name: Install kaldifeat
if: steps.my-cache.outputs.cache-hit != 'true'
shell: bash
run: |
.github/scripts/install-kaldifeat.sh
- name: Cache LibriSpeech test-clean and test-other datasets
id: libri-test-clean-and-test-other-data
uses: actions/cache@v2
with:
path: |
~/tmp/download
key: cache-libri-test-clean-and-test-other
- name: Download LibriSpeech test-clean and test-other
if: steps.libri-test-clean-and-test-other-data.outputs.cache-hit != 'true'
shell: bash
run: |
.github/scripts/download-librispeech-test-clean-and-test-other-dataset.sh
- name: Prepare manifests for LibriSpeech test-clean and test-other
shell: bash
run: |
.github/scripts/prepare-librispeech-test-clean-and-test-other-manifests.sh
- name: Cache LibriSpeech test-clean and test-other fbank features
id: libri-test-clean-and-test-other-fbank
uses: actions/cache@v2
with:
path: |
~/tmp/fbank-libri
key: cache-libri-fbank-test-clean-and-test-other-v2
- name: Compute fbank for LibriSpeech test-clean and test-other
if: steps.libri-test-clean-and-test-other-fbank.outputs.cache-hit != 'true'
shell: bash
run: |
.github/scripts/compute-fbank-librispeech-test-clean-and-test-other.sh
- name: Inference with pre-trained model
shell: bash
env:
GITHUB_EVENT_NAME: ${{ github.event_name }}
GITHUB_EVENT_LABEL_NAME: ${{ github.event.label.name }}
run: |
mkdir -p egs/librispeech/ASR/data
ln -sfv ~/tmp/fbank-libri egs/librispeech/ASR/data/fbank
ls -lh egs/librispeech/ASR/data/*
sudo apt-get -qq install git-lfs tree
export PYTHONPATH=$PWD:$PYTHONPATH
export PYTHONPATH=~/tmp/kaldifeat/kaldifeat/python:$PYTHONPATH
export PYTHONPATH=~/tmp/kaldifeat/build/lib:$PYTHONPATH
.github/scripts/run-librispeech-conformer-ctc3-2022-11-28.sh
- name: Display decoding results for librispeech conformer_ctc3
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
shell: bash
run: |
cd egs/librispeech/ASR/
tree ./conformer_ctc3/exp
cd conformer_ctc3
echo "results for conformer_ctc3"
echo "===ctc-decoding==="
find exp/ctc-decoding -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/ctc-decoding -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
echo "===1best==="
find exp/1best -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/1best -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
- name: Upload decoding results for librispeech conformer_ctc3
uses: actions/upload-artifact@v2
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
with:
name: torch-${{ matrix.torch }}-python-${{ matrix.python-version }}-ubuntu-latest-cpu-conformer_ctc3-2022-11-28
path: egs/librispeech/ASR/conformer_ctc3/exp/

157
.github/workflows/run-librispeech-pruned-transducer-stateless3-2022-05-13.yml vendored
View File

@ -1,157 +0,0 @@
# Copyright 2021 Fangjun Kuang (csukuangfj@gmail.com)
# See ../../LICENSE for clarification regarding multiple authors
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
name: run-librispeech-pruned-transducer-stateless3-2022-05-13
# stateless pruned transducer (reworked model) + giga speech
on:
push:
branches:
- master
pull_request:
types: [labeled]
schedule:
# minute (0-59)
# hour (0-23)
# day of the month (1-31)
# month (1-12)
# day of the week (0-6)
# nightly build at 15:50 UTC time every day
- cron: "50 15 * * *"
concurrency:
group: run_librispeech_pruned_transducer_stateless3_2022_05_13-${{ github.ref }}
cancel-in-progress: true
jobs:
run_librispeech_pruned_transducer_stateless3_2022_05_13:
if: github.event.label.name == 'ready' || github.event.label.name == 'run-decode' || github.event_name == 'push' || github.event_name == 'schedule'
runs-on: ${{ matrix.os }}
strategy:
matrix:
os: [ubuntu-latest]
python-version: [3.8]
fail-fast: false
steps:
- uses: actions/checkout@v2
with:
fetch-depth: 0
- name: Setup Python ${{ matrix.python-version }}
uses: actions/setup-python@v2
with:
python-version: ${{ matrix.python-version }}
cache: 'pip'
cache-dependency-path: '**/requirements-ci.txt'
- name: Install Python dependencies
run: |
grep -v '^#' ./requirements-ci.txt | xargs -n 1 -L 1 pip install
pip uninstall -y protobuf
pip install --no-binary protobuf protobuf==3.20.*
- name: Cache kaldifeat
id: my-cache
uses: actions/cache@v2
with:
path: |
~/tmp/kaldifeat
key: cache-tmp-${{ matrix.python-version }}-2023-05-22
- name: Install kaldifeat
if: steps.my-cache.outputs.cache-hit != 'true'
shell: bash
run: |
.github/scripts/install-kaldifeat.sh
- name: Cache LibriSpeech test-clean and test-other datasets
id: libri-test-clean-and-test-other-data
uses: actions/cache@v2
with:
path: |
~/tmp/download
key: cache-libri-test-clean-and-test-other
- name: Download LibriSpeech test-clean and test-other
if: steps.libri-test-clean-and-test-other-data.outputs.cache-hit != 'true'
shell: bash
run: |
.github/scripts/download-librispeech-test-clean-and-test-other-dataset.sh
- name: Prepare manifests for LibriSpeech test-clean and test-other
shell: bash
run: |
.github/scripts/prepare-librispeech-test-clean-and-test-other-manifests.sh
- name: Cache LibriSpeech test-clean and test-other fbank features
id: libri-test-clean-and-test-other-fbank
uses: actions/cache@v2
with:
path: |
~/tmp/fbank-libri
key: cache-libri-fbank-test-clean-and-test-other-v2
- name: Compute fbank for LibriSpeech test-clean and test-other
if: steps.libri-test-clean-and-test-other-fbank.outputs.cache-hit != 'true'
shell: bash
run: |
.github/scripts/compute-fbank-librispeech-test-clean-and-test-other.sh
- name: Inference with pre-trained model
shell: bash
env:
GITHUB_EVENT_NAME: ${{ github.event_name }}
GITHUB_EVENT_LABEL_NAME: ${{ github.event.label.name }}
run: |
mkdir -p egs/librispeech/ASR/data
ln -sfv ~/tmp/fbank-libri egs/librispeech/ASR/data/fbank
ls -lh egs/librispeech/ASR/data/*
sudo apt-get -qq install git-lfs tree
export PYTHONPATH=$PWD:$PYTHONPATH
export PYTHONPATH=~/tmp/kaldifeat/kaldifeat/python:$PYTHONPATH
export PYTHONPATH=~/tmp/kaldifeat/build/lib:$PYTHONPATH
.github/scripts/run-librispeech-pruned-transducer-stateless3-2022-05-13.sh
- name: Display decoding results for pruned_transducer_stateless3
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
shell: bash
run: |
cd egs/librispeech/ASR
tree pruned_transducer_stateless3/exp
cd pruned_transducer_stateless3/exp
echo "===greedy search==="
find greedy_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find greedy_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
echo "===fast_beam_search==="
find fast_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find fast_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
echo "===modified beam search==="
find modified_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find modified_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
- name: Upload decoding results for pruned_transducer_stateless3
uses: actions/upload-artifact@v2
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
with:
name: torch-${{ matrix.torch }}-python-${{ matrix.python-version }}-ubuntu-latest-cpu-pruned_transducer_stateless3-2022-04-29
path: egs/librispeech/ASR/pruned_transducer_stateless3/exp/

159
.github/workflows/run-librispeech-streaming-transducer-stateless2-2022-06-26.yml vendored
View File

@ -1,159 +0,0 @@
# Copyright 2021 Fangjun Kuang (csukuangfj@gmail.com)
# See ../../LICENSE for clarification regarding multiple authors
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
name: run-librispeech-streaming-2022-06-26
# streaming conformer stateless transducer2
on:
push:
branches:
- master
pull_request:
types: [labeled]
schedule:
# minute (0-59)
# hour (0-23)
# day of the month (1-31)
# month (1-12)
# day of the week (0-6)
# nightly build at 15:50 UTC time every day
- cron: "50 15 * * *"
concurrency:
group: run_librispeech_streaming_2022_06_26-${{ github.ref }}
cancel-in-progress: true
jobs:
run_librispeech_streaming_2022_06_26:
if: github.event.label.name == 'ready' || github.event.label.name == 'run-decode' || github.event_name == 'push' || github.event_name == 'schedule'
runs-on: ${{ matrix.os }}
strategy:
matrix:
os: [ubuntu-latest]
python-version: [3.8]
fail-fast: false
steps:
- uses: actions/checkout@v2
with:
fetch-depth: 0
- name: Setup Python ${{ matrix.python-version }}
uses: actions/setup-python@v2
with:
python-version: ${{ matrix.python-version }}
cache: 'pip'
cache-dependency-path: '**/requirements-ci.txt'
- name: Install Python dependencies
run: |
grep -v '^#' ./requirements-ci.txt | xargs -n 1 -L 1 pip install
pip uninstall -y protobuf
pip install --no-binary protobuf protobuf==3.20.*
- name: Cache kaldifeat
id: my-cache
uses: actions/cache@v2
with:
path: |
~/tmp/kaldifeat
key: cache-tmp-${{ matrix.python-version }}-2023-05-22
- name: Install kaldifeat
if: steps.my-cache.outputs.cache-hit != 'true'
shell: bash
run: |
.github/scripts/install-kaldifeat.sh
- name: Cache LibriSpeech test-clean and test-other datasets
id: libri-test-clean-and-test-other-data
uses: actions/cache@v2
with:
path: |
~/tmp/download
key: cache-libri-test-clean-and-test-other
- name: Download LibriSpeech test-clean and test-other
if: steps.libri-test-clean-and-test-other-data.outputs.cache-hit != 'true'
shell: bash
run: |
.github/scripts/download-librispeech-test-clean-and-test-other-dataset.sh
- name: Prepare manifests for LibriSpeech test-clean and test-other
shell: bash
run: |
.github/scripts/prepare-librispeech-test-clean-and-test-other-manifests.sh
- name: Cache LibriSpeech test-clean and test-other fbank features
id: libri-test-clean-and-test-other-fbank
uses: actions/cache@v2
with:
path: |
~/tmp/fbank-libri
key: cache-libri-fbank-test-clean-and-test-other-v2
- name: Compute fbank for LibriSpeech test-clean and test-other
if: steps.libri-test-clean-and-test-other-fbank.outputs.cache-hit != 'true'
shell: bash
run: |
.github/scripts/compute-fbank-librispeech-test-clean-and-test-other.sh
- name: Inference with pre-trained model
shell: bash
env:
GITHUB_EVENT_NAME: ${{ github.event_name }}
GITHUB_EVENT_LABEL_NAME: ${{ github.event.label.name }}
run: |
mkdir -p egs/librispeech/ASR/data
ln -sfv ~/tmp/fbank-libri egs/librispeech/ASR/data/fbank
ls -lh egs/librispeech/ASR/data/*
sudo apt-get -qq install git-lfs tree
export PYTHONPATH=$PWD:$PYTHONPATH
export PYTHONPATH=~/tmp/kaldifeat/kaldifeat/python:$PYTHONPATH
export PYTHONPATH=~/tmp/kaldifeat/build/lib:$PYTHONPATH
.github/scripts/run-librispeech-streaming-pruned-transducer-stateless2-2022-06-26.sh
- name: Display decoding results
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
shell: bash
run: |
cd egs/librispeech/ASR/
tree ./pruned_transducer_stateless2/exp
cd pruned_transducer_stateless2
echo "results for pruned_transducer_stateless2"
echo "===greedy search==="
find exp/greedy_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/greedy_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
echo "===fast_beam_search==="
find exp/fast_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/fast_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
echo "===modified_beam_search==="
find exp/modified_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/modified_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
- name: Upload decoding results for pruned_transducer_stateless2
uses: actions/upload-artifact@v2
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
with:
name: torch-${{ matrix.torch }}-python-${{ matrix.python-version }}-ubuntu-latest-cpu-pruned_transducer_stateless2-2022-06-26
path: egs/librispeech/ASR/pruned_transducer_stateless2/exp/

174
.github/workflows/run-librispeech-streaming-zipformer-2023-05-18.yml vendored
View File

@ -1,174 +0,0 @@
# Copyright 2022 Fangjun Kuang (csukuangfj@gmail.com)
# See ../../LICENSE for clarification regarding multiple authors
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
name: run-librispeech-streaming-zipformer-2023-05-18
# zipformer
on:
push:
branches:
- master
pull_request:
types: [labeled]
schedule:
# minute (0-59)
# hour (0-23)
# day of the month (1-31)
# month (1-12)
# day of the week (0-6)
# nightly build at 15:50 UTC time every day
- cron: "50 15 * * *"
concurrency:
group: run_librispeech_2023_05_18_streaming_zipformer-${{ github.ref }}
cancel-in-progress: true
jobs:
run_librispeech_2023_05_18_streaming_zipformer:
if: github.event.label.name == 'zipformer' ||github.event.label.name == 'ready' || github.event.label.name == 'run-decode' || github.event_name == 'push' || github.event_name == 'schedule'
runs-on: ${{ matrix.os }}
strategy:
matrix:
os: [ubuntu-latest]
python-version: [3.8]
fail-fast: false
steps:
- uses: actions/checkout@v2
with:
fetch-depth: 0
- name: Setup Python ${{ matrix.python-version }}
uses: actions/setup-python@v2
with:
python-version: ${{ matrix.python-version }}
cache: 'pip'
cache-dependency-path: '**/requirements-ci.txt'
- name: Install Python dependencies
run: |
grep -v '^#' ./requirements-ci.txt | xargs -n 1 -L 1 pip install
pip uninstall -y protobuf
pip install --no-binary protobuf protobuf==3.20.*
- name: Cache kaldifeat
id: my-cache
uses: actions/cache@v2
with:
path: |
~/tmp/kaldifeat
key: cache-tmp-${{ matrix.python-version }}-2023-05-22
- name: Install kaldifeat
if: steps.my-cache.outputs.cache-hit != 'true'
shell: bash
run: |
.github/scripts/install-kaldifeat.sh
- name: Cache LibriSpeech test-clean and test-other datasets
id: libri-test-clean-and-test-other-data
uses: actions/cache@v2
with:
path: |
~/tmp/download
key: cache-libri-test-clean-and-test-other
- name: Download LibriSpeech test-clean and test-other
if: steps.libri-test-clean-and-test-other-data.outputs.cache-hit != 'true'
shell: bash
run: |
.github/scripts/download-librispeech-test-clean-and-test-other-dataset.sh
- name: Prepare manifests for LibriSpeech test-clean and test-other
shell: bash
run: |
.github/scripts/prepare-librispeech-test-clean-and-test-other-manifests.sh
- name: Cache LibriSpeech test-clean and test-other fbank features
id: libri-test-clean-and-test-other-fbank
uses: actions/cache@v2
with:
path: |
~/tmp/fbank-libri
key: cache-libri-fbank-test-clean-and-test-other-v2
- name: Compute fbank for LibriSpeech test-clean and test-other
if: steps.libri-test-clean-and-test-other-fbank.outputs.cache-hit != 'true'
shell: bash
run: |
.github/scripts/compute-fbank-librispeech-test-clean-and-test-other.sh
- name: Inference with pre-trained model
shell: bash
env:
GITHUB_EVENT_NAME: ${{ github.event_name }}
GITHUB_EVENT_LABEL_NAME: ${{ github.event.label.name }}
run: |
mkdir -p egs/librispeech/ASR/data
ln -sfv ~/tmp/fbank-libri egs/librispeech/ASR/data/fbank
ls -lh egs/librispeech/ASR/data/*
sudo apt-get -qq install git-lfs tree
export PYTHONPATH=$PWD:$PYTHONPATH
export PYTHONPATH=~/tmp/kaldifeat/kaldifeat/python:$PYTHONPATH
export PYTHONPATH=~/tmp/kaldifeat/build/lib:$PYTHONPATH
.github/scripts/run-librispeech-streaming-zipformer-2023-05-18.sh
- name: Display decoding results for librispeech zipformer
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
shell: bash
run: |
cd egs/librispeech/ASR/
tree ./zipformer/exp
cd zipformer
echo "results for zipformer, simulated streaming decoding"
echo "===greedy search==="
find exp/greedy_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/greedy_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
echo "===fast_beam_search==="
find exp/fast_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/fast_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
echo "===modified beam search==="
find exp/modified_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/modified_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
echo "results for zipformer, chunk-wise streaming decoding"
echo "===greedy search==="
find exp/streaming/greedy_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/streaming/greedy_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
echo "===fast_beam_search==="
find exp/streaming/fast_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/streaming/fast_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
echo "===modified beam search==="
find exp/streaming/modified_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/streaming/modified_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
- name: Upload decoding results for librispeech zipformer
uses: actions/upload-artifact@v2
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
with:
name: torch-${{ matrix.torch }}-python-${{ matrix.python-version }}-ubuntu-latest-cpu-zipformer-2022-11-11
path: egs/librispeech/ASR/zipformer/exp/

