mirrors/icefall
1
0
Fork 0
mirror of https://github.com/k2-fsa/icefall.git synced 2025-12-11 06:55:27 +00:00
Code Issues Packages Projects Releases Wiki Activity

Merge branch 'k2-fsa:master' into phone

Browse Source
This commit is contained in:
Yifan Yang 2024-04-11 16:30:54 +08:00 committed by GitHub
commit 4858e2b036
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1590 changed files with 246927 additions and 9674 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
.flake8
View File

@ -13,8 +13,9 @@ per-file-ignores =
egs/librispeech/ASR/conv_emformer_transducer_stateless*/*.py: E501, E203
egs/librispeech/ASR/conformer_ctc*/*py: E501,
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
@ -23,6 +24,7 @@ exclude =
**/data/**,
icefall/shared/make_kn_lm.py,
icefall/__init__.py
icefall/ctc/__init__.py
ignore =
# E203 white space before ":"

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 .
log "Downloading pre-trained model from $repo_url"
repo_url=https://huggingface.co/csukuangfj/icefall-aishell-pruned-transducer-stateless3-2022-06-20
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

3
.github/scripts/run-gigaspeech-pruned-transducer-stateless2-2022-05-12.sh vendored
View File

@ -29,6 +29,9 @@ if [[ x"${GITHUB_EVENT_NAME}" == x"schedule" || x"${GITHUB_EVENT_LABEL_NAME}" ==
ls -lh data/fbank
ls -lh pruned_transducer_stateless2/exp
ln -s data/fbank/cuts_DEV.jsonl.gz data/fbank/gigaspeech_cuts_DEV.jsonl.gz
ln -s data/fbank/cuts_TEST.jsonl.gz data/fbank/gigaspeech_cuts_TEST.jsonl.gz
log "Decoding dev and test"
# use a small value for decoding with CPU

158
.github/scripts/run-gigaspeech-zipformer-2023-10-17.sh vendored Executable file
View File

@ -0,0 +1,158 @@
#!/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/gigaspeech/ASR
repo_url=https://huggingface.co/yfyeung/icefall-asr-gigaspeech-zipformer-2023-10-17
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.pt"
git lfs pull --include "exp/pretrained.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 30 \
--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-30.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 30 \
--avg 1 \
--use-averaged-model 0 \
--max-duration $max_duration \
--exp-dir zipformer/exp
done
rm zipformer/exp/*.pt
fi

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 \
--lang-dir $repo/data/lang_bpe_500 \
--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 \
--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 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

6
.github/scripts/run-librispeech-lstm-transducer-stateless2-2022-09-03.sh vendored
View File

@ -31,7 +31,7 @@ log "Test exporting with torch.jit.trace()"
./lstm_transducer_stateless2/export.py \
--exp-dir $repo/exp \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
--tokens $repo/data/lang_bpe_500/tokens.txt \
--epoch 99 \
--avg 1 \
--use-averaged-model 0 \
@ -55,7 +55,7 @@ for sym in 1 2 3; do
--method greedy_search \
--max-sym-per-frame $sym \
--checkpoint $repo/exp/pretrained.pt \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
--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
@ -68,7 +68,7 @@ for method in modified_beam_search beam_search fast_beam_search; do
--method $method \
--beam-size 4 \
--checkpoint $repo/exp/pretrained.pt \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
--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

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 \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
$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 \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
$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 \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
$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 \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
$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 \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
$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 \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
$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 \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
--epoch 99 \
--avg 1 \
--jit 1
./pruned_transducer_stateless3/export.py \
--exp-dir $repo/exp \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
--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 \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
$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 \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
$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 \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
--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 \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
$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 \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
--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 \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
$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 \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
$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 \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
--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 \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
$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 \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
$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-2022-12-15.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/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 "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 \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
--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 \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
$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 \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
$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 \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
--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 \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
--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 \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
--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 \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
--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 \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
$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 \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
$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 \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
--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 \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
--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

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 \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
$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 \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
$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

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 \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
--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 \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
$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

135
.github/scripts/run-multi-corpora-zipformer.sh vendored Executable file
View File

@ -0,0 +1,135 @@
#!/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/multi_zh-hans/ASR
log "==== Test icefall-asr-multi-zh-hans-zipformer-2023-9-2 ===="
repo_url=https://huggingface.co/zrjin/icefall-asr-multi-zh-hans-zipformer-2023-9-2/
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 epoch-20.pt epoch-99.pt
popd
ls -lh $repo/exp/*.pt
./zipformer/pretrained.py \
--checkpoint $repo/exp/epoch-99.pt \
--tokens $repo/data/lang_bpe_2000/tokens.txt \
--method greedy_search \
$repo/test_wavs/DEV_T0000000000.wav \
$repo/test_wavs/DEV_T0000000001.wav \
$repo/test_wavs/DEV_T0000000002.wav
for method in modified_beam_search fast_beam_search; do
log "$method"
./zipformer/pretrained.py \
--method $method \
--beam-size 4 \
--checkpoint $repo/exp/epoch-99.pt \
--tokens $repo/data/lang_bpe_2000/tokens.txt \
$repo/test_wavs/DEV_T0000000000.wav \
$repo/test_wavs/DEV_T0000000001.wav \
$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/
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 epoch-20.pt epoch-99.pt
popd
ls -lh $repo/exp/*.pt
./zipformer/pretrained.py \
--checkpoint $repo/exp/epoch-99.pt \
--tokens $repo/data/lang_bpe_2000/tokens.txt \
--use-ctc 1 \
--method greedy_search \
$repo/test_wavs/DEV_T0000000000.wav \
$repo/test_wavs/DEV_T0000000001.wav \
$repo/test_wavs/DEV_T0000000002.wav
for method in modified_beam_search fast_beam_search; do
log "$method"
./zipformer/pretrained.py \
--method $method \
--beam-size 4 \
--use-ctc 1 \
--checkpoint $repo/exp/epoch-99.pt \
--tokens $repo/data/lang_bpe_2000/tokens.txt \
$repo/test_wavs/DEV_T0000000000.wav \
$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