159
.github/workflows/run-librispeech-transducer-stateless2-2022-04-19.yml vendored
View File

@ -1,159 +0,0 @@
# Copyright 2021 Fangjun Kuang (csukuangfj@gmail.com)
# See ../../LICENSE for clarification regarding multiple authors
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
name: run-librispeech-2022-04-19
# stateless transducer + torchaudio rnn-t loss
on:
push:
branches:
- master
pull_request:
types: [labeled]
schedule:
# minute (0-59)
# hour (0-23)
# day of the month (1-31)
# month (1-12)
# day of the week (0-6)
# nightly build at 15:50 UTC time every day
- cron: "50 15 * * *"
concurrency:
group: run_librispeech_2022_04_19-${{ github.ref }}
cancel-in-progress: true
jobs:
run_librispeech_2022_04_19:
if: github.event.label.name == 'ready' || github.event.label.name == 'run-decode' || github.event_name == 'push' || github.event_name == 'schedule'
runs-on: ${{ matrix.os }}
strategy:
matrix:
os: [ubuntu-latest]
python-version: [3.8]
fail-fast: false
steps:
- uses: actions/checkout@v2
with:
fetch-depth: 0
- name: Setup Python ${{ matrix.python-version }}
uses: actions/setup-python@v2
with:
python-version: ${{ matrix.python-version }}
cache: 'pip'
cache-dependency-path: '**/requirements-ci.txt'
- name: Install Python dependencies
run: |
grep -v '^#' ./requirements-ci.txt | xargs -n 1 -L 1 pip install
pip uninstall -y protobuf
pip install --no-binary protobuf protobuf==3.20.*
- name: Cache kaldifeat
id: my-cache
uses: actions/cache@v2
with:
path: |
~/tmp/kaldifeat
key: cache-tmp-${{ matrix.python-version }}-2023-05-22
- name: Install kaldifeat
if: steps.my-cache.outputs.cache-hit != 'true'
shell: bash
run: |
.github/scripts/install-kaldifeat.sh
- name: Cache LibriSpeech test-clean and test-other datasets
id: libri-test-clean-and-test-other-data
uses: actions/cache@v2
with:
path: |
~/tmp/download
key: cache-libri-test-clean-and-test-other
- name: Download LibriSpeech test-clean and test-other
if: steps.libri-test-clean-and-test-other-data.outputs.cache-hit != 'true'
shell: bash
run: |
.github/scripts/download-librispeech-test-clean-and-test-other-dataset.sh
- name: Prepare manifests for LibriSpeech test-clean and test-other
shell: bash
run: |
.github/scripts/prepare-librispeech-test-clean-and-test-other-manifests.sh
- name: Cache LibriSpeech test-clean and test-other fbank features
id: libri-test-clean-and-test-other-fbank
uses: actions/cache@v2
with:
path: |
~/tmp/fbank-libri
key: cache-libri-fbank-test-clean-and-test-other-v2
- name: Compute fbank for LibriSpeech test-clean and test-other
if: steps.libri-test-clean-and-test-other-fbank.outputs.cache-hit != 'true'
shell: bash
run: |
.github/scripts/compute-fbank-librispeech-test-clean-and-test-other.sh
- name: Inference with pre-trained model
shell: bash
env:
GITHUB_EVENT_NAME: ${{ github.event_name }}
GITHUB_EVENT_LABEL_NAME: ${{ github.event.label.name }}
run: |
mkdir -p egs/librispeech/ASR/data
ln -sfv ~/tmp/fbank-libri egs/librispeech/ASR/data/fbank
ls -lh egs/librispeech/ASR/data/*
sudo apt-get -qq install git-lfs tree
export PYTHONPATH=$PWD:$PYTHONPATH
export PYTHONPATH=~/tmp/kaldifeat/kaldifeat/python:$PYTHONPATH
export PYTHONPATH=~/tmp/kaldifeat/build/lib:$PYTHONPATH
.github/scripts/run-librispeech-transducer-stateless2-2022-04-19.sh
- name: Display decoding results
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
shell: bash
run: |
cd egs/librispeech/ASR/
tree ./transducer_stateless2/exp
cd transducer_stateless2
echo "results for transducer_stateless2"
echo "===greedy search==="
find exp/greedy_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/greedy_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
echo "===fast_beam_search==="
find exp/fast_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/fast_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
echo "===modified_beam_search==="
find exp/modified_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/modified_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
- name: Upload decoding results for transducer_stateless2
uses: actions/upload-artifact@v2
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
with:
name: torch-${{ matrix.torch }}-python-${{ matrix.python-version }}-ubuntu-latest-cpu-transducer_stateless2-2022-04-19
path: egs/librispeech/ASR/transducer_stateless2/exp/

159
.github/workflows/run-librispeech-zipformer-2023-05-18.yml vendored
View File

@ -1,159 +0,0 @@
# Copyright 2022 Fangjun Kuang (csukuangfj@gmail.com)
# See ../../LICENSE for clarification regarding multiple authors
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
name: run-librispeech-zipformer-2023-05-18
# zipformer
on:
push:
branches:
- master
pull_request:
types: [labeled]
schedule:
# minute (0-59)
# hour (0-23)
# day of the month (1-31)
# month (1-12)
# day of the week (0-6)
# nightly build at 15:50 UTC time every day
- cron: "50 15 * * *"
concurrency:
group: run_librispeech_2023_05_18_zipformer-${{ github.ref }}
cancel-in-progress: true
jobs:
run_librispeech_2023_05_18_zipformer:
if: github.event.label.name == 'zipformer' ||github.event.label.name == 'ready' || github.event.label.name == 'run-decode' || github.event_name == 'push' || github.event_name == 'schedule'
runs-on: ${{ matrix.os }}
strategy:
matrix:
os: [ubuntu-latest]
python-version: [3.8]
fail-fast: false
steps:
- uses: actions/checkout@v2
with:
fetch-depth: 0
- name: Setup Python ${{ matrix.python-version }}
uses: actions/setup-python@v2
with:
python-version: ${{ matrix.python-version }}
cache: 'pip'
cache-dependency-path: '**/requirements-ci.txt'
- name: Install Python dependencies
run: |
grep -v '^#' ./requirements-ci.txt | xargs -n 1 -L 1 pip install
pip uninstall -y protobuf
pip install --no-binary protobuf protobuf==3.20.*
- name: Cache kaldifeat
id: my-cache
uses: actions/cache@v2
with:
path: |
~/tmp/kaldifeat
key: cache-tmp-${{ matrix.python-version }}-2023-05-22
- name: Install kaldifeat
if: steps.my-cache.outputs.cache-hit != 'true'
shell: bash
run: |
.github/scripts/install-kaldifeat.sh
- name: Cache LibriSpeech test-clean and test-other datasets
id: libri-test-clean-and-test-other-data
uses: actions/cache@v2
with:
path: |
~/tmp/download
key: cache-libri-test-clean-and-test-other
- name: Download LibriSpeech test-clean and test-other
if: steps.libri-test-clean-and-test-other-data.outputs.cache-hit != 'true'
shell: bash
run: |
.github/scripts/download-librispeech-test-clean-and-test-other-dataset.sh
- name: Prepare manifests for LibriSpeech test-clean and test-other
shell: bash
run: |
.github/scripts/prepare-librispeech-test-clean-and-test-other-manifests.sh
- name: Cache LibriSpeech test-clean and test-other fbank features
id: libri-test-clean-and-test-other-fbank
uses: actions/cache@v2
with:
path: |
~/tmp/fbank-libri
key: cache-libri-fbank-test-clean-and-test-other-v2
- name: Compute fbank for LibriSpeech test-clean and test-other
if: steps.libri-test-clean-and-test-other-fbank.outputs.cache-hit != 'true'
shell: bash
run: |
.github/scripts/compute-fbank-librispeech-test-clean-and-test-other.sh
- name: Inference with pre-trained model
shell: bash
env:
GITHUB_EVENT_NAME: ${{ github.event_name }}
GITHUB_EVENT_LABEL_NAME: ${{ github.event.label.name }}
run: |
mkdir -p egs/librispeech/ASR/data
ln -sfv ~/tmp/fbank-libri egs/librispeech/ASR/data/fbank
ls -lh egs/librispeech/ASR/data/*
sudo apt-get -qq install git-lfs tree
export PYTHONPATH=$PWD:$PYTHONPATH
export PYTHONPATH=~/tmp/kaldifeat/kaldifeat/python:$PYTHONPATH
export PYTHONPATH=~/tmp/kaldifeat/build/lib:$PYTHONPATH
.github/scripts/run-librispeech-zipformer-2023-05-18.sh
- name: Display decoding results for librispeech zipformer
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
shell: bash
run: |
cd egs/librispeech/ASR/
tree ./zipformer/exp
cd zipformer
echo "results for zipformer"
echo "===greedy search==="
find exp/greedy_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/greedy_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
echo "===fast_beam_search==="
find exp/fast_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/fast_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
echo "===modified beam search==="
find exp/modified_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/modified_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
- name: Upload decoding results for librispeech zipformer
uses: actions/upload-artifact@v2
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
with:
name: torch-${{ matrix.torch }}-python-${{ matrix.python-version }}-ubuntu-latest-cpu-zipformer-2022-11-11
path: egs/librispeech/ASR/zipformer/exp/

155
.github/workflows/run-librispeech-zipformer-ctc-2023-06-14.yml vendored
View File

@ -1,155 +0,0 @@
# Copyright 2022 Fangjun Kuang (csukuangfj@gmail.com)
# See ../../LICENSE for clarification regarding multiple authors
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
name: run-librispeech-zipformer-ctc-2023-06-14
# zipformer
on:
push:
branches:
- master
pull_request:
types: [labeled]
schedule:
# minute (0-59)
# hour (0-23)
# day of the month (1-31)
# month (1-12)
# day of the week (0-6)
# nightly build at 15:50 UTC time every day
- cron: "50 15 * * *"
concurrency:
group: run_librispeech_2023_06_14_zipformer-ctc-${{ github.ref }}
cancel-in-progress: true
jobs:
run_librispeech_2023_06_14_zipformer_ctc:
if: github.event.label.name == 'zipformer' ||github.event.label.name == 'ready' || github.event.label.name == 'run-decode' || github.event_name == 'push' || github.event_name == 'schedule'
runs-on: ${{ matrix.os }}
strategy:
matrix:
os: [ubuntu-latest]
python-version: [3.8]
fail-fast: false
steps:
- uses: actions/checkout@v2
with:
fetch-depth: 0
- name: Setup Python ${{ matrix.python-version }}
uses: actions/setup-python@v2
with:
python-version: ${{ matrix.python-version }}
cache: 'pip'
cache-dependency-path: '**/requirements-ci.txt'
- name: Install Python dependencies
run: |
grep -v '^#' ./requirements-ci.txt | xargs -n 1 -L 1 pip install
pip uninstall -y protobuf
pip install --no-binary protobuf protobuf==3.20.*
- name: Cache kaldifeat
id: my-cache
uses: actions/cache@v2
with:
path: |
~/tmp/kaldifeat
key: cache-tmp-${{ matrix.python-version }}-2023-05-22
- name: Install kaldifeat
if: steps.my-cache.outputs.cache-hit != 'true'
shell: bash
run: |
.github/scripts/install-kaldifeat.sh
- name: Cache LibriSpeech test-clean and test-other datasets
id: libri-test-clean-and-test-other-data
uses: actions/cache@v2
with:
path: |
~/tmp/download
key: cache-libri-test-clean-and-test-other
- name: Download LibriSpeech test-clean and test-other
if: steps.libri-test-clean-and-test-other-data.outputs.cache-hit != 'true'
shell: bash
run: |
.github/scripts/download-librispeech-test-clean-and-test-other-dataset.sh
- name: Prepare manifests for LibriSpeech test-clean and test-other
shell: bash
run: |
.github/scripts/prepare-librispeech-test-clean-and-test-other-manifests.sh
- name: Cache LibriSpeech test-clean and test-other fbank features
id: libri-test-clean-and-test-other-fbank
uses: actions/cache@v2
with:
path: |
~/tmp/fbank-libri
key: cache-libri-fbank-test-clean-and-test-other-v2
- name: Compute fbank for LibriSpeech test-clean and test-other
if: steps.libri-test-clean-and-test-other-fbank.outputs.cache-hit != 'true'
shell: bash
run: |
.github/scripts/compute-fbank-librispeech-test-clean-and-test-other.sh
- name: Inference with pre-trained model
shell: bash
env:
GITHUB_EVENT_NAME: ${{ github.event_name }}
GITHUB_EVENT_LABEL_NAME: ${{ github.event.label.name }}
run: |
mkdir -p egs/librispeech/ASR/data
ln -sfv ~/tmp/fbank-libri egs/librispeech/ASR/data/fbank
ls -lh egs/librispeech/ASR/data/*
sudo apt-get -qq install git-lfs tree
export PYTHONPATH=$PWD:$PYTHONPATH
export PYTHONPATH=~/tmp/kaldifeat/kaldifeat/python:$PYTHONPATH
export PYTHONPATH=~/tmp/kaldifeat/build/lib:$PYTHONPATH
.github/scripts/run-librispeech-zipformer-ctc-2023-06-14.sh
- name: Display decoding results for librispeech zipformer
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
shell: bash
run: |
cd egs/librispeech/ASR/
tree ./zipformer/exp
cd zipformer
echo "results for zipformer"
echo "===ctc-decoding==="
find exp/ctc-decoding -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/ctc-decoding -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
echo "===1best==="
find exp/1best -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/1best -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
- name: Upload decoding results for librispeech zipformer
uses: actions/upload-artifact@v2
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
with:
name: torch-${{ matrix.torch }}-python-${{ matrix.python-version }}-ubuntu-latest-cpu-zipformer-2022-11-11
path: egs/librispeech/ASR/zipformer/exp/

10
.github/workflows/run-multi-zh_hans-zipformer.yml → .github/workflows/run-multi-corpora-zipformer.yml vendored
View File

@ -14,7 +14,7 @@
# See the License for the specific language governing permissions and
# limitations under the License.
name: run-multi-zh_hans-zipformer
name: run-multi-corpora-zipformer
on:
push:
@ -24,12 +24,12 @@ on:
types: [labeled]
concurrency:
group: run_multi-zh_hans_zipformer-${{ github.ref }}
group: run_multi-corpora_zipformer-${{ github.ref }}
cancel-in-progress: true
jobs:
run_multi-zh_hans_zipformer:
if: github.event.label.name == 'onnx' || github.event.label.name == 'ready' || github.event_name == 'push' || github.event.label.name == 'multi-zh_hans' || github.event.label.name == 'zipformer'
run_multi-corpora_zipformer:
if: github.event.label.name == 'onnx' || github.event.label.name == 'ready' || github.event_name == 'push' || github.event.label.name == 'multi-zh_hans' || github.event.label.name == 'zipformer' || github.event.label.name == 'multi-corpora'
runs-on: ${{ matrix.os }}
strategy:
matrix:
@ -81,4 +81,4 @@ jobs:
export PYTHONPATH=~/tmp/kaldifeat/kaldifeat/python:$PYTHONPATH
export PYTHONPATH=~/tmp/kaldifeat/build/lib:$PYTHONPATH
.github/scripts/run-multi-zh_hans-zipformer.sh
.github/scripts/run-multi-corpora-zipformer.sh

87
.github/workflows/run-pretrained-ctc.yml vendored
View File

@ -1,87 +0,0 @@
# Copyright 2021 Fangjun Kuang (csukuangfj@gmail.com)
# See ../../LICENSE for clarification regarding multiple authors
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
name: run-pre-trained-ctc
on:
push:
branches:
- master
pull_request:
types: [labeled]
workflow_dispatch:
inputs:
test-run:
description: 'Test (y/n)?'
required: true
default: 'y'
concurrency:
group: run_pre_trained_ctc-${{ github.ref }}
cancel-in-progress: true
jobs:
run_pre_trained_ctc:
if: github.event.label.name == 'ready' || github.event_name == 'push' || github.event.inputs.test-run == 'y' || github.event.label.name == 'ctc'
runs-on: ${{ matrix.os }}
strategy:
matrix:
os: [ubuntu-latest]
python-version: [3.8]
fail-fast: false
steps:
- uses: actions/checkout@v2
with:
fetch-depth: 0
- name: Setup Python ${{ matrix.python-version }}
uses: actions/setup-python@v2
with:
python-version: ${{ matrix.python-version }}
cache: 'pip'
cache-dependency-path: '**/requirements-ci.txt'
- name: Install Python dependencies
run: |
grep -v '^#' ./requirements-ci.txt | xargs -n 1 -L 1 pip install
pip uninstall -y protobuf
pip install --no-binary protobuf protobuf==3.20.*
- name: Cache kaldifeat
id: my-cache
uses: actions/cache@v2
with:
path: |
~/tmp/kaldifeat
key: cache-tmp-${{ matrix.python-version }}-2023-05-22
- name: Install kaldifeat
if: steps.my-cache.outputs.cache-hit != 'true'
shell: bash
run: |
.github/scripts/install-kaldifeat.sh
- name: Inference with pre-trained model
shell: bash
run: |
sudo apt-get -qq install git-lfs tree
export PYTHONPATH=$PWD:$PYTHONPATH
export PYTHONPATH=~/tmp/kaldifeat/kaldifeat/python:$PYTHONPATH
export PYTHONPATH=~/tmp/kaldifeat/build/lib:$PYTHONPATH
.github/scripts/run-pre-trained-ctc.sh