46
.github/scripts/run-pre-trained-conformer-ctc.sh vendored
View File

@ -1,46 +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://github.com/csukuangfj/icefall-asr-conformer-ctc-bpe-500
git lfs install
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/*.flac
log "CTC decoding"
./conformer_ctc/pretrained.py \
--method ctc-decoding \
--num-classes 500 \
--checkpoint $repo/exp/pretrained.pt \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
$repo/test_wavs/1089-134686-0001.flac \
$repo/test_wavs/1221-135766-0001.flac \
$repo/test_wavs/1221-135766-0002.flac
log "HLG decoding"
./conformer_ctc/pretrained.py \
--method 1best \
--num-classes 500 \
--checkpoint $repo/exp/pretrained.pt \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
--words-file $repo/data/lang_bpe_500/words.txt \
--HLG $repo/data/lang_bpe_500/HLG.pt \
$repo/test_wavs/1089-134686-0001.flac \
$repo/test_wavs/1221-135766-0001.flac \
$repo/test_wavs/1221-135766-0002.flac

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 \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
$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 \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
$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 \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
$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 \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
$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 \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
$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 \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
$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 \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
$repo/test_wavs/1089-134686-0001.wav \
$repo/test_wavs/1221-135766-0001.wav \
$repo/test_wavs/1221-135766-0002.wav

44
.github/scripts/run-swbd-conformer-ctc-2023-08-26.sh vendored Executable file
View File

@ -0,0 +1,44 @@
#!/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/swbd/ASR
repo_url=https://huggingface.co/zrjin/icefall-asr-swbd-conformer-ctc-2023-8-26
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 epoch-98.pt epoch-99.pt
popd
ls -lh $repo/exp/*.pt
for method in ctc-decoding 1best; do
log "$method"
./conformer_ctc/pretrained.py \
--method $method \
--checkpoint $repo/exp/epoch-99.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 \
--G $repo/data/lm/G_4_gram.pt \
$repo/test_wavs/1089-134686-0001.wav \
$repo/test_wavs/1221-135766-0001.wav \
$repo/test_wavs/1221-135766-0002.wav
done

42
.github/scripts/run-wenetspeech-pruned-transducer-stateless2.sh vendored
View File

@ -17,7 +17,6 @@ git lfs install
git clone $repo_url
repo=$(basename $repo_url)
log "Display test files"
tree $repo/
ls -lh $repo/test_wavs/*.wav
@ -29,25 +28,24 @@ popd
log "Test exporting to ONNX format"
./pruned_transducer_stateless2/export.py \
./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 \
--onnx 1
--avg 1
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
@ -59,19 +57,17 @@ log "Decode with ONNX models"
./pruned_transducer_stateless2/onnx_check.py \
--jit-filename $repo/exp/cpu_jit.pt \
--onnx-encoder-filename $repo/exp/encoder.onnx \
--onnx-decoder-filename $repo/exp/decoder.onnx \
--onnx-joiner-filename $repo/exp/joiner.onnx \
--onnx-joiner-encoder-proj-filename $repo/exp/joiner_encoder_proj.onnx \
--onnx-joiner-decoder-proj-filename $repo/exp/joiner_decoder_proj.onnx
--onnx-encoder-filename $repo/exp/encoder-epoch-10-avg-2.onnx \
--onnx-decoder-filename $repo/exp/decoder-epoch-10-avg-2.onnx \
--onnx-joiner-filename $repo/exp/joiner-epoch-10-avg-2.onnx \
--onnx-joiner-encoder-proj-filename $repo/exp/joiner_encoder_proj-epoch-10-avg-2.onnx \
--onnx-joiner-decoder-proj-filename $repo/exp/joiner_decoder_proj-epoch-10-avg-2.onnx
./pruned_transducer_stateless2/onnx_pretrained.py \
--tokens $repo/data/lang_char/tokens.txt \
--encoder-model-filename $repo/exp/encoder.onnx \
--decoder-model-filename $repo/exp/decoder.onnx \
--joiner-model-filename $repo/exp/joiner.onnx \
--joiner-encoder-proj-model-filename $repo/exp/joiner_encoder_proj.onnx \
--joiner-decoder-proj-model-filename $repo/exp/joiner_decoder_proj.onnx \
--encoder-model-filename $repo/exp/encoder-epoch-99-avg-1.onnx \
--decoder-model-filename $repo/exp/decoder-epoch-99-avg-1.onnx \
--joiner-model-filename $repo/exp/joiner-epoch-99-avg-1.onnx \
$repo/test_wavs/DEV_T0000000000.wav \
$repo/test_wavs/DEV_T0000000001.wav \
$repo/test_wavs/DEV_T0000000002.wav
@ -104,9 +100,9 @@ for sym in 1 2 3; do
--lang-dir $repo/data/lang_char \
--decoding-method greedy_search \
--max-sym-per-frame $sym \
$repo/test_wavs/DEV_T0000000000.wav \
$repo/test_wavs/DEV_T0000000001.wav \
$repo/test_wavs/DEV_T0000000002.wav
$repo/test_wavs/DEV_T0000000000.wav \
$repo/test_wavs/DEV_T0000000001.wav \
$repo/test_wavs/DEV_T0000000002.wav
done
for method in modified_beam_search beam_search fast_beam_search; do
@ -117,7 +113,7 @@ for method in modified_beam_search beam_search fast_beam_search; do
--beam-size 4 \
--checkpoint $repo/exp/epoch-99.pt \
--lang-dir $repo/data/lang_char \
$repo/test_wavs/DEV_T0000000000.wav \
$repo/test_wavs/DEV_T0000000001.wav \
$repo/test_wavs/DEV_T0000000002.wav
$repo/test_wavs/DEV_T0000000000.wav \
$repo/test_wavs/DEV_T0000000001.wav \
$repo/test_wavs/DEV_T0000000002.wav
done

16
.github/scripts/test-ncnn-export.sh vendored
View File

@ -45,7 +45,6 @@ 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-30-avg-10-averaged.pt"
cd exp
@ -56,11 +55,10 @@ log "Export via torch.jit.trace()"
./conv_emformer_transducer_stateless2/export-for-ncnn.py \
--exp-dir $repo/exp \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
--epoch 99 \
--avg 1 \
--use-averaged-model 0 \
\
--tokens $repo/data/lang_bpe_500/tokens.txt \
--num-encoder-layers 12 \
--chunk-length 32 \
--cnn-module-kernel 31 \
@ -91,7 +89,6 @@ 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-iter-468000-avg-16.pt"
cd exp
@ -102,7 +99,7 @@ log "Export via torch.jit.trace()"
./lstm_transducer_stateless2/export-for-ncnn.py \
--exp-dir $repo/exp \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
--tokens $repo/data/lang_bpe_500/tokens.txt \
--epoch 99 \
--avg 1 \
--use-averaged-model 0
@ -140,7 +137,6 @@ 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.pt"
cd exp
@ -148,7 +144,7 @@ ln -s pretrained.pt epoch-99.pt
popd
./pruned_transducer_stateless7_streaming/export-for-ncnn.py \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
--tokens $repo/data/lang_bpe_500/tokens.txt \
--exp-dir $repo/exp \
--use-averaged-model 0 \
--epoch 99 \
@ -195,14 +191,14 @@ git lfs pull --include "data/lang_char_bpe/Linv.pt"
git lfs pull --include "exp/pretrained.pt"
cd exp
ln -s pretrained.pt epoch-99.pt
ln -s pretrained.pt epoch-9999.pt
popd
./pruned_transducer_stateless7_streaming/export-for-ncnn-zh.py \
--lang-dir $repo/data/lang_char_bpe \
--tokens $repo/data/lang_char_bpe/tokens.txt \
--exp-dir $repo/exp \
--use-averaged-model 0 \
--epoch 99 \
--epoch 9999 \
--avg 1 \
--decode-chunk-len 32 \
--num-encoder-layers "2,4,3,2,4" \