158
.github/workflows/run-pretrained-transducer-stateless-librispeech-100h.yml vendored
View File

@ -1,158 +0,0 @@
# Copyright 2021 Fangjun Kuang (csukuangfj@gmail.com)
# See ../../LICENSE for clarification regarding multiple authors
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
name: run-pre-trained-trandsucer-stateless-multi-datasets-librispeech-100h
on:
push:
branches:
- master
pull_request:
types: [labeled]
schedule:
# minute (0-59)
# hour (0-23)
# day of the month (1-31)
# month (1-12)
# day of the week (0-6)
# nightly build at 15:50 UTC time every day
- cron: "50 15 * * *"
concurrency:
group: run_pre_trained_transducer_stateless_multi_datasets_librispeech_100h-${{ github.ref }}
cancel-in-progress: true
jobs:
run_pre_trained_transducer_stateless_multi_datasets_librispeech_100h:
if: github.event.label.name == 'ready' || github.event.label.name == 'run-decode' || github.event_name == 'push' || github.event_name == 'schedule'
runs-on: ${{ matrix.os }}
strategy:
matrix:
os: [ubuntu-latest]
python-version: [3.8]
fail-fast: false
steps:
- uses: actions/checkout@v2
with:
fetch-depth: 0
- name: Setup Python ${{ matrix.python-version }}
uses: actions/setup-python@v2
with:
python-version: ${{ matrix.python-version }}
cache: 'pip'
cache-dependency-path: '**/requirements-ci.txt'
- name: Install Python dependencies
run: |
grep -v '^#' ./requirements-ci.txt | xargs -n 1 -L 1 pip install
pip uninstall -y protobuf
pip install --no-binary protobuf protobuf==3.20.*
- name: Cache kaldifeat
id: my-cache
uses: actions/cache@v2
with:
path: |
~/tmp/kaldifeat
key: cache-tmp-${{ matrix.python-version }}-2023-05-22
- name: Install kaldifeat
if: steps.my-cache.outputs.cache-hit != 'true'
shell: bash
run: |
.github/scripts/install-kaldifeat.sh
- name: Cache LibriSpeech test-clean and test-other datasets
id: libri-test-clean-and-test-other-data
uses: actions/cache@v2
with:
path: |
~/tmp/download
key: cache-libri-test-clean-and-test-other
- name: Download LibriSpeech test-clean and test-other
if: steps.libri-test-clean-and-test-other-data.outputs.cache-hit != 'true'
shell: bash
run: |
.github/scripts/download-librispeech-test-clean-and-test-other-dataset.sh
- name: Prepare manifests for LibriSpeech test-clean and test-other
shell: bash
run: |
.github/scripts/prepare-librispeech-test-clean-and-test-other-manifests.sh
- name: Cache LibriSpeech test-clean and test-other fbank features
id: libri-test-clean-and-test-other-fbank
uses: actions/cache@v2
with:
path: |
~/tmp/fbank-libri
key: cache-libri-fbank-test-clean-and-test-other-v2
- name: Compute fbank for LibriSpeech test-clean and test-other
if: steps.libri-test-clean-and-test-other-fbank.outputs.cache-hit != 'true'
shell: bash
run: |
.github/scripts/compute-fbank-librispeech-test-clean-and-test-other.sh
- name: Inference with pre-trained model
shell: bash
env:
GITHUB_EVENT_NAME: ${{ github.event_name }}
GITHUB_EVENT_LABEL_NAME: ${{ github.event.label.name }}
run: |
mkdir -p egs/librispeech/ASR/data
ln -sfv ~/tmp/fbank-libri egs/librispeech/ASR/data/fbank
ls -lh egs/librispeech/ASR/data/*
sudo apt-get -qq install git-lfs tree
export PYTHONPATH=$PWD:$PYTHONPATH
export PYTHONPATH=~/tmp/kaldifeat/kaldifeat/python:$PYTHONPATH
export PYTHONPATH=~/tmp/kaldifeat/build/lib:$PYTHONPATH
.github/scripts/run-pre-trained-transducer-stateless-librispeech-100h.sh
- name: Display decoding results for transducer_stateless_multi_datasets
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
shell: bash
run: |
cd egs/librispeech/ASR/
tree ./transducer_stateless_multi_datasets/exp
cd transducer_stateless_multi_datasets
echo "results for transducer_stateless_multi_datasets"
echo "===greedy search==="
find exp/greedy_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/greedy_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
echo "===fast_beam_search==="
find exp/fast_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/fast_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
echo "===modified beam search==="
find exp/modified_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/modified_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
- name: Upload decoding results for transducer_stateless_multi_datasets
uses: actions/upload-artifact@v2
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
with:
name: torch-${{ matrix.torch }}-python-${{ matrix.python-version }}-ubuntu-latest-cpu-transducer_stateless_multi_datasets-100h-2022-02-21
path: egs/librispeech/ASR/transducer_stateless_multi_datasets/exp/

158
.github/workflows/run-pretrained-transducer-stateless-librispeech-multi-datasets.yml vendored
View File

@ -1,158 +0,0 @@
# Copyright 2021 Fangjun Kuang (csukuangfj@gmail.com)
# See ../../LICENSE for clarification regarding multiple authors
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
name: run-pre-trained-trandsucer-stateless-multi-datasets-librispeech-960h
on:
push:
branches:
- master
pull_request:
types: [labeled]
schedule:
# minute (0-59)
# hour (0-23)
# day of the month (1-31)
# month (1-12)
# day of the week (0-6)
# nightly build at 15:50 UTC time every day
- cron: "50 15 * * *"
concurrency:
group: run_pre_trained_transducer_stateless_multi_datasets_librispeech_960h-${{ github.ref }}
cancel-in-progress: true
jobs:
run_pre_trained_transducer_stateless_multi_datasets_librispeech_960h:
if: github.event.label.name == 'ready' || github.event.label.name == 'run-decode' || github.event_name == 'push' || github.event_name == 'schedule'
runs-on: ${{ matrix.os }}
strategy:
matrix:
os: [ubuntu-latest]
python-version: [3.8]
fail-fast: false
steps:
- uses: actions/checkout@v2
with:
fetch-depth: 0
- name: Setup Python ${{ matrix.python-version }}
uses: actions/setup-python@v2
with:
python-version: ${{ matrix.python-version }}
cache: 'pip'
cache-dependency-path: '**/requirements-ci.txt'
- name: Install Python dependencies
run: |
grep -v '^#' ./requirements-ci.txt | xargs -n 1 -L 1 pip install
pip uninstall -y protobuf
pip install --no-binary protobuf protobuf==3.20.*
- name: Cache kaldifeat
id: my-cache
uses: actions/cache@v2
with:
path: |
~/tmp/kaldifeat
key: cache-tmp-${{ matrix.python-version }}-2023-05-22
- name: Install kaldifeat
if: steps.my-cache.outputs.cache-hit != 'true'
shell: bash
run: |
.github/scripts/install-kaldifeat.sh
- name: Cache LibriSpeech test-clean and test-other datasets
id: libri-test-clean-and-test-other-data
uses: actions/cache@v2
with:
path: |
~/tmp/download
key: cache-libri-test-clean-and-test-other
- name: Download LibriSpeech test-clean and test-other
if: steps.libri-test-clean-and-test-other-data.outputs.cache-hit != 'true'
shell: bash
run: |
.github/scripts/download-librispeech-test-clean-and-test-other-dataset.sh
- name: Prepare manifests for LibriSpeech test-clean and test-other
shell: bash
run: |
.github/scripts/prepare-librispeech-test-clean-and-test-other-manifests.sh
- name: Cache LibriSpeech test-clean and test-other fbank features
id: libri-test-clean-and-test-other-fbank
uses: actions/cache@v2
with:
path: |
~/tmp/fbank-libri
key: cache-libri-fbank-test-clean-and-test-other-v2
- name: Compute fbank for LibriSpeech test-clean and test-other
if: steps.libri-test-clean-and-test-other-fbank.outputs.cache-hit != 'true'
shell: bash
run: |
.github/scripts/compute-fbank-librispeech-test-clean-and-test-other.sh
- name: Inference with pre-trained model
shell: bash
env:
GITHUB_EVENT_NAME: ${{ github.event_name }}
GITHUB_EVENT_LABEL_NAME: ${{ github.event.label.name }}
run: |
mkdir -p egs/librispeech/ASR/data
ln -sfv ~/tmp/fbank-libri egs/librispeech/ASR/data/fbank
ls -lh egs/librispeech/ASR/data/*
sudo apt-get -qq install git-lfs tree
export PYTHONPATH=$PWD:$PYTHONPATH
export PYTHONPATH=~/tmp/kaldifeat/kaldifeat/python:$PYTHONPATH
export PYTHONPATH=~/tmp/kaldifeat/build/lib:$PYTHONPATH
.github/scripts/run-pre-trained-transducer-stateless-librispeech-960h.sh
- name: Display decoding results for transducer_stateless_multi_datasets
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
shell: bash
run: |
cd egs/librispeech/ASR/
tree ./transducer_stateless_multi_datasets/exp
cd transducer_stateless_multi_datasets
echo "results for transducer_stateless_multi_datasets"
echo "===greedy search==="
find exp/greedy_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/greedy_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
echo "===fast_beam_search==="
find exp/fast_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/fast_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
echo "===modified beam search==="
find exp/modified_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/modified_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
- name: Upload decoding results for transducer_stateless_multi_datasets
uses: actions/upload-artifact@v2
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
with:
name: torch-${{ matrix.torch }}-python-${{ matrix.python-version }}-ubuntu-latest-cpu-transducer_stateless_multi_datasets-100h-2022-03-01
path: egs/librispeech/ASR/transducer_stateless_multi_datasets/exp/

80
.github/workflows/run-pretrained-transducer-stateless-modified-2-aishell.yml vendored
View File

@ -1,80 +0,0 @@
# Copyright 2021 Fangjun Kuang (csukuangfj@gmail.com)
# See ../../LICENSE for clarification regarding multiple authors
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
name: run-pre-trained-trandsucer-stateless-modified-2-aishell
on:
push:
branches:
- master
pull_request:
types: [labeled]
concurrency:
group: run_pre_trained_transducer_stateless_modified_2_aishell-${{ github.ref }}
cancel-in-progress: true
jobs:
run_pre_trained_transducer_stateless_modified_2_aishell:
if: github.event.label.name == 'ready' || github.event_name == 'push'
runs-on: ${{ matrix.os }}
strategy:
matrix:
os: [ubuntu-latest]
python-version: [3.8]
fail-fast: false
steps:
- uses: actions/checkout@v2
with:
fetch-depth: 0
- name: Setup Python ${{ matrix.python-version }}
uses: actions/setup-python@v2
with:
python-version: ${{ matrix.python-version }}
cache: 'pip'
cache-dependency-path: '**/requirements-ci.txt'
- name: Install Python dependencies
run: |
grep -v '^#' ./requirements-ci.txt | xargs -n 1 -L 1 pip install
pip uninstall -y protobuf
pip install --no-binary protobuf protobuf==3.20.*
- name: Cache kaldifeat
id: my-cache
uses: actions/cache@v2
with:
path: |
~/tmp/kaldifeat
key: cache-tmp-${{ matrix.python-version }}-2023-05-22
- name: Install kaldifeat
if: steps.my-cache.outputs.cache-hit != 'true'
shell: bash
run: |
.github/scripts/install-kaldifeat.sh
- name: Inference with pre-trained model
shell: bash
run: |
sudo apt-get -qq install git-lfs tree
export PYTHONPATH=$PWD:$PYTHONPATH
export PYTHONPATH=~/tmp/kaldifeat/kaldifeat/python:$PYTHONPATH
export PYTHONPATH=~/tmp/kaldifeat/build/lib:$PYTHONPATH
.github/scripts/run-pre-trained-transducer-stateless-modified-2-aishell.sh

80
.github/workflows/run-pretrained-transducer-stateless-modified-aishell.yml vendored
View File

@ -1,80 +0,0 @@
# Copyright 2021 Fangjun Kuang (csukuangfj@gmail.com)
# See ../../LICENSE for clarification regarding multiple authors
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
name: run-pre-trained-trandsucer-stateless-modified-aishell
on:
push:
branches:
- master
pull_request:
types: [labeled]
concurrency:
group: run_pre_trained_transducer_stateless_modified_aishell-${{ github.ref }}
cancel-in-progress: true
jobs:
run_pre_trained_transducer_stateless_modified_aishell:
if: github.event.label.name == 'ready' || github.event_name == 'push'
runs-on: ${{ matrix.os }}
strategy:
matrix:
os: [ubuntu-latest]
python-version: [3.8]
fail-fast: false
steps:
- uses: actions/checkout@v2
with:
fetch-depth: 0
- name: Setup Python ${{ matrix.python-version }}
uses: actions/setup-python@v2
with:
python-version: ${{ matrix.python-version }}
cache: 'pip'
cache-dependency-path: '**/requirements-ci.txt'
- name: Install Python dependencies
run: |
grep -v '^#' ./requirements-ci.txt | xargs -n 1 -L 1 pip install
pip uninstall -y protobuf
pip install --no-binary protobuf protobuf==3.20.*
- name: Cache kaldifeat
id: my-cache
uses: actions/cache@v2
with:
path: |
~/tmp/kaldifeat
key: cache-tmp-${{ matrix.python-version }}-2023-05-22
- name: Install kaldifeat
if: steps.my-cache.outputs.cache-hit != 'true'
shell: bash
run: |
.github/scripts/install-kaldifeat.sh
- name: Inference with pre-trained model
shell: bash
run: |
sudo apt-get -qq install git-lfs tree
export PYTHONPATH=$PWD:$PYTHONPATH
export PYTHONPATH=~/tmp/kaldifeat/kaldifeat/python:$PYTHONPATH
export PYTHONPATH=~/tmp/kaldifeat/build/lib:$PYTHONPATH
.github/scripts/run-pre-trained-transducer-stateless-modified-aishell.sh

158
.github/workflows/run-pretrained-transducer-stateless.yml vendored
View File

@ -1,158 +0,0 @@
# Copyright 2021 Fangjun Kuang (csukuangfj@gmail.com)
# See ../../LICENSE for clarification regarding multiple authors
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
name: run-pre-trained-transducer-stateless
on:
push:
branches:
- master
pull_request:
types: [labeled]
schedule:
# minute (0-59)
# hour (0-23)
# day of the month (1-31)
# month (1-12)
# day of the week (0-6)
# nightly build at 15:50 UTC time every day
- cron: "50 15 * * *"
concurrency:
group: run_pre_trained_transducer_stateless-${{ github.ref }}
cancel-in-progress: true
jobs:
run_pre_trained_transducer_stateless:
if: github.event.label.name == 'ready' || github.event.label.name == 'run-decode' || github.event_name == 'push' || github.event_name == 'schedule'
runs-on: ${{ matrix.os }}
strategy:
matrix:
os: [ubuntu-latest]
python-version: [3.8]
fail-fast: false
steps:
- uses: actions/checkout@v2
with:
fetch-depth: 0
- name: Setup Python ${{ matrix.python-version }}
uses: actions/setup-python@v2
with:
python-version: ${{ matrix.python-version }}
cache: 'pip'
cache-dependency-path: '**/requirements-ci.txt'
- name: Install Python dependencies
run: |
grep -v '^#' ./requirements-ci.txt | xargs -n 1 -L 1 pip install
pip uninstall -y protobuf
pip install --no-binary protobuf protobuf==3.20.*
- name: Cache kaldifeat
id: my-cache
uses: actions/cache@v2
with:
path: |
~/tmp/kaldifeat
key: cache-tmp-${{ matrix.python-version }}-2023-05-22
- name: Install kaldifeat
if: steps.my-cache.outputs.cache-hit != 'true'
shell: bash
run: |
.github/scripts/install-kaldifeat.sh
- name: Cache LibriSpeech test-clean and test-other datasets
id: libri-test-clean-and-test-other-data
uses: actions/cache@v2
with:
path: |
~/tmp/download
key: cache-libri-test-clean-and-test-other
- name: Download LibriSpeech test-clean and test-other
if: steps.libri-test-clean-and-test-other-data.outputs.cache-hit != 'true'
shell: bash
run: |
.github/scripts/download-librispeech-test-clean-and-test-other-dataset.sh
- name: Prepare manifests for LibriSpeech test-clean and test-other
shell: bash
run: |
.github/scripts/prepare-librispeech-test-clean-and-test-other-manifests.sh
- name: Cache LibriSpeech test-clean and test-other fbank features
id: libri-test-clean-and-test-other-fbank
uses: actions/cache@v2
with:
path: |
~/tmp/fbank-libri
key: cache-libri-fbank-test-clean-and-test-other-v2
- name: Compute fbank for LibriSpeech test-clean and test-other
if: steps.libri-test-clean-and-test-other-fbank.outputs.cache-hit != 'true'
shell: bash
run: |
.github/scripts/compute-fbank-librispeech-test-clean-and-test-other.sh
- name: Inference with pre-trained model
shell: bash
env:
GITHUB_EVENT_NAME: ${{ github.event_name }}
GITHUB_EVENT_LABEL_NAME: ${{ github.event.label.name }}
run: |
mkdir -p egs/librispeech/ASR/data
ln -sfv ~/tmp/fbank-libri egs/librispeech/ASR/data/fbank
ls -lh egs/librispeech/ASR/data/*
sudo apt-get -qq install git-lfs tree
export PYTHONPATH=$PWD:$PYTHONPATH
export PYTHONPATH=~/tmp/kaldifeat/kaldifeat/python:$PYTHONPATH
export PYTHONPATH=~/tmp/kaldifeat/build/lib:$PYTHONPATH
.github/scripts/run-pre-trained-transducer-stateless.sh
- name: Display decoding results for transducer_stateless
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
shell: bash
run: |
cd egs/librispeech/ASR/
tree ./transducer_stateless/exp
cd transducer_stateless
echo "results for transducer_stateless"
echo "===greedy search==="
find exp/greedy_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/greedy_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
echo "===fast_beam_search==="
find exp/fast_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/fast_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
echo "===modified beam search==="
find exp/modified_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/modified_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
- name: Upload decoding results for transducer_stateless
uses: actions/upload-artifact@v2
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
with:
name: torch-${{ matrix.torch }}-python-${{ matrix.python-version }}-ubuntu-latest-cpu-transducer_stateless-2022-02-07
path: egs/librispeech/ASR/transducer_stateless/exp/

185
.github/workflows/run-yesno-recipe.yml vendored
View File

@ -1,185 +0,0 @@
# Copyright 2021 Fangjun Kuang (csukuangfj@gmail.com)
# See ../../LICENSE for clarification regarding multiple authors
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
name: run-yesno-recipe
on:
push:
branches:
- master
pull_request:
branches:
- master
concurrency:
group: run-yesno-recipe-${{ github.ref }}
cancel-in-progress: true
jobs:
run-yesno-recipe:
runs-on: ${{ matrix.os }}
strategy:
matrix:
# os: [ubuntu-latest, macos-10.15]
# TODO: enable macOS for CPU testing
os: [ubuntu-latest]
python-version: [3.8]
fail-fast: false
steps:
- uses: actions/checkout@v2
with:
fetch-depth: 0
- name: Setup Python ${{ matrix.python-version }}
uses: actions/setup-python@v2
with:
python-version: ${{ matrix.python-version }}
cache: 'pip'
cache-dependency-path: '**/requirements-ci.txt'
- name: Install libnsdfile and libsox
if: startsWith(matrix.os, 'ubuntu')
run: |
sudo apt update
sudo apt install -q -y libsndfile1-dev libsndfile1 ffmpeg
sudo apt install -q -y --fix-missing sox libsox-dev libsox-fmt-all
- name: Install Python dependencies
run: |
grep -v '^#' ./requirements-ci.txt | xargs -n 1 -L 1 pip install
pip uninstall -y protobuf
pip install --no-binary protobuf protobuf==3.20.*
pip install --no-deps --force-reinstall k2==1.24.4.dev20231021+cpu.torch1.13.1 -f https://k2-fsa.github.io/k2/cpu.html
pip install kaldifeat==1.25.1.dev20231022+cpu.torch1.13.1 -f https://csukuangfj.github.io/kaldifeat/cpu.html
- name: Run yesno recipe
shell: bash
working-directory: ${{github.workspace}}
run: |
export PYTHONPATH=$PWD:$PYTHONPATH
echo $PYTHONPATH
cd egs/yesno/ASR
./prepare.sh
python3 ./tdnn/train.py
python3 ./tdnn/decode.py
- name: Test exporting to pretrained.pt
shell: bash
working-directory: ${{github.workspace}}
run: |
export PYTHONPATH=$PWD:$PYTHONPATH
echo $PYTHONPATH
cd egs/yesno/ASR
python3 ./tdnn/export.py --epoch 14 --avg 2
python3 ./tdnn/pretrained.py \
--checkpoint ./tdnn/exp/pretrained.pt \
--HLG ./data/lang_phone/HLG.pt \
--words-file ./data/lang_phone/words.txt \
download/waves_yesno/0_0_0_1_0_0_0_1.wav \
download/waves_yesno/0_0_1_0_0_0_1_0.wav
- name: Test exporting to torchscript
shell: bash
working-directory: ${{github.workspace}}
run: |
export PYTHONPATH=$PWD:$PYTHONPATH
echo $PYTHONPATH
cd egs/yesno/ASR
python3 ./tdnn/export.py --epoch 14 --avg 2 --jit 1
python3 ./tdnn/jit_pretrained.py \
--nn-model ./tdnn/exp/cpu_jit.pt \
--HLG ./data/lang_phone/HLG.pt \
--words-file ./data/lang_phone/words.txt \
download/waves_yesno/0_0_0_1_0_0_0_1.wav \
download/waves_yesno/0_0_1_0_0_0_1_0.wav
- name: Test exporting to onnx
shell: bash
working-directory: ${{github.workspace}}
run: |
export PYTHONPATH=$PWD:$PYTHONPATH
echo $PYTHONPATH
cd egs/yesno/ASR
python3 ./tdnn/export_onnx.py --epoch 14 --avg 2
echo "Test float32 model"
python3 ./tdnn/onnx_pretrained.py \
--nn-model ./tdnn/exp/model-epoch-14-avg-2.onnx \
--HLG ./data/lang_phone/HLG.pt \
--words-file ./data/lang_phone/words.txt \
download/waves_yesno/0_0_0_1_0_0_0_1.wav \
download/waves_yesno/0_0_1_0_0_0_1_0.wav
echo "Test int8 model"
python3 ./tdnn/onnx_pretrained.py \
--nn-model ./tdnn/exp/model-epoch-14-avg-2.int8.onnx \
--HLG ./data/lang_phone/HLG.pt \
--words-file ./data/lang_phone/words.txt \
download/waves_yesno/0_0_0_1_0_0_0_1.wav \
download/waves_yesno/0_0_1_0_0_0_1_0.wav
- name: Test decoding with H
shell: bash
working-directory: ${{github.workspace}}
run: |
export PYTHONPATH=$PWD:$PYTHONPATH
echo $PYTHONPATH
cd egs/yesno/ASR
python3 ./tdnn/export.py --epoch 14 --avg 2 --jit 1
python3 ./tdnn/jit_pretrained_decode_with_H.py \
--nn-model ./tdnn/exp/cpu_jit.pt \
--H ./data/lang_phone/H.fst \
--tokens ./data/lang_phone/tokens.txt \
./download/waves_yesno/0_0_0_1_0_0_0_1.wav \
./download/waves_yesno/0_0_1_0_0_0_1_0.wav \
./download/waves_yesno/0_0_1_0_0_1_1_1.wav
- name: Test decoding with HL
shell: bash
working-directory: ${{github.workspace}}
run: |
export PYTHONPATH=$PWD:$PYTHONPATH
echo $PYTHONPATH
cd egs/yesno/ASR
python3 ./tdnn/export.py --epoch 14 --avg 2 --jit 1
python3 ./tdnn/jit_pretrained_decode_with_HL.py \
--nn-model ./tdnn/exp/cpu_jit.pt \
--HL ./data/lang_phone/HL.fst \
--words ./data/lang_phone/words.txt \
./download/waves_yesno/0_0_0_1_0_0_0_1.wav \
./download/waves_yesno/0_0_1_0_0_0_1_0.wav \
./download/waves_yesno/0_0_1_0_0_1_1_1.wav
- name: Show generated files
shell: bash
working-directory: ${{github.workspace}}
run: |
cd egs/yesno/ASR
ls -lh tdnn/exp
ls -lh data/lang_phone

8
.github/workflows/style_check.yml vendored
View File

@ -49,7 +49,7 @@ jobs:
- name: Install Python dependencies
run: |
python3 -m pip install --upgrade pip black==22.3.0 flake8==5.0.4 click==8.1.0
python3 -m pip install --upgrade pip black==22.3.0 flake8==5.0.4 click==8.1.0 isort==5.10.1
# Click issue fixed in https://github.com/psf/black/pull/2966
- name: Run flake8
@ -67,3 +67,9 @@ jobs:
working-directory: ${{github.workspace}}
run: |
black --check --diff .
- name: Run isort
shell: bash
working-directory: ${{github.workspace}}
run: |
isort --check --diff .

106
.github/workflows/test.yml vendored
View File

@ -1,95 +1,75 @@
# Copyright 2021 Fangjun Kuang (csukuangfj@gmail.com)
# See ../../LICENSE for clarification regarding multiple authors
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
name: test
on:
push:
branches:
- master
pull_request:
branches:
- master
workflow_dispatch:
concurrency:
group: test-${{ github.ref }}
cancel-in-progress: true
jobs:
generate_build_matrix:
if: github.repository_owner == 'csukuangfj' || github.repository_owner == 'k2-fsa'
# see https://github.com/pytorch/pytorch/pull/50633
runs-on: ubuntu-latest
outputs:
matrix: ${{ steps.set-matrix.outputs.matrix }}
steps:
- uses: actions/checkout@v4
with:
fetch-depth: 0
- name: Generating build matrix
id: set-matrix
run: |
# outputting for debugging purposes
python ./.github/scripts/docker/generate_build_matrix.py
MATRIX=$(python ./.github/scripts/docker/generate_build_matrix.py)
echo "::set-output name=matrix::${MATRIX}"
test:
runs-on: ${{ matrix.os }}
needs: generate_build_matrix
name: py${{ matrix.python-version }} torch${{ matrix.torch-version }} v${{ matrix.version }}
runs-on: ubuntu-latest
strategy:
matrix:
os: [ubuntu-latest]
python-version: ["3.8"]
torch: ["1.13.0"]
torchaudio: ["0.13.0"]
k2-version: ["1.24.3.dev20230719"]
fail-fast: false
matrix:
${{ fromJson(needs.generate_build_matrix.outputs.matrix) }}
steps:
- uses: actions/checkout@v2
- uses: actions/checkout@v4
with:
fetch-depth: 0
- name: Setup Python ${{ matrix.python-version }}
uses: actions/setup-python@v1
with:
python-version: ${{ matrix.python-version }}
- name: Install libnsdfile and libsox
if: startsWith(matrix.os, 'ubuntu')
run: |
sudo apt update
sudo apt install -q -y libsndfile1-dev libsndfile1 ffmpeg
sudo apt install -q -y --fix-missing libsox-dev libsox-fmt-all
- name: Install Python dependencies
run: |
python3 -m pip install --upgrade pip pytest
# numpy 1.20.x does not support python 3.6
pip install numpy==1.19
pip install torch==${{ matrix.torch }}+cpu -f https://download.pytorch.org/whl/cpu/torch_stable.html
pip install torchaudio==${{ matrix.torchaudio }}+cpu -f https://download.pytorch.org/whl/cpu/torch_stable.html
pip install k2==${{ matrix.k2-version }}+cpu.torch${{ matrix.torch }} -f https://k2-fsa.github.io/k2/cpu.html
pip install git+https://github.com/lhotse-speech/lhotse
# icefall requirements
pip uninstall -y protobuf
pip install --no-binary protobuf protobuf==3.20.*
pip install kaldifst
pip install onnxruntime matplotlib
pip install -r requirements.txt
- name: Install graphviz
if: startsWith(matrix.os, 'ubuntu')
- name: Free space
shell: bash
run: |
python3 -m pip install -qq graphviz
sudo apt-get -qq install graphviz
df -h
rm -rf /opt/hostedtoolcache
df -h
echo "pwd: $PWD"
echo "github.workspace ${{ github.workspace }}"
- name: Run tests
if: startsWith(matrix.os, 'ubuntu')
uses: addnab/docker-run-action@v3
with:
image: ghcr.io/${{ github.repository_owner }}/icefall:cpu-py${{ matrix.python-version }}-torch${{ matrix.torch-version }}-v${{ matrix.version }}
options: |
--volume ${{ github.workspace }}/:/icefall
shell: bash
run: |
ls -lh
export PYTHONPATH=$PWD:$PWD/lhotse:$PYTHONPATH
echo $PYTHONPATH
export PYTHONPATH=/icefall:$PYTHONPATH
cd /icefall
git config --global --add safe.directory /icefall
pytest -v -s ./test
# runt tests for conformer ctc
cd egs/librispeech/ASR/conformer_ctc
pytest -v -s

65
.github/workflows/yesno.yml vendored Normal file
View File

@ -0,0 +1,65 @@
name: yesno
on:
push:
branches:
- master
pull_request:
branches:
- master
workflow_dispatch:
concurrency:
group: yesno-${{ github.ref }}
cancel-in-progress: true
jobs:
generate_build_matrix:
if: github.repository_owner == 'csukuangfj' || github.repository_owner == 'k2-fsa'
# see https://github.com/pytorch/pytorch/pull/50633
runs-on: ubuntu-latest
outputs:
matrix: ${{ steps.set-matrix.outputs.matrix }}
steps:
- uses: actions/checkout@v4
with:
fetch-depth: 0
- name: Generating build matrix
id: set-matrix
run: |
# outputting for debugging purposes
python ./.github/scripts/docker/generate_build_matrix.py
MATRIX=$(python ./.github/scripts/docker/generate_build_matrix.py)
echo "::set-output name=matrix::${MATRIX}"
yesno:
needs: generate_build_matrix
name: py${{ matrix.python-version }} torch${{ matrix.torch-version }} v${{ matrix.version }}
runs-on: ubuntu-latest
strategy:
fail-fast: false
matrix:
${{ fromJson(needs.generate_build_matrix.outputs.matrix) }}
steps:
- uses: actions/checkout@v4
with:
fetch-depth: 0
- name: Run the yesno recipe
uses: addnab/docker-run-action@v3
with:
image: ghcr.io/${{ github.repository_owner }}/icefall:cpu-py${{ matrix.python-version }}-torch${{ matrix.torch-version }}-v${{ matrix.version }}
options: |
--volume ${{ github.workspace }}/:/icefall
shell: bash
run: |
export PYTHONPATH=/icefall:$PYTHONPATH
cd /icefall
git config --global --add safe.directory /icefall
python3 -m torch.utils.collect_env
python3 -m k2.version
.github/scripts/yesno/ASR/run.sh

2
.pre-commit-config.yaml
View File

@ -26,7 +26,7 @@ repos:
# E121,E123,E126,E226,E24,E704,W503,W504
- repo: https://github.com/pycqa/isort
rev: 5.11.5
rev: 5.12.0
hooks:
- id: isort
args: ["--profile=black"]