138
.github/scripts/test-onnx-export.sh vendored
View File

@ -10,7 +10,123 @@ log() {
cd egs/librispeech/ASR
log "=========================================================================="
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)
pushd $repo
git lfs pull --include "exp/pretrained.pt"
cd exp
ln -s pretrained.pt epoch-99.pt
popd
log "Export via torch.jit.script()"
./zipformer/export.py \
--exp-dir $repo/exp \
--tokens $repo/data/lang_bpe_500/tokens.txt \
--epoch 99 \
--avg 1 \
--jit 1
log "Test export to ONNX format"
./zipformer/export-onnx.py \
--tokens $repo/data/lang_bpe_500/tokens.txt \
--use-averaged-model 0 \
--epoch 99 \
--avg 1 \
--exp-dir $repo/exp \
--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"
ls -lh $repo/exp
log "Run onnx_check.py"
./zipformer/onnx_check.py \
--jit-filename $repo/exp/jit_script.pt \
--onnx-encoder-filename $repo/exp/encoder-epoch-99-avg-1.onnx \
--onnx-decoder-filename $repo/exp/decoder-epoch-99-avg-1.onnx \
--onnx-joiner-filename $repo/exp/joiner-epoch-99-avg-1.onnx
log "Run onnx_pretrained.py"
./zipformer/onnx_pretrained.py \
--encoder-model-filename $repo/exp/encoder-epoch-99-avg-1.onnx \
--decoder-model-filename $repo/exp/decoder-epoch-99-avg-1.onnx \
--joiner-model-filename $repo/exp/joiner-epoch-99-avg-1.onnx \
--tokens $repo/data/lang_bpe_500/tokens.txt \
$repo/test_wavs/1089-134686-0001.wav
rm -rf $repo
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)
pushd $repo
git lfs pull --include "exp/pretrained.pt"
cd exp
ln -s pretrained.pt epoch-99.pt
popd
log "Test export streaming model to ONNX format"
./zipformer/export-onnx-streaming.py \
--tokens $repo/data/lang_bpe_500/tokens.txt \
--use-averaged-model 0 \
--epoch 99 \
--avg 1 \
--exp-dir $repo/exp \
--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 True \
--chunk-size 16 \
--left-context-frames 64
ls -lh $repo/exp
log "Run onnx_pretrained-streaming.py"
./zipformer/onnx_pretrained-streaming.py \
--encoder-model-filename $repo/exp/encoder-epoch-99-avg-1-chunk-16-left-64.onnx \
--decoder-model-filename $repo/exp/decoder-epoch-99-avg-1-chunk-16-left-64.onnx \
--joiner-model-filename $repo/exp/joiner-epoch-99-avg-1-chunk-16-left-64.onnx \
--tokens $repo/data/lang_bpe_500/tokens.txt \
$repo/test_wavs/1089-134686-0001.wav
rm -rf $repo
log "--------------------------------------------------------------------------"
log "=========================================================================="
repo_url=https://huggingface.co/Zengwei/icefall-asr-librispeech-pruned-transducer-stateless7-streaming-2022-12-29
@ -39,7 +155,7 @@ log "Export via torch.jit.trace()"
log "Test exporting to ONNX format"
./pruned_transducer_stateless7_streaming/export-onnx.py \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
--tokens $repo/data/lang_bpe_500/tokens.txt \
--use-averaged-model 0 \
--epoch 99 \
--avg 1 \
@ -88,7 +204,7 @@ popd
log "Export via torch.jit.script()"
./pruned_transducer_stateless3/export.py \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
--tokens $repo/data/lang_bpe_500/tokens.txt \
--epoch 9999 \
--avg 1 \
--exp-dir $repo/exp/ \
@ -97,7 +213,7 @@ log "Export via torch.jit.script()"
log "Test exporting to ONNX format"
./pruned_transducer_stateless3/export-onnx.py \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
--tokens $repo/data/lang_bpe_500/tokens.txt \
--epoch 9999 \
--avg 1 \
--exp-dir $repo/exp/
@ -126,7 +242,6 @@ log "Run onnx_pretrained.py"
rm -rf $repo
log "--------------------------------------------------------------------------"
log "=========================================================================="
repo_url=https://huggingface.co/csukuangfj/icefall-asr-librispeech-pruned-transducer-stateless5-2022-05-13
GIT_LFS_SKIP_SMUDGE=1 git clone $repo_url
@ -143,7 +258,7 @@ popd
log "Export via torch.jit.script()"
./pruned_transducer_stateless5/export.py \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
--tokens $repo/data/lang_bpe_500/tokens.txt \
--epoch 99 \
--avg 1 \
--use-averaged-model 0 \
@ -159,7 +274,7 @@ log "Export via torch.jit.script()"
log "Test exporting to ONNX format"
./pruned_transducer_stateless5/export-onnx.py \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
--tokens $repo/data/lang_bpe_500/tokens.txt \
--epoch 99 \
--avg 1 \
--use-averaged-model 0 \
@ -205,7 +320,6 @@ 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.pt"
cd exp
@ -215,7 +329,7 @@ popd
log "Export via torch.jit.script()"
./pruned_transducer_stateless7/export.py \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
--tokens $repo/data/lang_bpe_500/tokens.txt \
--use-averaged-model 0 \
--epoch 99 \
--avg 1 \
@ -226,7 +340,7 @@ log "Export via torch.jit.script()"
log "Test exporting to ONNX format"
./pruned_transducer_stateless7/export-onnx.py \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
--tokens $repo/data/lang_bpe_500/tokens.txt \
--use-averaged-model 0 \
--epoch 99 \
--avg 1 \
@ -270,7 +384,7 @@ popd
log "Test exporting to ONNX format"
./conv_emformer_transducer_stateless2/export-onnx.py \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
--tokens $repo/data/lang_bpe_500/tokens.txt \
--use-averaged-model 0 \
--epoch 99 \
--avg 1 \
@ -310,7 +424,7 @@ popd
log "Export via torch.jit.trace()"
./lstm_transducer_stateless2/export.py \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
--tokens $repo/data/lang_bpe_500/tokens.txt \
--use-averaged-model 0 \
--epoch 99 \
--avg 1 \
@ -320,7 +434,7 @@ log "Export via torch.jit.trace()"
log "Test exporting to ONNX format"
./lstm_transducer_stateless2/export-onnx.py \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
--tokens $repo/data/lang_bpe_500/tokens.txt \
--use-averaged-model 0 \
--epoch 99 \
--avg 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:

52
.github/workflows/build-docker-image.yml vendored Normal file
View File

@ -0,0 +1,52 @@
# see also
# https://docs.github.com/en/actions/publishing-packages/publishing-docker-images#publishing-images-to-github-packages
name: Build docker image
on:
workflow_dispatch:
concurrency:
group: build_docker-${{ github.ref }}
cancel-in-progress: true
jobs:
build-docker-image:
name: ${{ matrix.image }}
runs-on: ${{ matrix.os }}
strategy:
fail-fast: false
matrix:
os: [ubuntu-latest]
image: ["torch2.2.2-cuda12.1", "torch2.2.2-cuda11.8", "torch2.2.1-cuda12.1", "torch2.2.1-cuda11.8", "torch2.2.0-cuda12.1", "torch2.2.0-cuda11.8", "torch2.1.0-cuda12.1", "torch2.1.0-cuda11.8", "torch2.0.0-cuda11.7", "torch1.13.0-cuda11.6", "torch1.12.1-cuda11.3", "torch1.9.0-cuda10.2"]
steps:
# refer to https://github.com/actions/checkout
- uses: actions/checkout@v2
with:
fetch-depth: 0
- name: Rename
shell: bash
run: |
image=${{ matrix.image }}
mv -v ./docker/$image.dockerfile ./Dockerfile
- name: Free space
shell: bash
run: |
df -h
rm -rf /opt/hostedtoolcache
df -h
- name: Log in to Docker Hub
uses: docker/login-action@v2
with:
username: ${{ secrets.DOCKER_USERNAME }}
password: ${{ secrets.DOCKER_PASSWORD }}
- name: Build and push
uses: docker/build-push-action@v4
with:
context: .
file: ./Dockerfile
push: true
tags: k2fsa/icefall:${{ matrix.image }}