426
README.md
View File

@ -2,46 +2,86 @@
<img src="https://raw.githubusercontent.com/k2-fsa/icefall/master/docs/source/_static/logo.png" width=168>
</div>
## Introduction
# Introduction
icefall contains ASR recipes for various datasets
using <https://github.com/k2-fsa/k2>.
The icefall project contains speech-related recipes for various datasets
using [k2-fsa](https://github.com/k2-fsa/k2) and [lhotse](https://github.com/lhotse-speech/lhotse).
You can use <https://github.com/k2-fsa/sherpa> to deploy models
trained with icefall.
You can use [sherpa](https://github.com/k2-fsa/sherpa), [sherpa-ncnn](https://github.com/k2-fsa/sherpa-ncnn) or [sherpa-onnx](https://github.com/k2-fsa/sherpa-onnx) for deployment with models
in icefall; these frameworks also support models not included in icefall; please refer to respective documents for more details.
You can try pre-trained models from within your browser without the need
to download or install anything by visiting <https://huggingface.co/spaces/k2-fsa/automatic-speech-recognition>
See <https://k2-fsa.github.io/icefall/huggingface/spaces.html> for more details.
to download or install anything by visiting this [huggingface space](https://huggingface.co/spaces/k2-fsa/automatic-speech-recognition).
Please refer to [document](https://k2-fsa.github.io/icefall/huggingface/spaces.html) for more details.
## Installation
# Installation
Please refer to <https://icefall.readthedocs.io/en/latest/installation/index.html>
Please refer to [document](https://icefall.readthedocs.io/en/latest/installation/index.html)
for installation.
## Recipes
# Recipes
Please refer to <https://icefall.readthedocs.io/en/latest/recipes/index.html>
for more information.
Please refer to [document](https://icefall.readthedocs.io/en/latest/recipes/index.html)
for more details.
We provide the following recipes:
## ASR: Automatic Speech Recognition
### Supported Datasets
- [yesno][yesno]
- [LibriSpeech][librispeech]
- [GigaSpeech][gigaspeech]
- [AMI][ami]
- [Aidatatang_200zh][aidatatang_200zh]
- [Aishell][aishell]
- [Aishell2][aishell2]
- [Aishell4][aishell4]
- [Alimeeting][alimeeting]
- [AMI][ami]
- [CommonVoice][commonvoice]
- [Corpus of Spontaneous Japanese][csj]
- [GigaSpeech][gigaspeech]
- [LibriCSS][libricss]
- [LibriSpeech][librispeech]
- [Libriheavy][libriheavy]
- [Multi-Dialect Broadcast News Arabic Speech Recognition][mgb2]
- [PeopleSpeech][peoplespeech]
- [SPGISpeech][spgispeech]
- [Switchboard][swbd]
- [TIMIT][timit]
- [TED-LIUM3][tedlium3]
- [Aidatatang_200zh][aidatatang_200zh]
- [WenetSpeech][wenetspeech]
- [Alimeeting][alimeeting]
- [Switchboard][swbd]
- [TAL_CSASR][tal_csasr]
- [Voxpopuli][voxpopuli]
- [XBMU-AMDO31][xbmu-amdo31]
- [WenetSpeech][wenetspeech]
### yesno
More datasets will be added in the future.
### Supported Models
The [LibriSpeech][librispeech] recipe supports the most comprehensive set of models, you are welcome to try them out.
#### CTC
- TDNN LSTM CTC
- Conformer CTC
- Zipformer CTC
#### MMI
- Conformer MMI
- Zipformer MMI
#### Transducer
- Conformer-based Encoder
- LSTM-based Encoder
- Zipformer-based Encoder
- LSTM-based Predictor
- [Stateless Predictor](https://research.google/pubs/rnn-transducer-with-stateless-prediction-network/)
#### Whisper
- [OpenAi Whisper](https://arxiv.org/abs/2212.04356) (We support fine-tuning on AiShell-1.)
If you are willing to contribute to icefall, please refer to [contributing](https://icefall.readthedocs.io/en/latest/contributing/index.html) for more details.
We would like to highlight the performance of some of the recipes here.
### [yesno][yesno]
This is the simplest ASR recipe in `icefall` and can be run on CPU.
Training takes less than 30 seconds and gives you the following WER:
@ -52,350 +92,264 @@ Training takes less than 30 seconds and gives you the following WER:
We provide a Colab notebook for this recipe: [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1tIjjzaJc3IvGyKiMCDWO-TSnBgkcuN3B?usp=sharing)
### LibriSpeech
### [LibriSpeech][librispeech]
Please see <https://github.com/k2-fsa/icefall/blob/master/egs/librispeech/ASR/RESULTS.md>
Please see [RESULTS.md](https://github.com/k2-fsa/icefall/blob/master/egs/librispeech/ASR/RESULTS.md)
for the **latest** results.
We provide 5 models for this recipe:
- [conformer CTC model][LibriSpeech_conformer_ctc]
- [TDNN LSTM CTC model][LibriSpeech_tdnn_lstm_ctc]
- [Transducer: Conformer encoder + LSTM decoder][LibriSpeech_transducer]
- [Transducer: Conformer encoder + Embedding decoder][LibriSpeech_transducer_stateless]
- [Transducer: Zipformer encoder + Embedding decoder][LibriSpeech_zipformer]
#### Conformer CTC Model
The best WER we currently have is:
#### [Conformer CTC](https://github.com/k2-fsa/icefall/tree/master/egs/librispeech/ASR/conformer_ctc)
| | test-clean | test-other |
|-----|------------|------------|
| WER | 2.42 | 5.73 |
We provide a Colab notebook to run a pre-trained conformer CTC model: [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1huyupXAcHsUrKaWfI83iMEJ6J0Nh0213?usp=sharing)
We provide a Colab notebook to test the pre-trained model: [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1huyupXAcHsUrKaWfI83iMEJ6J0Nh0213?usp=sharing)
#### TDNN LSTM CTC Model
The WER for this model is:
#### [TDNN LSTM CTC](https://github.com/k2-fsa/icefall/tree/master/egs/librispeech/ASR/tdnn_lstm_ctc)
| | test-clean | test-other |
|-----|------------|------------|
| WER | 6.59 | 17.69 |
We provide a Colab notebook to run a pre-trained TDNN LSTM CTC model: [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1-iSfQMp2So-We_Uu49N4AAcMInB72u9z?usp=sharing)
We provide a Colab notebook to test the pre-trained model: [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1-iSfQMp2So-We_Uu49N4AAcMInB72u9z?usp=sharing)
#### Transducer: Conformer encoder + LSTM decoder
Using Conformer as encoder and LSTM as decoder.
The best WER with greedy search is:
#### [Transducer (Conformer Encoder + LSTM Predictor)](https://github.com/k2-fsa/icefall/tree/master/egs/librispeech/ASR/transducer)
| | test-clean | test-other |
|-----|------------|------------|
| WER | 3.07 | 7.51 |
|---------------|------------|------------|
| greedy_search | 3.07 | 7.51 |
We provide a Colab notebook to run a pre-trained RNN-T conformer model: [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1_u6yK9jDkPwG_NLrZMN2XK7Aeq4suMO2?usp=sharing)
We provide a Colab notebook to test the pre-trained model: [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1_u6yK9jDkPwG_NLrZMN2XK7Aeq4suMO2?usp=sharing)
#### Transducer: Conformer encoder + Embedding decoder
Using Conformer as encoder. The decoder consists of 1 embedding layer
and 1 convolutional layer.
The best WER using modified beam search with beam size 4 is:
#### [Transducer (Conformer Encoder + Stateless Predictor)](https://github.com/k2-fsa/icefall/tree/master/egs/librispeech/ASR/transducer)
| | test-clean | test-other |
|-----|------------|------------|
| WER | 2.56 | 6.27 |
Note: No auxiliary losses are used in the training and no LMs are used
in the decoding.
We provide a Colab notebook to run a pre-trained transducer conformer + stateless decoder model: [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1CO1bXJ-2khDckZIW8zjOPHGSKLHpTDlp?usp=sharing)
|---------------------------------------|------------|------------|
| modified_beam_search (`beam_size=4`) | 2.56 | 6.27 |
#### k2 pruned RNN-T
We provide a Colab notebook to test the pre-trained model: [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1CO1bXJ-2khDckZIW8zjOPHGSKLHpTDlp?usp=sharing)
#### [Transducer (Zipformer Encoder + Stateless Predictor)](https://github.com/k2-fsa/icefall/tree/master/egs/librispeech/ASR/zipformer)
WER (modified_beam_search `beam_size=4` unless further stated)
1. LibriSpeech-960hr
| Encoder | Params | test-clean | test-other | epochs | devices |
|-----------------|--------|------------|------------|---------|------------|
| zipformer | 65.5M | 2.21 | 4.79 | 50 | 4 32G-V100 |
| zipformer-small | 23.2M | 2.42 | 5.73 | 50 | 2 32G-V100 |
| zipformer-large | 148.4M | 2.06 | 4.63 | 50 | 4 32G-V100 |
| zipformer-large | 148.4M | 2.00 | 4.38 | 174 | 8 80G-A100 |
| Zipformer | 65.5M | 2.21 | 4.79 | 50 | 4 32G-V100 |
| Zipformer-small | 23.2M | 2.42 | 5.73 | 50 | 2 32G-V100 |
| Zipformer-large | 148.4M | 2.06 | 4.63 | 50 | 4 32G-V100 |
| Zipformer-large | 148.4M | 2.00 | 4.38 | 174 | 8 80G-A100 |
Note: No auxiliary losses are used in the training and no LMs are used
in the decoding.
2. LibriSpeech-960hr + GigaSpeech
#### k2 pruned RNN-T + GigaSpeech
| | test-clean | test-other |
|-----|------------|------------|
| WER | 1.78 | 4.08 |
Note: No auxiliary losses are used in the training and no LMs are used
in the decoding.
#### k2 pruned RNN-T + GigaSpeech + CommonVoice
| | test-clean | test-other |
|-----|------------|------------|
| WER | 1.90 | 3.98 |
Note: No auxiliary losses are used in the training and no LMs are used
in the decoding.
| Encoder | Params | test-clean | test-other |
|-----------------|--------|------------|------------|
| Zipformer | 65.5M | 1.78 | 4.08 |
### GigaSpeech
3. LibriSpeech-960hr + GigaSpeech + CommonVoice
We provide three models for this recipe:
| Encoder | Params | test-clean | test-other |
|-----------------|--------|------------|------------|
| Zipformer | 65.5M | 1.90 | 3.98 |
- [Conformer CTC model][GigaSpeech_conformer_ctc]
- [Pruned stateless RNN-T: Conformer encoder + Embedding decoder + k2 pruned RNN-T loss][GigaSpeech_pruned_transducer_stateless2].
- [Transducer: Zipformer encoder + Embedding decoder][GigaSpeech_zipformer]
#### Conformer CTC
### [GigaSpeech][gigaspeech]
#### [Conformer CTC](https://github.com/k2-fsa/icefall/tree/master/egs/gigaspeech/ASR/conformer_ctc)
| | Dev | Test |
|-----|-------|-------|
| WER | 10.47 | 10.58 |
#### Pruned stateless RNN-T: Conformer encoder + Embedding decoder + k2 pruned RNN-T loss
#### [Transducer (pruned_transducer_stateless2)](https://github.com/k2-fsa/icefall/tree/master/egs/gigaspeech/ASR/pruned_transducer_stateless2)
Conformer Encoder + Stateless Predictor + k2 Pruned RNN-T Loss
| | Dev | Test |
|----------------------|-------|-------|
| greedy search | 10.51 | 10.73 |
| fast beam search | 10.50 | 10.69 |
| modified beam search | 10.40 | 10.51 |
| greedy_search | 10.51 | 10.73 |
| fast_beam_search | 10.50 | 10.69 |
| modified_beam_search | 10.40 | 10.51 |
#### Transducer: Zipformer encoder + Embedding decoder
#### [Transducer (Zipformer Encoder + Stateless Predictor)](https://github.com/k2-fsa/icefall/tree/master/egs/gigaspeech/ASR/zipformer)
| | Dev | Test |
|----------------------|-------|-------|
| greedy search | 10.31 | 10.50 |
| fast beam search | 10.26 | 10.48 |
| modified beam search | 10.25 | 10.38 |
| greedy_search | 10.31 | 10.50 |
| fast_beam_search | 10.26 | 10.48 |
| modified_beam_search | 10.25 | 10.38 |
### Aishell
### [Aishell][aishell]
We provide three models for this recipe: [conformer CTC model][Aishell_conformer_ctc],
[TDNN LSTM CTC model][Aishell_tdnn_lstm_ctc], and [Transducer Stateless Model][Aishell_pruned_transducer_stateless7],
#### Conformer CTC Model
The best CER we currently have is:
| | test |
|-----|------|
| CER | 4.26 |
#### TDNN LSTM CTC Model
The CER for this model is:
#### [TDNN LSTM CTC](https://github.com/k2-fsa/icefall/tree/master/egs/aishell/ASR/tdnn_lstm_ctc)
| | test |
|-----|-------|
| CER | 10.16 |
We provide a Colab notebook to run a pre-trained TDNN LSTM CTC model: [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1jbyzYq3ytm6j2nlEt-diQm-6QVWyDDEa?usp=sharing)
We provide a Colab notebook to test the pre-trained model: [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1jbyzYq3ytm6j2nlEt-diQm-6QVWyDDEa?usp=sharing)
#### Transducer Stateless Model
The best CER we currently have is:
#### [Transducer (Conformer Encoder + Stateless Predictor)](https://github.com/k2-fsa/icefall/tree/master/egs/aishell/ASR/transducer_stateless)
| | test |
|-----|------|
| CER | 4.38 |
We provide a Colab notebook to run a pre-trained TransducerStateless model: [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/14XaT2MhnBkK-3_RqqWq3K90Xlbin-GZC?usp=sharing)
We provide a Colab notebook to test the pre-trained model: [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/14XaT2MhnBkK-3_RqqWq3K90Xlbin-GZC?usp=sharing)
#### [Transducer (Zipformer Encoder + Stateless Predictor)](https://github.com/k2-fsa/icefall/tree/master/egs/aishell/ASR/zipformer)
WER (modified_beam_search `beam_size=4`)
| Encoder | Params | dev | test | epochs |
|-----------------|--------|-----|------|---------|
| Zipformer | 73.4M | 4.13| 4.40 | 55 |
| Zipformer-small | 30.2M | 4.40| 4.67 | 55 |
| Zipformer-large | 157.3M | 4.03| 4.28 | 56 |
### Aishell2
### [Aishell4][aishell4]
We provide one model for this recipe: [Transducer Stateless Model][Aishell2_pruned_transducer_stateless5].
#### Transducer Stateless Model
The best WER we currently have is:
| | dev-ios | test-ios |
|-----|------------|------------|
| WER | 5.32 | 5.56 |
### Aishell4
We provide one model for this recipe: [Pruned stateless RNN-T: Conformer encoder + Embedding decoder + k2 pruned RNN-T loss][Aishell4_pruned_transducer_stateless5].
#### Pruned stateless RNN-T: Conformer encoder + Embedding decoder + k2 pruned RNN-T loss (trained with all subsets)
The best CER we currently have is:
#### [Transducer (pruned_transducer_stateless5)](https://github.com/k2-fsa/icefall/tree/master/egs/aishell4/ASR/pruned_transducer_stateless5)
1 Trained with all subsets:
| | test |
|-----|------------|
| CER | 29.08 |
We provide a Colab notebook to run a pre-trained Pruned Transducer Stateless model: [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1z3lkURVv9M7uTiIgf3Np9IntMHEknaks?usp=sharing)
We provide a Colab notebook to test the pre-trained model: [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1z3lkURVv9M7uTiIgf3Np9IntMHEknaks?usp=sharing)
### TIMIT
### [TIMIT][timit]
We provide two models for this recipe: [TDNN LSTM CTC model][TIMIT_tdnn_lstm_ctc]
and [TDNN LiGRU CTC model][TIMIT_tdnn_ligru_ctc].
#### [TDNN LSTM CTC](https://github.com/k2-fsa/icefall/tree/master/egs/timit/ASR/tdnn_lstm_ctc)
#### TDNN LSTM CTC Model
The best PER we currently have is:
||TEST|
|--|--|
| |TEST|
|---|----|
|PER| 19.71% |
We provide a Colab notebook to run a pre-trained TDNN LSTM CTC model: [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1Hs9DA4V96uapw_30uNp32OMJgkuR5VVd?usp=sharing)
We provide a Colab notebook to test the pre-trained model: [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1Hs9DA4V96uapw_30uNp32OMJgkuR5VVd?usp=sharing)
#### TDNN LiGRU CTC Model
#### [TDNN LiGRU CTC](https://github.com/k2-fsa/icefall/tree/master/egs/timit/ASR/tdnn_ligru_ctc)
The PER for this model is:
||TEST|
|--|--|
| |TEST|
|---|----|
|PER| 17.66% |
We provide a Colab notebook to run a pre-trained TDNN LiGRU CTC model: [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1z3lkURVv9M7uTiIgf3Np9IntMHEknaks?usp=sharing)
We provide a Colab notebook to test the pre-trained model: [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1z3lkURVv9M7uTiIgf3Np9IntMHEknaks?usp=sharing)
### TED-LIUM3
### [TED-LIUM3][tedlium3]
We provide two models for this recipe: [Transducer Stateless: Conformer encoder + Embedding decoder][TED-LIUM3_transducer_stateless] and [Pruned Transducer Stateless: Conformer encoder + Embedding decoder + k2 pruned RNN-T loss][TED-LIUM3_pruned_transducer_stateless].
#### Transducer Stateless: Conformer encoder + Embedding decoder
The best WER using modified beam search with beam size 4 is:
#### [Transducer (Conformer Encoder + Stateless Predictor)](https://github.com/k2-fsa/icefall/tree/master/egs/tedlium3/ASR/transducer_stateless)
| | dev | test |
|-----|-------|--------|
| WER | 6.91 | 6.33 |
|--------------------------------------|-------|--------|
| modified_beam_search (`beam_size=4`) | 6.91 | 6.33 |
Note: No auxiliary losses are used in the training and no LMs are used in the decoding.
We provide a Colab notebook to run a pre-trained Transducer Stateless model: [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1MmY5bBxwvKLNT4A2DJnwiqRXhdchUqPN?usp=sharing)
We provide a Colab notebook to test the pre-trained model: [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1MmY5bBxwvKLNT4A2DJnwiqRXhdchUqPN?usp=sharing)
#### Pruned Transducer Stateless: Conformer encoder + Embedding decoder + k2 pruned RNN-T loss
The best WER using modified beam search with beam size 4 is:
#### [Transducer (pruned_transducer_stateless)](https://github.com/k2-fsa/icefall/tree/master/egs/tedlium3/ASR/pruned_transducer_stateless)
| | dev | test |
|-----|-------|--------|
| WER | 6.77 | 6.14 |
|--------------------------------------|-------|--------|
| modified_beam_search (`beam_size=4`) | 6.77 | 6.14 |
We provide a Colab notebook to run a pre-trained Pruned Transducer Stateless model: [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1je_1zGrOkGVVd4WLzgkXRHxl-I27yWtz?usp=sharing)
We provide a Colab notebook to test the pre-trained model: [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1je_1zGrOkGVVd4WLzgkXRHxl-I27yWtz?usp=sharing)
### Aidatatang_200zh
### [Aidatatang_200zh][aidatatang_200zh]
We provide one model for this recipe: [Pruned stateless RNN-T: Conformer encoder + Embedding decoder + k2 pruned RNN-T loss][Aidatatang_200zh_pruned_transducer_stateless2].
#### Pruned stateless RNN-T: Conformer encoder + Embedding decoder + k2 pruned RNN-T loss
#### [Transducer (pruned_transducer_stateless2)](https://github.com/k2-fsa/icefall/tree/master/egs/aidatatang_200zh/ASR/pruned_transducer_stateless2)
| | Dev | Test |
|----------------------|-------|-------|
| greedy search | 5.53 | 6.59 |
| fast beam search | 5.30 | 6.34 |
| modified beam search | 5.27 | 6.33 |
| greedy_search | 5.53 | 6.59 |
| fast_beam_search | 5.30 | 6.34 |
| modified_beam_search | 5.27 | 6.33 |
We provide a Colab notebook to run a pre-trained Pruned Transducer Stateless model: [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1wNSnSj3T5oOctbh5IGCa393gKOoQw2GH?usp=sharing)
We provide a Colab notebook to test the pre-trained model: [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1wNSnSj3T5oOctbh5IGCa393gKOoQw2GH?usp=sharing)
### WenetSpeech
### [WenetSpeech][wenetspeech]
We provide some models for this recipe: [Pruned stateless RNN-T_2: Conformer encoder + Embedding decoder + k2 pruned RNN-T loss][WenetSpeech_pruned_transducer_stateless2] and [Pruned stateless RNN-T_5: Conformer encoder + Embedding decoder + k2 pruned RNN-T loss][WenetSpeech_pruned_transducer_stateless5].
#### Pruned stateless RNN-T_2: Conformer encoder + Embedding decoder + k2 pruned RNN-T loss (trained with L subset, offline ASR)
#### [Transducer (pruned_transducer_stateless2)](https://github.com/k2-fsa/icefall/tree/master/egs/wenetspeech/ASR/pruned_transducer_stateless2)
| | Dev | Test-Net | Test-Meeting |
|----------------------|-------|----------|--------------|
| greedy search | 7.80 | 8.75 | 13.49 |
| modified beam search| 7.76 | 8.71 | 13.41 |
| fast beam search | 7.94 | 8.74 | 13.80 |
| greedy_search | 7.80 | 8.75 | 13.49 |
| fast_beam_search | 7.94 | 8.74 | 13.80 |
| modified_beam_search | 7.76 | 8.71 | 13.41 |
We provide a Colab notebook to test the pre-trained model: [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1EV4e1CHa1GZgEF-bZgizqI9RyFFehIiN?usp=sharing)
#### [Transducer **Streaming** (pruned_transducer_stateless5) ](https://github.com/k2-fsa/icefall/tree/master/egs/wenetspeech/ASR/pruned_transducer_stateless5)
#### Pruned stateless RNN-T_5: Conformer encoder + Embedding decoder + k2 pruned RNN-T loss (trained with L subset)
**Streaming**:
| | Dev | Test-Net | Test-Meeting |
|----------------------|-------|----------|--------------|
| greedy_search | 8.78 | 10.12 | 16.16 |
| modified_beam_search | 8.53| 9.95 | 15.81 |
| fast_beam_search| 9.01 | 10.47 | 16.28 |
| modified_beam_search | 8.53| 9.95 | 15.81 |
We provide a Colab notebook to run a pre-trained Pruned Transducer Stateless2 model: [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1EV4e1CHa1GZgEF-bZgizqI9RyFFehIiN?usp=sharing)
### Alimeeting
### [Alimeeting][alimeeting]
We provide one model for this recipe: [Pruned stateless RNN-T: Conformer encoder + Embedding decoder + k2 pruned RNN-T loss][Alimeeting_pruned_transducer_stateless2].
#### Pruned stateless RNN-T: Conformer encoder + Embedding decoder + k2 pruned RNN-T loss (trained with far subset)
#### [Transducer (pruned_transducer_stateless2)](https://github.com/k2-fsa/icefall/tree/master/egs/alimeeting/ASR/pruned_transducer_stateless2)
| | Eval | Test-Net |
|----------------------|--------|----------|
| greedy search | 31.77 | 34.66 |
| fast beam search | 31.39 | 33.02 |
| modified beam search | 30.38 | 34.25 |
| greedy_search | 31.77 | 34.66 |
| fast_beam_search | 31.39 | 33.02 |
| modified_beam_search | 30.38 | 34.25 |
We provide a Colab notebook to run a pre-trained Pruned Transducer Stateless model: [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1tKr3f0mL17uO_ljdHGKtR7HOmthYHwJG?usp=sharing)
We provide a Colab notebook to test the pre-trained model: [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1tKr3f0mL17uO_ljdHGKtR7HOmthYHwJG?usp=sharing)
### TAL_CSASR
### [TAL_CSASR][tal_csasr]
We provide one model for this recipe: [Pruned stateless RNN-T: Conformer encoder + Embedding decoder + k2 pruned RNN-T loss][TAL_CSASR_pruned_transducer_stateless5].
#### Pruned stateless RNN-T: Conformer encoder + Embedding decoder + k2 pruned RNN-T loss
#### [Transducer (pruned_transducer_stateless5)](https://github.com/k2-fsa/icefall/tree/master/egs/tal_csasr/ASR/pruned_transducer_stateless5)
The best results for Chinese CER(%) and English WER(%) respectively (zh: Chinese, en: English):
|decoding-method | dev | dev_zh | dev_en | test | test_zh | test_en |
|--|--|--|--|--|--|--|
|greedy_search| 7.30 | 6.48 | 19.19 |7.39| 6.66 | 19.13|
|modified_beam_search| 7.15 | 6.35 | 18.95 | 7.22| 6.50 | 18.70 |
|fast_beam_search| 7.18 | 6.39| 18.90 | 7.27| 6.55 | 18.77|
|modified_beam_search| 7.15 | 6.35 | 18.95 | 7.22| 6.50 | 18.70 |
We provide a Colab notebook to run a pre-trained Pruned Transducer Stateless model: [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1DmIx-NloI1CMU5GdZrlse7TRu4y3Dpf8?usp=sharing)
We provide a Colab notebook to test the pre-trained model: [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1DmIx-NloI1CMU5GdZrlse7TRu4y3Dpf8?usp=sharing)
## Deployment with C++
## TTS: Text-to-Speech
Once you have trained a model in icefall, you may want to deploy it with C++,
without Python dependencies.
### Supported Datasets
Please refer to the documentation
<https://icefall.readthedocs.io/en/latest/recipes/Non-streaming-ASR/librispeech/conformer_ctc.html#deployment-with-c>
- [LJSpeech][ljspeech]
- [VCTK][vctk]
### Supported Models
- [VITS](https://arxiv.org/abs/2106.06103)
# Deployment with C++
Once you have trained a model in icefall, you may want to deploy it with C++ without Python dependencies.
Please refer to the [document](https://icefall.readthedocs.io/en/latest/recipes/Non-streaming-ASR/librispeech/conformer_ctc.html#deployment-with-c)
for how to do this.
We also provide a Colab notebook, showing you how to run a torch scripted model in [k2][k2] with C++.
Please see: [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1BIGLWzS36isskMXHKcqC9ysN6pspYXs_?usp=sharing)
[LibriSpeech_tdnn_lstm_ctc]: egs/librispeech/ASR/tdnn_lstm_ctc
[LibriSpeech_conformer_ctc]: egs/librispeech/ASR/conformer_ctc
[LibriSpeech_transducer]: egs/librispeech/ASR/transducer
[LibriSpeech_transducer_stateless]: egs/librispeech/ASR/transducer_stateless
[LibriSpeech_zipformer]: egs/librispeech/ASR/zipformer
[Aishell_tdnn_lstm_ctc]: egs/aishell/ASR/tdnn_lstm_ctc
[Aishell_conformer_ctc]: egs/aishell/ASR/conformer_ctc
[Aishell_pruned_transducer_stateless7]: egs/aishell/ASR/pruned_transducer_stateless7_bbpe
[Aishell2_pruned_transducer_stateless5]: egs/aishell2/ASR/pruned_transducer_stateless5
[Aishell4_pruned_transducer_stateless5]: egs/aishell4/ASR/pruned_transducer_stateless5
[TIMIT_tdnn_lstm_ctc]: egs/timit/ASR/tdnn_lstm_ctc
[TIMIT_tdnn_ligru_ctc]: egs/timit/ASR/tdnn_ligru_ctc
[TED-LIUM3_transducer_stateless]: egs/tedlium3/ASR/transducer_stateless
[TED-LIUM3_pruned_transducer_stateless]: egs/tedlium3/ASR/pruned_transducer_stateless
[GigaSpeech_conformer_ctc]: egs/gigaspeech/ASR/conformer_ctc
[GigaSpeech_pruned_transducer_stateless2]: egs/gigaspeech/ASR/pruned_transducer_stateless2
[GigaSpeech_zipformer]: egs/gigaspeech/ASR/zipformer
[Aidatatang_200zh_pruned_transducer_stateless2]: egs/aidatatang_200zh/ASR/pruned_transducer_stateless2
[WenetSpeech_pruned_transducer_stateless2]: egs/wenetspeech/ASR/pruned_transducer_stateless2
[WenetSpeech_pruned_transducer_stateless5]: egs/wenetspeech/ASR/pruned_transducer_stateless5
[Alimeeting_pruned_transducer_stateless2]: egs/alimeeting/ASR/pruned_transducer_stateless2
[TAL_CSASR_pruned_transducer_stateless5]: egs/tal_csasr/ASR/pruned_transducer_stateless5
[yesno]: egs/yesno/ASR
[librispeech]: egs/librispeech/ASR
[aishell]: egs/aishell/ASR
@ -411,3 +365,15 @@ Please see: [![Open In Colab](https://colab.research.google.com/assets/colab-bad
[ami]: egs/ami
[swbd]: egs/swbd/ASR
[k2]: https://github.com/k2-fsa/k2
[commonvoice]: egs/commonvoice/ASR
[csj]: egs/csj/ASR
[libricss]: egs/libricss/SURT
[libriheavy]: egs/libriheavy/ASR
[mgb2]: egs/mgb2/ASR
[peoplespeech]: egs/peoplespeech/ASR
[spgispeech]: egs/spgispeech/ASR
[voxpopuli]: egs/voxpopuli/ASR
[xbmu-amdo31]: egs/xbmu-amdo31/ASR
[vctk]: egs/vctk/TTS
[ljspeech]: egs/ljspeech/TTS

7
docker/torch1.12.1-cuda11.3.dockerfile
View File

@ -5,8 +5,8 @@ ENV LC_ALL C.UTF-8
ARG DEBIAN_FRONTEND=noninteractive
# python 3.7
ARG K2_VERSION="1.24.4.dev20230725+cuda11.3.torch1.12.1"
ARG KALDIFEAT_VERSION="1.25.1.dev20231022+cuda11.3.torch1.12.1"
ARG K2_VERSION="1.24.4.dev20240223+cuda11.3.torch1.12.1"
ARG KALDIFEAT_VERSION="1.25.4.dev20240223+cuda11.3.torch1.12.1"
ARG TORCHAUDIO_VERSION="0.12.1+cu113"
LABEL authors="Fangjun Kuang <csukuangfj@gmail.com>"
@ -44,7 +44,6 @@ RUN pip install --no-cache-dir \
k2==${K2_VERSION} -f https://k2-fsa.github.io/k2/cuda.html \
git+https://github.com/lhotse-speech/lhotse \
kaldifeat==${KALDIFEAT_VERSION} -f https://csukuangfj.github.io/kaldifeat/cuda.html \
\
kaldi_native_io \
kaldialign \
kaldifst \
@ -56,6 +55,8 @@ RUN pip install --no-cache-dir \
onnx \
onnxruntime \
onnxmltools \
onnxoptimizer \
onnxsim \
multi_quantization \
typeguard \
numpy \

7
docker/torch1.13.0-cuda11.6.dockerfile
View File

@ -5,8 +5,8 @@ ENV LC_ALL C.UTF-8
ARG DEBIAN_FRONTEND=noninteractive
# python 3.9
ARG K2_VERSION="1.24.4.dev20231021+cuda11.6.torch1.13.0"
ARG KALDIFEAT_VERSION="1.25.1.dev20231022+cuda11.6.torch1.13.0"
ARG K2_VERSION="1.24.4.dev20240223+cuda11.6.torch1.13.0"
ARG KALDIFEAT_VERSION="1.25.4.dev20240223+cuda11.6.torch1.13.0"
ARG TORCHAUDIO_VERSION="0.13.0+cu116"
LABEL authors="Fangjun Kuang <csukuangfj@gmail.com>"
@ -44,7 +44,6 @@ RUN pip install --no-cache-dir \
k2==${K2_VERSION} -f https://k2-fsa.github.io/k2/cuda.html \
git+https://github.com/lhotse-speech/lhotse \
kaldifeat==${KALDIFEAT_VERSION} -f https://csukuangfj.github.io/kaldifeat/cuda.html \
\
kaldi_native_io \
kaldialign \
kaldifst \
@ -56,6 +55,8 @@ RUN pip install --no-cache-dir \
onnx \
onnxruntime \
onnxmltools \
onnxoptimizer \
onnxsim \
multi_quantization \
typeguard \
numpy \

7
docker/torch1.9.0-cuda10.2.dockerfile
View File

@ -5,8 +5,8 @@ ENV LC_ALL C.UTF-8
ARG DEBIAN_FRONTEND=noninteractive
# python 3.7
ARG K2_VERSION="1.24.3.dev20230726+cuda10.2.torch1.9.0"
ARG KALDIFEAT_VERSION="1.25.1.dev20231022+cuda10.2.torch1.9.0"
ARG K2_VERSION="1.24.4.dev20240223+cuda10.2.torch1.9.0"
ARG KALDIFEAT_VERSION="1.25.4.dev20240223+cuda10.2.torch1.9.0"
ARG TORCHAUDIO_VERSION="0.9.0"
LABEL authors="Fangjun Kuang <csukuangfj@gmail.com>"
@ -58,7 +58,6 @@ RUN pip uninstall -y tqdm && \
k2==${K2_VERSION} -f https://k2-fsa.github.io/k2/cuda.html \
kaldifeat==${KALDIFEAT_VERSION} -f https://csukuangfj.github.io/kaldifeat/cuda.html \
git+https://github.com/lhotse-speech/lhotse \
\
kaldi_native_io \
kaldialign \
kaldifst \
@ -70,6 +69,8 @@ RUN pip uninstall -y tqdm && \
onnx \
onnxruntime \
onnxmltools \
onnxoptimizer \
onnxsim \
multi_quantization \
typeguard \
numpy \

8
docker/torch2.0.0-cuda11.7.dockerfile
View File

@ -1,12 +1,13 @@
FROM pytorch/pytorch:2.0.0-cuda11.7-cudnn8-devel
# python 3.10
ENV LC_ALL C.UTF-8
ARG DEBIAN_FRONTEND=noninteractive
# python 3.10
ARG K2_VERSION="1.24.4.dev20231021+cuda11.7.torch2.0.0"
ARG KALDIFEAT_VERSION="1.25.1.dev20231022+cuda11.7.torch2.0.0"
ARG K2_VERSION="1.24.4.dev20240223+cuda11.7.torch2.0.0"
ARG KALDIFEAT_VERSION="1.25.4.dev20240223+cuda11.7.torch2.0.0"
ARG TORCHAUDIO_VERSION="2.0.0+cu117"
LABEL authors="Fangjun Kuang <csukuangfj@gmail.com>"
@ -44,7 +45,6 @@ RUN pip install --no-cache-dir \
k2==${K2_VERSION} -f https://k2-fsa.github.io/k2/cuda.html \
git+https://github.com/lhotse-speech/lhotse \
kaldifeat==${KALDIFEAT_VERSION} -f https://csukuangfj.github.io/kaldifeat/cuda.html \
\
kaldi_native_io \
kaldialign \
kaldifst \
@ -56,6 +56,8 @@ RUN pip install --no-cache-dir \
onnx \
onnxruntime \
onnxmltools \
onnxoptimizer \
onnxsim \
multi_quantization \
typeguard \
numpy \