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

61
.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: ${{ matrix.os }}
strategy:
matrix:
os: [ubuntu-18.04]
python-version: [3.7, 3.8, 3.9]
fail-fast: false
matrix:
os: [ubuntu-latest]
python-version: [3.8]
steps:
- uses: actions/checkout@v2
- uses: actions/checkout@v4
with:
fetch-depth: 0
@ -62,19 +47,33 @@ jobs:
with:
path: |
~/tmp/kaldifeat
key: cache-tmp-${{ matrix.python-version }}-2022-09-25
key: cache-tmp-${{ matrix.python-version }}-2023-05-22
- name: Install kaldifeat
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

112
.github/workflows/run-docker-image.yml vendored Normal file
View File

@ -0,0 +1,112 @@
name: Run docker image
on:
workflow_dispatch:
concurrency:
group: run_docker_image-${{ github.ref }}
cancel-in-progress: true
jobs:
run-docker-image:
name: ${{ matrix.image }}
runs-on: ${{ matrix.os }}
strategy:
fail-fast: false
matrix:
os: [ubuntu-latest]
image: ["torch2.2.2-cuda12.1", "torch2.2.2-cuda11.8", "torch2.2.1-cuda12.1", "torch2.2.1-cuda11.8", "torch2.2.0-cuda12.1", "torch2.2.0-cuda11.8", "torch2.1.0-cuda12.1", "torch2.1.0-cuda11.8", "torch2.0.0-cuda11.7", "torch1.13.0-cuda11.6", "torch1.12.1-cuda11.3", "torch1.9.0-cuda10.2"]
steps:
# refer to https://github.com/actions/checkout
- uses: actions/checkout@v2
with:
fetch-depth: 0
- name: Free space
shell: bash
run: |
df -h
rm -rf /opt/hostedtoolcache
df -h
- name: Run the build process with Docker
uses: addnab/docker-run-action@v3
with:
image: k2fsa/icefall:${{ matrix.image }}
shell: bash
run: |
uname -a
cat /etc/*release
find / -name libcuda* 2>/dev/null
ls -lh /usr/local/
ls -lh /usr/local/cuda*
nvcc --version
ls -lh /usr/local/cuda-*/compat/*
# For torch1.9.0-cuda10.2
export LD_LIBRARY_PATH=/usr/local/cuda-10.2/compat:$LD_LIBRARY_PATH
# For torch1.12.1-cuda11.3
export LD_LIBRARY_PATH=/usr/local/cuda-11.3/compat:$LD_LIBRARY_PATH
# For torch2.0.0-cuda11.7
export LD_LIBRARY_PATH=/usr/local/cuda-11.7/compat:$LD_LIBRARY_PATH
# For torch2.1.0-cuda11.8
export LD_LIBRARY_PATH=/usr/local/cuda-11.8/compat:$LD_LIBRARY_PATH
# For torch2.1.0-cuda12.1
export LD_LIBRARY_PATH=/usr/local/cuda-12.1/compat:$LD_LIBRARY_PATH
which nvcc
cuda_dir=$(dirname $(which nvcc))
echo "cuda_dir: $cuda_dir"
find $cuda_dir -name libcuda.so*
echo "--------------------"
find / -name libcuda.so* 2>/dev/null
# for torch1.13.0-cuda11.6
if [ -e /opt/conda/lib/stubs/libcuda.so ]; then
cd /opt/conda/lib/stubs && ln -s libcuda.so libcuda.so.1 && cd -
export LD_LIBRARY_PATH=/opt/conda/lib/stubs:$LD_LIBRARY_PATH
fi
find / -name libcuda.so* 2>/dev/null
echo "LD_LIBRARY_PATH: $LD_LIBRARY_PATH"
python3 --version
which python3
python3 -m pip list
echo "----------torch----------"
python3 -m torch.utils.collect_env
echo "----------k2----------"
python3 -c "import k2; print(k2.__file__)"
python3 -c "import k2; print(k2.__dev_version__)"
python3 -m k2.version
echo "----------lhotse----------"
python3 -c "import lhotse; print(lhotse.__file__)"
python3 -c "import lhotse; print(lhotse.__version__)"
echo "----------kaldifeat----------"
python3 -c "import kaldifeat; print(kaldifeat.__file__)"
python3 -c "import kaldifeat; print(kaldifeat.__version__)"
echo "Test yesno recipe"
cd egs/yesno/ASR
./prepare.sh
./tdnn/train.py
./tdnn/decode.py

8
.github/workflows/run-gigaspeech-2022-05-13.yml vendored
View File

@ -43,8 +43,8 @@ jobs:
runs-on: ${{ matrix.os }}
strategy:
matrix:
os: [ubuntu-18.04]
python-version: [3.7, 3.8, 3.9]
os: [ubuntu-latest]
python-version: [3.8]
fail-fast: false
@ -72,7 +72,7 @@ jobs:
with:
path: |
~/tmp/kaldifeat
key: cache-tmp-${{ matrix.python-version }}-2022-09-25
key: cache-tmp-${{ matrix.python-version }}-2023-05-22
- name: Install kaldifeat
if: steps.my-cache.outputs.cache-hit != 'true'
@ -122,5 +122,5 @@ jobs:
uses: actions/upload-artifact@v2
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
with:
name: torch-${{ matrix.torch }}-python-${{ matrix.python-version }}-ubuntu-18.04-cpu-gigaspeech-pruned_transducer_stateless2-2022-05-12
name: torch-${{ matrix.torch }}-python-${{ matrix.python-version }}-ubuntu-latest-cpu-gigaspeech-pruned_transducer_stateless2-2022-05-12
path: egs/gigaspeech/ASR/pruned_transducer_stateless2/exp/