8
docker/torch2.1.0-cuda11.8.dockerfile
View File

@ -1,12 +1,13 @@
FROM pytorch/pytorch:2.1.0-cuda11.8-cudnn8-devel
# python 3.10
ENV LC_ALL C.UTF-8
ARG DEBIAN_FRONTEND=noninteractive
# python 3.10
ARG K2_VERSION="1.24.4.dev20231021+cuda11.8.torch2.1.0"
ARG KALDIFEAT_VERSION="1.25.1.dev20231022+cuda11.8.torch2.1.0"
ARG K2_VERSION="1.24.4.dev20240223+cuda11.8.torch2.1.0"
ARG KALDIFEAT_VERSION="1.25.4.dev20240223+cuda11.8.torch2.1.0"
ARG TORCHAUDIO_VERSION="2.1.0+cu118"
LABEL authors="Fangjun Kuang <csukuangfj@gmail.com>"
@ -44,7 +45,6 @@ RUN pip install --no-cache-dir \
k2==${K2_VERSION} -f https://k2-fsa.github.io/k2/cuda.html \
git+https://github.com/lhotse-speech/lhotse \
kaldifeat==${KALDIFEAT_VERSION} -f https://csukuangfj.github.io/kaldifeat/cuda.html \
\
kaldi_native_io \
kaldialign \
kaldifst \
@ -56,6 +56,8 @@ RUN pip install --no-cache-dir \
onnx \
onnxruntime \
onnxmltools \
onnxoptimizer \
onnxsim \
multi_quantization \
typeguard \
numpy \

8
docker/torch2.1.0-cuda12.1.dockerfile
View File

@ -1,12 +1,13 @@
FROM pytorch/pytorch:2.1.0-cuda12.1-cudnn8-devel
# python 3.10
ENV LC_ALL C.UTF-8
ARG DEBIAN_FRONTEND=noninteractive
# python 3.10
ARG K2_VERSION="1.24.4.dev20231021+cuda12.1.torch2.1.0"
ARG KALDIFEAT_VERSION="1.25.1.dev20231022+cuda12.1.torch2.1.0"
ARG K2_VERSION="1.24.4.dev20240223+cuda12.1.torch2.1.0"
ARG KALDIFEAT_VERSION="1.25.4.dev20240223+cuda12.1.torch2.1.0"
ARG TORCHAUDIO_VERSION="2.1.0+cu121"
LABEL authors="Fangjun Kuang <csukuangfj@gmail.com>"
@ -44,7 +45,6 @@ RUN pip install --no-cache-dir \
k2==${K2_VERSION} -f https://k2-fsa.github.io/k2/cuda.html \
git+https://github.com/lhotse-speech/lhotse \
kaldifeat==${KALDIFEAT_VERSION} -f https://csukuangfj.github.io/kaldifeat/cuda.html \
\
kaldi_native_io \
kaldialign \
kaldifst \
@ -56,6 +56,8 @@ RUN pip install --no-cache-dir \
onnx \
onnxruntime \
onnxmltools \
onnxoptimizer \
onnxsim \
multi_quantization \
typeguard \
numpy \

73
docker/torch2.2.0-cuda11.8.dockerfile Normal file
View File

@ -0,0 +1,73 @@
FROM pytorch/pytorch:2.2.0-cuda11.8-cudnn8-devel
# python 3.10
ENV LC_ALL C.UTF-8
ARG DEBIAN_FRONTEND=noninteractive
# python 3.10
ARG K2_VERSION="1.24.4.dev20240223+cuda11.8.torch2.2.0"
ARG KALDIFEAT_VERSION="1.25.4.dev20240223+cuda11.8.torch2.2.0"
ARG TORCHAUDIO_VERSION="2.2.0+cu118"
LABEL authors="Fangjun Kuang <csukuangfj@gmail.com>"
LABEL k2_version=${K2_VERSION}
LABEL kaldifeat_version=${KALDIFEAT_VERSION}
LABEL github_repo="https://github.com/k2-fsa/icefall"
RUN apt-get update && \
apt-get install -y --no-install-recommends \
curl \
vim \
libssl-dev \
autoconf \
automake \
bzip2 \
ca-certificates \
ffmpeg \
g++ \
gfortran \
git \
libtool \
make \
patch \
sox \
subversion \
unzip \
valgrind \
wget \
zlib1g-dev \
&& rm -rf /var/lib/apt/lists/*
# Install dependencies
RUN pip install --no-cache-dir \
torchaudio==${TORCHAUDIO_VERSION} -f https://download.pytorch.org/whl/torch_stable.html \
k2==${K2_VERSION} -f https://k2-fsa.github.io/k2/cuda.html \
git+https://github.com/lhotse-speech/lhotse \
kaldifeat==${KALDIFEAT_VERSION} -f https://csukuangfj.github.io/kaldifeat/cuda.html \
kaldi_native_io \
kaldialign \
kaldifst \
kaldilm \
sentencepiece>=0.1.96 \
tensorboard \
typeguard \
dill \
onnx \
onnxruntime \
onnxmltools \
onnxoptimizer \
onnxsim \
multi_quantization \
typeguard \
numpy \
pytest \
graphviz
RUN git clone https://github.com/k2-fsa/icefall /workspace/icefall && \
cd /workspace/icefall && \
pip install --no-cache-dir -r requirements.txt
ENV PYTHONPATH /workspace/icefall:$PYTHONPATH
WORKDIR /workspace/icefall

73
docker/torch2.2.0-cuda12.1.dockerfile Normal file
View File

@ -0,0 +1,73 @@
FROM pytorch/pytorch:2.2.0-cuda12.1-cudnn8-devel
# python 3.10
ENV LC_ALL C.UTF-8
ARG DEBIAN_FRONTEND=noninteractive
# python 3.10
ARG K2_VERSION="1.24.4.dev20240223+cuda12.1.torch2.2.0"
ARG KALDIFEAT_VERSION="1.25.4.dev20240223+cuda12.1.torch2.2.0"
ARG TORCHAUDIO_VERSION="2.2.0+cu121"
LABEL authors="Fangjun Kuang <csukuangfj@gmail.com>"
LABEL k2_version=${K2_VERSION}
LABEL kaldifeat_version=${KALDIFEAT_VERSION}
LABEL github_repo="https://github.com/k2-fsa/icefall"
RUN apt-get update && \
apt-get install -y --no-install-recommends \
curl \
vim \
libssl-dev \
autoconf \
automake \
bzip2 \
ca-certificates \
ffmpeg \
g++ \
gfortran \
git \
libtool \
make \
patch \
sox \
subversion \
unzip \
valgrind \
wget \
zlib1g-dev \
&& rm -rf /var/lib/apt/lists/*
# Install dependencies
RUN pip install --no-cache-dir \
torchaudio==${TORCHAUDIO_VERSION} -f https://download.pytorch.org/whl/torch_stable.html \
k2==${K2_VERSION} -f https://k2-fsa.github.io/k2/cuda.html \
git+https://github.com/lhotse-speech/lhotse \
kaldifeat==${KALDIFEAT_VERSION} -f https://csukuangfj.github.io/kaldifeat/cuda.html \
kaldi_native_io \
kaldialign \
kaldifst \
kaldilm \
sentencepiece>=0.1.96 \
tensorboard \
typeguard \
dill \
onnx \
onnxruntime \
onnxmltools \
onnxoptimizer \
onnxsim \
multi_quantization \
typeguard \
numpy \
pytest \
graphviz
RUN git clone https://github.com/k2-fsa/icefall /workspace/icefall && \
cd /workspace/icefall && \
pip install --no-cache-dir -r requirements.txt
ENV PYTHONPATH /workspace/icefall:$PYTHONPATH
WORKDIR /workspace/icefall

73
docker/torch2.2.1-cuda11.8.dockerfile Normal file
View File

@ -0,0 +1,73 @@
FROM pytorch/pytorch:2.2.1-cuda11.8-cudnn8-devel
# python 3.10
ENV LC_ALL C.UTF-8
ARG DEBIAN_FRONTEND=noninteractive
# python 3.10
ARG K2_VERSION="1.24.4.dev20240223+cuda11.8.torch2.2.1"
ARG KALDIFEAT_VERSION="1.25.4.dev20240223+cuda11.8.torch2.2.1"
ARG TORCHAUDIO_VERSION="2.2.1+cu118"
LABEL authors="Fangjun Kuang <csukuangfj@gmail.com>"
LABEL k2_version=${K2_VERSION}
LABEL kaldifeat_version=${KALDIFEAT_VERSION}
LABEL github_repo="https://github.com/k2-fsa/icefall"
RUN apt-get update && \
apt-get install -y --no-install-recommends \
curl \
vim \
libssl-dev \
autoconf \
automake \
bzip2 \
ca-certificates \
ffmpeg \
g++ \
gfortran \
git \
libtool \
make \
patch \
sox \
subversion \
unzip \
valgrind \
wget \
zlib1g-dev \
&& rm -rf /var/lib/apt/lists/*
# Install dependencies
RUN pip install --no-cache-dir \
torchaudio==${TORCHAUDIO_VERSION} -f https://download.pytorch.org/whl/torch_stable.html \
k2==${K2_VERSION} -f https://k2-fsa.github.io/k2/cuda.html \
git+https://github.com/lhotse-speech/lhotse \
kaldifeat==${KALDIFEAT_VERSION} -f https://csukuangfj.github.io/kaldifeat/cuda.html \
kaldi_native_io \
kaldialign \
kaldifst \
kaldilm \
sentencepiece>=0.1.96 \
tensorboard \
typeguard \
dill \
onnx \
onnxruntime \
onnxmltools \
onnxoptimizer \
onnxsim \
multi_quantization \
typeguard \
numpy \
pytest \
graphviz
RUN git clone https://github.com/k2-fsa/icefall /workspace/icefall && \
cd /workspace/icefall && \
pip install --no-cache-dir -r requirements.txt
ENV PYTHONPATH /workspace/icefall:$PYTHONPATH
WORKDIR /workspace/icefall

73
docker/torch2.2.1-cuda12.1.dockerfile Normal file
View File

@ -0,0 +1,73 @@
FROM pytorch/pytorch:2.2.1-cuda12.1-cudnn8-devel
# python 3.10
ENV LC_ALL C.UTF-8
ARG DEBIAN_FRONTEND=noninteractive
# python 3.10
ARG K2_VERSION="1.24.4.dev20240223+cuda12.1.torch2.2.1"
ARG KALDIFEAT_VERSION="1.25.4.dev20240223+cuda12.1.torch2.2.1"
ARG TORCHAUDIO_VERSION="2.2.1+cu121"
LABEL authors="Fangjun Kuang <csukuangfj@gmail.com>"
LABEL k2_version=${K2_VERSION}
LABEL kaldifeat_version=${KALDIFEAT_VERSION}
LABEL github_repo="https://github.com/k2-fsa/icefall"
RUN apt-get update && \
apt-get install -y --no-install-recommends \
curl \
vim \
libssl-dev \
autoconf \
automake \
bzip2 \
ca-certificates \
ffmpeg \
g++ \
gfortran \
git \
libtool \
make \
patch \
sox \
subversion \
unzip \
valgrind \
wget \
zlib1g-dev \
&& rm -rf /var/lib/apt/lists/*
# Install dependencies
RUN pip install --no-cache-dir \
torchaudio==${TORCHAUDIO_VERSION} -f https://download.pytorch.org/whl/torch_stable.html \
k2==${K2_VERSION} -f https://k2-fsa.github.io/k2/cuda.html \
git+https://github.com/lhotse-speech/lhotse \
kaldifeat==${KALDIFEAT_VERSION} -f https://csukuangfj.github.io/kaldifeat/cuda.html \
kaldi_native_io \
kaldialign \
kaldifst \
kaldilm \
sentencepiece>=0.1.96 \
tensorboard \
typeguard \
dill \
onnx \
onnxruntime \
onnxmltools \
onnxoptimizer \
onnxsim \
multi_quantization \
typeguard \
numpy \
pytest \
graphviz
RUN git clone https://github.com/k2-fsa/icefall /workspace/icefall && \
cd /workspace/icefall && \
pip install --no-cache-dir -r requirements.txt
ENV PYTHONPATH /workspace/icefall:$PYTHONPATH
WORKDIR /workspace/icefall

73
docker/torch2.2.2-cuda11.8.dockerfile Normal file
View File

@ -0,0 +1,73 @@
FROM pytorch/pytorch:2.2.2-cuda11.8-cudnn8-devel
# python 3.10
ENV LC_ALL C.UTF-8
ARG DEBIAN_FRONTEND=noninteractive
# python 3.10
ARG K2_VERSION="1.24.4.dev20240328+cuda11.8.torch2.2.2"
ARG KALDIFEAT_VERSION="1.25.4.dev20240329+cuda11.8.torch2.2.2"
ARG TORCHAUDIO_VERSION="2.2.2+cu118"
LABEL authors="Fangjun Kuang <csukuangfj@gmail.com>"
LABEL k2_version=${K2_VERSION}
LABEL kaldifeat_version=${KALDIFEAT_VERSION}
LABEL github_repo="https://github.com/k2-fsa/icefall"
RUN apt-get update && \
apt-get install -y --no-install-recommends \
curl \
vim \
libssl-dev \
autoconf \
automake \
bzip2 \
ca-certificates \
ffmpeg \
g++ \
gfortran \
git \
libtool \
make \
patch \
sox \
subversion \
unzip \
valgrind \
wget \
zlib1g-dev \
&& rm -rf /var/lib/apt/lists/*
# Install dependencies
RUN pip install --no-cache-dir \
torchaudio==${TORCHAUDIO_VERSION} -f https://download.pytorch.org/whl/torch_stable.html \
k2==${K2_VERSION} -f https://k2-fsa.github.io/k2/cuda.html \
git+https://github.com/lhotse-speech/lhotse \
kaldifeat==${KALDIFEAT_VERSION} -f https://csukuangfj.github.io/kaldifeat/cuda.html \
kaldi_native_io \
kaldialign \
kaldifst \
kaldilm \
sentencepiece>=0.1.96 \
tensorboard \
typeguard \
dill \
onnx \
onnxruntime \
onnxmltools \
onnxoptimizer \
onnxsim \
multi_quantization \
typeguard \
numpy \
pytest \
graphviz
RUN git clone https://github.com/k2-fsa/icefall /workspace/icefall && \
cd /workspace/icefall && \
pip install --no-cache-dir -r requirements.txt
ENV PYTHONPATH /workspace/icefall:$PYTHONPATH
WORKDIR /workspace/icefall

73
docker/torch2.2.2-cuda12.1.dockerfile Normal file
View File

@ -0,0 +1,73 @@
FROM pytorch/pytorch:2.2.2-cuda12.1-cudnn8-devel
# python 3.10
ENV LC_ALL C.UTF-8
ARG DEBIAN_FRONTEND=noninteractive
# python 3.10
ARG K2_VERSION="1.24.4.dev20240328+cuda12.1.torch2.2.2"
ARG KALDIFEAT_VERSION="1.25.4.dev20240329+cuda12.1.torch2.2.2"
ARG TORCHAUDIO_VERSION="2.2.2+cu121"
LABEL authors="Fangjun Kuang <csukuangfj@gmail.com>"
LABEL k2_version=${K2_VERSION}
LABEL kaldifeat_version=${KALDIFEAT_VERSION}
LABEL github_repo="https://github.com/k2-fsa/icefall"
RUN apt-get update && \
apt-get install -y --no-install-recommends \
curl \
vim \
libssl-dev \
autoconf \
automake \
bzip2 \
ca-certificates \
ffmpeg \
g++ \
gfortran \
git \
libtool \
make \
patch \
sox \
subversion \
unzip \
valgrind \
wget \
zlib1g-dev \
&& rm -rf /var/lib/apt/lists/*
# Install dependencies
RUN pip install --no-cache-dir \
torchaudio==${TORCHAUDIO_VERSION} -f https://download.pytorch.org/whl/torch_stable.html \
k2==${K2_VERSION} -f https://k2-fsa.github.io/k2/cuda.html \
git+https://github.com/lhotse-speech/lhotse \
kaldifeat==${KALDIFEAT_VERSION} -f https://csukuangfj.github.io/kaldifeat/cuda.html \
kaldi_native_io \
kaldialign \
kaldifst \
kaldilm \
sentencepiece>=0.1.96 \
tensorboard \
typeguard \
dill \
onnx \
onnxruntime \
onnxmltools \
onnxoptimizer \
onnxsim \
multi_quantization \
typeguard \
numpy \
pytest \
graphviz
RUN git clone https://github.com/k2-fsa/icefall /workspace/icefall && \
cd /workspace/icefall && \
pip install --no-cache-dir -r requirements.txt
ENV PYTHONPATH /workspace/icefall:$PYTHONPATH
WORKDIR /workspace/icefall

6
docs/source/decoding-with-langugage-models/LODR.rst
View File

@ -30,7 +30,7 @@ of langugae model integration.
First, let's have a look at some background information. As the predecessor of LODR, Density Ratio (DR) is first proposed `here <https://arxiv.org/abs/2002.11268>`_
to address the language information mismatch between the training
corpus (source domain) and the testing corpus (target domain). Assuming that the source domain and the test domain
are acoustically similar, DR derives the following formular for decoding with Bayes' theorem:
are acoustically similar, DR derives the following formula for decoding with Bayes' theorem:
.. math::
@ -41,7 +41,7 @@ are acoustically similar, DR derives the following formular for decoding with Ba
where :math:`\lambda_1` and :math:`\lambda_2` are the weights of LM scores for target domain and source domain respectively.
Here, the source domain LM is trained on the training corpus. The only difference in the above formular compared to
Here, the source domain LM is trained on the training corpus. The only difference in the above formula compared to
shallow fusion is the subtraction of the source domain LM.
Some works treat the predictor and the joiner of the neural transducer as its internal LM. However, the LM is
@ -58,7 +58,7 @@ during decoding for transducer model:
In LODR, an additional bi-gram LM estimated on the source domain (e.g training corpus) is required. Compared to DR,
the only difference lies in the choice of source domain LM. According to the original `paper <https://arxiv.org/abs/2203.16776>`_,
LODR achieves similar performance compared DR in both intra-domain and cross-domain settings.
LODR achieves similar performance compared to DR in both intra-domain and cross-domain settings.
As a bi-gram is much faster to evaluate, LODR is usually much faster.
Now, we will show you how to use LODR in ``icefall``.