65
.github/workflows/run-aishell-2022-06-20.yml → .github/workflows/run-gigaspeech-zipformer-2023-10-17.yml vendored
View File

@ -14,14 +14,14 @@
# 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
name: run-gigaspeech-zipformer-2023-10-17
# zipformer
on:
push:
branches:
- master
pull_request:
types: [labeled]
@ -34,18 +34,20 @@ on:
# nightly build at 15:50 UTC time every day
- cron: "50 15 * * *"
workflow_dispatch:
concurrency:
group: run_aishell_2022_06_20-${{ github.ref }}
group: run_gigaspeech_2023_10_17_zipformer-${{ 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'
run_gigaspeech_2023_10_17_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: ${{ matrix.os }}
strategy:
matrix:
os: [ubuntu-18.04]
python-version: [3.7, 3.8, 3.9]
os: [ubuntu-latest]
python-version: [3.8]
fail-fast: false
@ -73,7 +75,7 @@ jobs:
with:
path: |
~/tmp/kaldifeat
key: cache-tmp-${{ matrix.python-version }}-2022-09-25
key: cache-tmp-${{ matrix.python-version }}-2023-05-22
- name: Install kaldifeat
if: steps.my-cache.outputs.cache-hit != 'true'
@ -86,38 +88,53 @@ jobs:
env:
GITHUB_EVENT_NAME: ${{ github.event_name }}
GITHUB_EVENT_LABEL_NAME: ${{ github.event.label.name }}
HF_TOKEN: ${{ secrets.HF_TOKEN }}
run: |
mkdir -p egs/gigaspeech/ASR/data
ln -sfv ~/tmp/fbank-libri egs/gigaspeech/ASR/data/fbank
ls -lh egs/gigaspeech/ASR/data/*
sudo apt-get -qq install git-lfs tree
export PYTHONPATH=$PWD:$PYTHONPATH
export PYTHONPATH=~/tmp/kaldifeat/kaldifeat/python:$PYTHONPATH
export PYTHONPATH=~/tmp/kaldifeat/build/lib:$PYTHONPATH
.github/scripts/run-aishell-pruned-transducer-stateless3-2022-06-20.sh
.github/scripts/run-gigaspeech-zipformer-2023-10-17.sh
- name: Display decoding results for aishell pruned_transducer_stateless3
- 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
run: |
cd egs/aishell/ASR/
tree ./pruned_transducer_stateless3/exp
cd egs/gigaspeech/ASR/
tree ./zipformer/exp
cd pruned_transducer_stateless3
echo "results for pruned_transducer_stateless3"
cd zipformer
echo "results for zipformer"
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
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" {} + | sort -n -k2
find exp/fast_beam_search -name "log-*" -exec grep -n --color "best for dev" {} + | sort -n -k2
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" {} + | sort -n -k2
find exp/modified_beam_search -name "log-*" -exec grep -n --color "best for dev" {} + | sort -n -k2
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 aishell pruned_transducer_stateless3
- name: Upload decoding results for gigaspeech zipformer
uses: actions/upload-artifact@v2
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
with:
name: aishell-torch-${{ matrix.torch }}-python-${{ matrix.python-version }}-ubuntu-18.04-cpu-pruned_transducer_stateless3-2022-06-20
path: egs/aishell/ASR/pruned_transducer_stateless3/exp/
name: torch-${{ matrix.torch }}-python-${{ matrix.python-version }}-ubuntu-latest-cpu-zipformer-2022-11-11
path: egs/gigaspeech/ASR/zipformer/exp/

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-18.04]
python-version: [3.7, 3.8, 3.9]
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 }}-2022-09-25
- 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-18.04-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-18.04]
python-version: [3.7, 3.8, 3.9]
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 }}-2022-09-25
- 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-${{ matrix.python-version }}-ubuntu-18.04-cpu-pruned_transducer_stateless2-2022-04-29
path: egs/librispeech/ASR/pruned_transducer_stateless2/exp/
- name: Upload decoding results for pruned_transducer_stateless3
uses: actions/upload-artifact@v2
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
with:
name: torch-${{ matrix.torch }}-python-${{ matrix.python-version }}-ubuntu-18.04-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-18.04]
python-version: [3.7, 3.8, 3.9]
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 }}-2022-09-25
- 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-18.04-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 }}-2022-09-25
- 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-18.04-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 }}-2022-09-25
- 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-18.04-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 }}-2022-09-25
- 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-18.04-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 }}-2022-09-25
- 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-18.04-cpu-zipformer_mmi-2022-12-08
path: egs/librispeech/ASR/zipformer_mmi/exp/

163
.github/workflows/run-librispeech-2022-12-15-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-2022-12-15-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_2022_12_15_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 }}-2022-09-25
- 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:$PYTHONPATH
.github/scripts/run-librispeech-pruned-transducer-stateless7-ctc-bs-2022-12-15.sh
- name: Display decoding results for librispeech pruned_transducer_stateless7_ctc_bs
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
shell: bash
run: |
cd egs/librispeech/ASR/
tree ./pruned_transducer_stateless7_ctc_bs/exp
cd pruned_transducer_stateless7_ctc_bs
echo "results for pruned_transducer_stateless7_ctc_bs"
echo "===greedy search==="
find exp/greedy_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/greedy_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
echo "===fast_beam_search==="
find exp/fast_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/fast_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
echo "===modified beam search==="
find exp/modified_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/modified_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
echo "===ctc decoding==="
find exp/ctc-decoding -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/ctc-decoding -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
echo "===1best==="
find exp/1best -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/1best -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
- name: Upload decoding results for librispeech pruned_transducer_stateless7_ctc_bs
uses: actions/upload-artifact@v2
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
with:
name: torch-${{ matrix.torch }}-python-${{ matrix.python-version }}-ubuntu-18.04-cpu-pruned_transducer_stateless7-ctc-bs-2022-12-15
path: egs/librispeech/ASR/pruned_transducer_stateless7_ctc_bs/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 }}-2022-09-25
- 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-18.04-cpu-pruned_transducer_stateless7-streaming-2022-12-29
path: egs/librispeech/ASR/pruned_transducer_stateless7_streaming/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 }}-2022-09-25
- 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-18.04-cpu-conformer_ctc3-2022-11-28
path: egs/librispeech/ASR/conformer_ctc3/exp/

6
.github/workflows/run-librispeech-lstm-transducer-stateless2-2022-09-03.yml vendored
View File

@ -26,7 +26,7 @@ jobs:
runs-on: ${{ matrix.os }}
strategy:
matrix:
os: [ubuntu-18.04]
os: [ubuntu-latest]
python-version: [3.8]
fail-fast: false
@ -55,7 +55,7 @@ jobs:
with:
path: |
~/tmp/kaldifeat
key: cache-tmp-${{ matrix.python-version }}-2022-09-25
key: cache-tmp-${{ matrix.python-version }}-2023-05-22
- name: Install kaldifeat
if: steps.my-cache.outputs.cache-hit != 'true'
@ -159,5 +159,5 @@ jobs:
uses: actions/upload-artifact@v2
if: github.event_name == 'schedule' || github.event.label.name == 'shallow-fusion' || github.event.label.name == 'LODR'
with:
name: torch-${{ matrix.torch }}-python-${{ matrix.python-version }}-ubuntu-18.04-cpu-lstm_transducer_stateless2-2022-09-03
name: torch-${{ matrix.torch }}-python-${{ matrix.python-version }}-ubuntu-latest-cpu-lstm_transducer_stateless2-2022-09-03
path: egs/librispeech/ASR/lstm_transducer_stateless2/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-18.04]
python-version: [3.7, 3.8, 3.9]
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 }}-2022-09-25
- 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-18.04-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-18.04]
python-version: [3.7, 3.8, 3.9]
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 }}-2022-09-25
- 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-18.04-cpu-pruned_transducer_stateless2-2022-06-26
path: egs/librispeech/ASR/pruned_transducer_stateless2/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-18.04]
python-version: [3.7, 3.8, 3.9]
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 }}-2022-09-25
- 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-18.04-cpu-transducer_stateless2-2022-04-19
path: egs/librispeech/ASR/transducer_stateless2/exp/

22
.github/workflows/run-pretrained-conformer-ctc.yml → .github/workflows/run-multi-corpora-zipformer.yml vendored
View File

@ -1,4 +1,4 @@
# Copyright 2021 Fangjun Kuang (csukuangfj@gmail.com)
# Copyright 2023 Xiaomi Corp. (author: Zengrui Jin)
# See ../../LICENSE for clarification regarding multiple authors
#
@ -14,7 +14,7 @@
# See the License for the specific language governing permissions and
# limitations under the License.
name: run-pre-trained-conformer-ctc
name: run-multi-corpora-zipformer
on:
push:
@ -24,17 +24,17 @@ on:
types: [labeled]
concurrency:
group: run_pre_trained_conformer_ctc-${{ github.ref }}
group: run_multi-corpora_zipformer-${{ github.ref }}
cancel-in-progress: true
jobs:
run_pre_trained_conformer_ctc:
if: github.event.label.name == 'ready' || github.event_name == 'push'
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: ${{ matrix.os }}
strategy:
matrix:
os: [ubuntu-18.04]
python-version: [3.7, 3.8, 3.9]
os: [ubuntu-latest]
python-version: [3.8]
fail-fast: false
@ -62,7 +62,7 @@ jobs:
with:
path: |
~/tmp/kaldifeat
key: cache-tmp-${{ matrix.python-version }}-2022-09-25
key: cache-tmp-${{ matrix.python-version }}-2023-05-22
- name: Install kaldifeat
if: steps.my-cache.outputs.cache-hit != 'true'
@ -72,9 +72,13 @@ jobs:
- 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-pre-trained-conformer-ctc.sh
.github/scripts/run-multi-corpora-zipformer.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-18.04]
python-version: [3.7, 3.8, 3.9]
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 }}-2022-09-25
- 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-18.04-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-18.04]
python-version: [3.7, 3.8, 3.9]
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 }}-2022-09-25
- 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-18.04-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-18.04]
python-version: [3.7, 3.8, 3.9]
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 }}-2022-09-25
- 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

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-18.04]
python-version: [3.7, 3.8, 3.9]
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 }}-2022-09-25
- 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-18.04-cpu-transducer_stateless-2022-02-07
path: egs/librispeech/ASR/transducer_stateless/exp/

22
.github/workflows/run-pretrained-transducer-stateless-modified-aishell.yml → .github/workflows/run-swbd-conformer-ctc.yml vendored
View File