2
docs/source/decoding-with-langugage-models/shallow-fusion.rst
View File

@ -139,7 +139,7 @@ A few parameters can be tuned to further boost the performance of shallow fusion
- ``--lm-scale``
Controls the scale of the LM. If too small, the external language model may not be fully utilized; if too large,
the LM score may dominant during decoding, leading to bad WER. A typical value of this is around 0.3.
the LM score might be dominant during decoding, leading to bad WER. A typical value of this is around 0.3.
- ``--beam-size``

52
docs/source/docker/intro.rst
View File

@ -20,7 +20,11 @@ We describe the following items in this section:
View available tags
===================
You can use the following command to view available tags:
CUDA-enabled docker images
--------------------------
You can use the following command to view available tags for CUDA-enabled
docker images:
.. code-block:: bash
@ -30,6 +34,12 @@ which will give you something like below:
.. code-block:: bash
"torch2.2.2-cuda12.1"
"torch2.2.2-cuda11.8"
"torch2.2.1-cuda12.1"
"torch2.2.1-cuda11.8"
"torch2.2.0-cuda12.1"
"torch2.2.0-cuda11.8"
"torch2.1.0-cuda12.1"
"torch2.1.0-cuda11.8"
"torch2.0.0-cuda11.7"
@ -43,8 +53,25 @@ which will give you something like below:
Please select an appropriate combination of `torch`_ and CUDA.
Download a docker image
=======================
CPU-only docker images
----------------------
To view CPU-only docker images, please visit `<https://github.com/k2-fsa/icefall/pkgs/container/icefall>`_
for available tags.
You can select different combinations of ``Python`` and ``torch``. For instance,
to select ``Python 3.8`` and ``torch 2.1.2``, you can use the following tag
.. code-block:: bash
cpu-py3.8-torch2.1.2-v1.1
where ``v1.1`` is the current version of the docker image. You may see
``ghcr.io/k2-fsa/icefall:cpu-py3.8-torch2.1.2-v1.2`` or some other versions.
We recommend that you always use the latest version.
Download a docker image (CUDA)
==============================
Suppose that you select the tag ``torch1.13.0-cuda11.6``, you can use
the following command to download it:
@ -53,6 +80,16 @@ the following command to download it:
sudo docker image pull k2fsa/icefall:torch1.13.0-cuda11.6
Download a docker image (CPU)
==============================
Suppose that you select the tag ``cpu-py3.8-torch2.1.2-v1.1``, you can use
the following command to download it:
.. code-block:: bash
sudo docker pull ghcr.io/k2-fsa/icefall:cpu-py3.8-torch2.1.2-v1.1
Run a docker image with GPU
===========================
@ -65,7 +102,7 @@ Run a docker image with CPU
.. code-block:: bash
sudo docker run --rm -it k2fsa/icefall:torch1.13.0-cuda11.6 /bin/bash
sudo docker run --rm -it ghcr.io/k2-fsa/icefall:cpu-py3.8-torch2.1.2-v1.1 /bin/bash
Run yesno within a docker container
===================================
@ -74,8 +111,13 @@ After starting the container, the following interface is presented:
.. code-block:: bash
# GPU-enabled docker
root@60c947eac59c:/workspace/icefall#
# CPU-only docker
root@60c947eac59c:# mkdir /workspace; git clone https://github.com/k2-fsa/icefall
root@60c947eac59c:# export PYTHONPATH=/workspace/icefall:$PYTHONPATH
It shows the current user is ``root`` and the current working directory
is ``/workspace/icefall``.
@ -107,7 +149,7 @@ to switch to the ``yesno`` recipe and run
.. hint::
If you are running without GPU, it may report the following error:
If you are running without GPU with a GPU-enabled docker, it may report the following error:
.. code-block:: bash

8
docs/source/for-dummies/environment-setup.rst
View File

@ -66,13 +66,17 @@ to install dependencies of `icefall`_:
pip install torch==2.0.0+cpu torchaudio==2.0.0+cpu -f https://download.pytorch.org/whl/torch_stable.html
# If you are using macOS or Windows, please use the following command to install torch and torchaudio
# If you are using macOS, please use the following command to install torch and torchaudio
# pip install torch==2.0.0 torchaudio==2.0.0 -f https://download.pytorch.org/whl/torch_stable.html
# Now install k2
# Please refer to https://k2-fsa.github.io/k2/installation/from_wheels.html#linux-cpu-example
pip install k2==1.24.3.dev20230726+cpu.torch2.0.0 -f https://k2-fsa.github.io/k2/cpu.html
pip install k2==1.24.4.dev20231220+cpu.torch2.0.0 -f https://k2-fsa.github.io/k2/cpu.html
# For users from China
# 中国国内用户，如果访问不了 huggingface, 请使用
# pip install k2==1.24.4.dev20231220+cpu.torch2.0.0 -f https://k2-fsa.github.io/k2/cpu-cn.html
# Install the latest version of lhotse

6
docs/source/for-dummies/model-export.rst
View File

@ -85,7 +85,7 @@ We can also use it to decode files with the following command:
# Please refer to https://csukuangfj.github.io/kaldifeat/installation/from_wheels.html
# for how to install kaldifeat
pip install kaldifeat==1.25.0.dev20230726+cpu.torch2.0.0 -f https://csukuangfj.github.io/kaldifeat/cpu.html
pip install kaldifeat==1.25.3.dev20231221+cpu.torch2.0.0 -f https://csukuangfj.github.io/kaldifeat/cpu.html
./tdnn/pretrained.py \
--checkpoint ./tdnn/exp/pretrained.pt \
@ -162,7 +162,7 @@ To use ``tdnn/exp/cpu_jit.pt`` with `icefall`_ to decode files, we can use:
# Please refer to https://csukuangfj.github.io/kaldifeat/installation/from_wheels.html
# for how to install kaldifeat
pip install kaldifeat==1.25.0.dev20230726+cpu.torch2.0.0 -f https://csukuangfj.github.io/kaldifeat/cpu.html
pip install kaldifeat==1.25.3.dev20231221+cpu.torch2.0.0 -f https://csukuangfj.github.io/kaldifeat/cpu.html
./tdnn/jit_pretrained.py \
@ -249,7 +249,7 @@ To use the generated ONNX model files for decoding with `onnxruntime`_, we can u
# Please refer to https://csukuangfj.github.io/kaldifeat/installation/from_wheels.html
# for how to install kaldifeat
pip install kaldifeat==1.25.0.dev20230726+cpu.torch2.0.0 -f https://csukuangfj.github.io/kaldifeat/cpu.html
pip install kaldifeat==1.25.3.dev20231221+cpu.torch2.0.0 -f https://csukuangfj.github.io/kaldifeat/cpu.html
./tdnn/onnx_pretrained.py \
--nn-model ./tdnn/exp/model-epoch-14-avg-2.onnx \

3
docs/source/installation/index.rst
View File

@ -206,6 +206,9 @@ We will install `k2`_ from pre-compiled wheels by following
.. code-block:: bash
(test-icefall) kuangfangjun:~$ pip install k2==1.24.3.dev20230725+cuda11.6.torch1.13.0 -f https://k2-fsa.github.io/k2/cuda.html
# For users from China
# 中国国内用户，如果访问不了 huggingface, 请使用
# pip install k2==1.24.3.dev20230725+cuda11.6.torch1.13.0 -f https://k2-fsa.github.io/k2/cuda-cn.html
Looking in indexes: https://pypi.tuna.tsinghua.edu.cn/simple
Looking in links: https://k2-fsa.github.io/k2/cuda.html

225
docs/source/recipes/Finetune/adapter/finetune_adapter.rst Normal file
View File

@ -0,0 +1,225 @@
Finetune from a pre-trained Zipformer model with adapters
=========================================================
This tutorial shows you how to fine-tune a pre-trained **Zipformer**
transducer model on a new dataset with adapters.
Adapters are compact and efficient module that can be integrated into a pre-trained model
to improve the model's performance on a new domain. Adapters are injected
between different modules in the well-trained neural network. During training, only the parameters
in the adapters will be updated. It achieves competitive performance
while requiring much less GPU memory than full fine-tuning. For more details about adapters,
please refer to the original `paper <https://arxiv.org/pdf/1902.00751.pdf#/>`_ for more details.
.. HINT::
We assume you have read the page :ref:`install icefall` and have setup
the environment for ``icefall``.
.. HINT::
We recommend you to use a GPU or several GPUs to run this recipe
For illustration purpose, we fine-tune the Zipformer transducer model
pre-trained on `LibriSpeech`_ on the small subset of `GigaSpeech`_. You could use your
own data for fine-tuning if you create a manifest for your new dataset.
Data preparation
----------------
Please follow the instructions in the `GigaSpeech recipe <https://github.com/k2-fsa/icefall/tree/master/egs/gigaspeech/ASR>`_
to prepare the fine-tune data used in this tutorial. We only require the small subset in GigaSpeech for this tutorial.
Model preparation
-----------------
We are using the Zipformer model trained on full LibriSpeech (960 hours) as the intialization. The
checkpoint of the model can be downloaded via the following command:
.. code-block:: bash
$ GIT_LFS_SKIP_SMUDGE=1 git clone https://huggingface.co/Zengwei/icefall-asr-librispeech-zipformer-2023-05-15
$ cd icefall-asr-librispeech-zipformer-2023-05-15/exp
$ git lfs pull --include "pretrained.pt"
$ ln -s pretrained.pt epoch-99.pt
$ cd ../data/lang_bpe_500
$ git lfs pull --include bpe.model
$ cd ../../..
Before fine-tuning, let's test the model's WER on the new domain. The following command performs
decoding on the GigaSpeech test sets:
.. code-block:: bash
./zipformer/decode_gigaspeech.py \
--epoch 99 \
--avg 1 \
--exp-dir icefall-asr-librispeech-zipformer-2023-05-15/exp \
--use-averaged-model 0 \
--max-duration 1000 \
--decoding-method greedy_search
You should see the following numbers:
.. code-block::
For dev, WER of different settings are:
greedy_search 20.06 best for dev
For test, WER of different settings are:
greedy_search 19.27 best for test
Fine-tune with adapter
----------------------
We insert 4 adapters with residual connection in each ``Zipformer2EncoderLayer``.
The original model parameters remain untouched during training and only the parameters of
the adapters are updated. The following command starts a fine-tuning experiment with adapters:
.. code-block:: bash
$ do_finetune=1
$ use_adapters=1
$ adapter_dim=8
$ ./zipformer_adapter/train.py \
--world-size 2 \
--num-epochs 20 \
--start-epoch 1 \
--exp-dir zipformer_adapter/exp_giga_finetune_adapters${use_adapters}_adapter_dim${adapter_dim} \
--use-fp16 1 \
--base-lr 0.045 \
--use-adapters $use_adapters --adapter-dim $adapter_dim \
--bpe-model data/lang_bpe_500/bpe.model \
--do-finetune $do_finetune \
--master-port 13022 \
--finetune-ckpt icefall-asr-librispeech-zipformer-2023-05-15/exp/pretrained.pt \
--max-duration 1000
The following arguments are related to fine-tuning:
- ``--do-finetune``
If True, do fine-tuning by initializing the model from a pre-trained checkpoint.
**Note that if you want to resume your fine-tuning experiment from certain epochs, you
need to set this to False.**
- ``use-adapters``
If adapters are used during fine-tuning.
- ``--adapter-dim``
The bottleneck dimension of the adapter module. Typically a small number.
You should notice that in the training log, the total number of trainale parameters is shown:
.. code-block::
2024-02-22 21:22:03,808 INFO [train.py:1277] A total of 761344 trainable parameters (1.148% of the whole model)
The trainable parameters only makes up 1.15% of the entire model parameters, so the training will be much faster
and requires less memory than full fine-tuning.
Decoding
--------
After training, let's test the WERs. To test the WERs on the GigaSpeech set,
you can execute the following command:
.. code-block:: bash
$ epoch=20
$ avg=10
$ use_adapters=1
$ adapter_dim=8
% ./zipformer/decode.py \
--epoch $epoch \
--avg $avg \
--use-averaged-model 1 \
--exp-dir zipformer_adapter/exp_giga_finetune_adapters${use_adapters}_adapter_dim${adapter_dim} \
--max-duration 600 \
--use-adapters $use_adapters \
--adapter-dim $adapter_dim \
--decoding-method greedy_search
You should see the following numbers:
.. code-block::
For dev, WER of different settings are:
greedy_search 15.44 best for dev
For test, WER of different settings are:
greedy_search 15.42 best for test
The WER on test set is improved from 19.27 to 15.42, demonstrating the effectiveness of adapters.
The same model can be used to perform decoding on LibriSpeech test sets. You can deactivate the adapters
to keep the same performance of the original model:
.. code-block:: bash
$ epoch=20
$ avg=1
$ use_adapters=0
$ adapter_dim=8
% ./zipformer/decode.py \
--epoch $epoch \
--avg $avg \
--use-averaged-model 1 \
--exp-dir zipformer_adapter/exp_giga_finetune_adapters${use_adapters}_adapter_dim${adapter_dim} \
--max-duration 600 \
--use-adapters $use_adapters \
--adapter-dim $adapter_dim \
--decoding-method greedy_search
.. code-block::
For dev, WER of different settings are:
greedy_search 2.23 best for test-clean
For test, WER of different settings are:
greedy_search 4.96 best for test-other
The numbers are the same as reported in `icefall <https://github.com/k2-fsa/icefall/blob/master/egs/librispeech/ASR/RESULTS.md#normal-scaled-model-number-of-model-parameters-65549011-ie-6555-m>`_. So adapter-based
fine-tuning is also very flexible as the same model can be used for decoding on the original and target domain.
Export the model
----------------
After training, the model can be exported to ``onnx`` format easily using the following command:
.. code-block:: bash
$ use_adapters=1
$ adapter_dim=16
$ ./zipformer_adapter/export-onnx.py \
--tokens icefall-asr-librispeech-zipformer-2023-05-15/data/lang_bpe_500/tokens.txt \
--use-averaged-model 1 \
--epoch 20 \
--avg 10 \
--exp-dir zipformer_adapter/exp_giga_finetune_adapters${use_adapters}_adapter_dim${adapter_dim} \
--use-adapters $use_adapters \
--adapter-dim $adapter_dim \
--num-encoder-layers "2,2,3,4,3,2" \
--downsampling-factor "1,2,4,8,4,2" \
--feedforward-dim "512,768,1024,1536,1024,768" \
--num-heads "4,4,4,8,4,4" \
--encoder-dim "192,256,384,512,384,256" \
--query-head-dim 32 \
--value-head-dim 12 \
--pos-head-dim 4 \
--pos-dim 48 \
--encoder-unmasked-dim "192,192,256,256,256,192" \
--cnn-module-kernel "31,31,15,15,15,31" \
--decoder-dim 512 \
--joiner-dim 512 \
--causal False \
--chunk-size "16,32,64,-1" \
--left-context-frames "64,128,256,-1"

Some files were not shown because too many files have changed in this diff Show More