@ -1,4 +1,4 @@
# Copyright 2021 Fangjun Kuang (csukuangfj@gmail.com)
# Copyright 2023 Xiaomi Corp. (author: Zengrui Jin)
# See ../../LICENSE for clarification regarding multiple authors
#
@ -14,7 +14,7 @@
# See the License for the specific language governing permissions and
# limitations under the License.
name: run-pre-trained-trandsucer-stateless-modified-aishell
name: run-swbd-conformer_ctc
on:
push:
@ -24,17 +24,17 @@ on:
types: [labeled]
concurrency:
group: run_pre_trained_transducer_stateless_modified_aishell-${{ github.ref }}
group: run-swbd-conformer_ctc-${{ github.ref }}
cancel-in-progress: true
jobs:
run_pre_trained_transducer_stateless_modified_aishell:
if: github.event.label.name == 'ready' || github.event_name == 'push'
run-swbd-conformer_ctc:
if: github.event.label.name == 'onnx' || github.event.label.name == 'ready' || github.event_name == 'push' || github.event.label.name == 'swbd'
runs-on: ${{ matrix.os }}
strategy:
matrix:
os: [ubuntu-18.04]
python-version: [3.7, 3.8, 3.9]
os: [ubuntu-latest]
python-version: [3.8]
fail-fast: false
@ -62,7 +62,7 @@ jobs:
with:
path: |
~/tmp/kaldifeat
key: cache-tmp-${{ matrix.python-version }}-2022-09-25
key: cache-tmp-${{ matrix.python-version }}-2023-05-22
- name: Install kaldifeat
if: steps.my-cache.outputs.cache-hit != 'true'
@ -72,9 +72,13 @@ jobs:
- 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-pre-trained-transducer-stateless-modified-aishell.sh
.github/scripts/run-swbd-conformer-ctc-2023-08-26.sh

4
.github/workflows/run-wenetspeech-pruned-transducer-stateless2.yml vendored
View File

@ -33,7 +33,7 @@ jobs:
runs-on: ${{ matrix.os }}
strategy:
matrix:
os: [ubuntu-18.04]
os: [ubuntu-latest]
python-version: [3.8]
fail-fast: false
@ -62,7 +62,7 @@ jobs:
with:
path: |
~/tmp/kaldifeat
key: cache-tmp-${{ matrix.python-version }}-2022-09-25
key: cache-tmp-${{ matrix.python-version }}-2023-05-22
- name: Install kaldifeat
if: steps.my-cache.outputs.cache-hit != 'true'

84
.github/workflows/run-yesno-recipe.yml vendored
View File

@ -1,84 +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-18.04, 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: Install graphviz
shell: bash
run: |
sudo apt-get -qq install graphviz
- 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 | grep -v kaldifst | xargs -n 1 -L 1 pip install
pip uninstall -y protobuf
pip install --no-binary protobuf protobuf==3.20.*
- 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
# TODO: Check that the WER is less than some value

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 .

2
.github/workflows/test-ncnn-export.yml vendored
View File

@ -54,7 +54,7 @@ jobs:
with:
path: |
~/tmp/kaldifeat
key: cache-tmp-${{ matrix.python-version }}-2022-09-25
key: cache-tmp-${{ matrix.python-version }}-2023-05-22
- name: Install kaldifeat
if: steps.my-cache.outputs.cache-hit != 'true'

2
.github/workflows/test-onnx-export.yml vendored
View File

@ -54,7 +54,7 @@ jobs:
with:
path: |
~/tmp/kaldifeat
key: cache-tmp-${{ matrix.python-version }}-2022-09-25
key: cache-tmp-${{ matrix.python-version }}-2023-05-22
- name: Install kaldifeat
if: steps.my-cache.outputs.cache-hit != 'true'

204
.github/workflows/test.yml vendored
View File

@ -1,167 +1,111 @@
# 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: ["1.10.0"]
torchaudio: ["0.10.0"]
k2-version: ["1.23.2.dev20221201"]
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
- name: Free space
shell: bash
run: |
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:
python-version: ${{ matrix.python-version }}
image: ghcr.io/${{ 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
- 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
pytest -v -s ./test
- 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
# runt tests for conformer ctc
cd egs/librispeech/ASR/conformer_ctc
pytest -v -s
pip install k2==${{ matrix.k2-version }}+cpu.torch${{ matrix.torch }} -f https://k2-fsa.org/nightly/
pip install git+https://github.com/lhotse-speech/lhotse
# icefall requirements
pip uninstall -y protobuf
pip install --no-binary protobuf protobuf==3.20.*
cd ../pruned_transducer_stateless
pytest -v -s
pip install kaldifst
pip install onnxruntime
pip install -r requirements.txt
cd ../pruned_transducer_stateless2
pytest -v -s
- name: Install graphviz
if: startsWith(matrix.os, 'ubuntu')
shell: bash
run: |
python3 -m pip install -qq graphviz
sudo apt-get -qq install graphviz
cd ../pruned_transducer_stateless3
pytest -v -s
- name: Install graphviz
if: startsWith(matrix.os, 'macos')
shell: bash
run: |
python3 -m pip install -qq graphviz
brew install -q graphviz
cd ../pruned_transducer_stateless4
pytest -v -s
- name: Run tests
if: startsWith(matrix.os, 'ubuntu')
run: |
ls -lh
export PYTHONPATH=$PWD:$PWD/lhotse:$PYTHONPATH
echo $PYTHONPATH
pytest -v -s ./test
# runt tests for conformer ctc
cd egs/librispeech/ASR/conformer_ctc
pytest -v -s
echo $PYTHONPATH
cd ../pruned_transducer_stateless7
pytest -v -s
cd ../pruned_transducer_stateless
pytest -v -s
cd ../transducer_stateless
pytest -v -s
cd ../pruned_transducer_stateless2
pytest -v -s
# cd ../transducer
# pytest -v -s
cd ../pruned_transducer_stateless3
pytest -v -s
cd ../transducer_stateless2
pytest -v -s
cd ../pruned_transducer_stateless4
pytest -v -s
cd ../transducer_lstm
pytest -v -s
cd ../pruned_transducer_stateless7
pytest -v -s
cd ../zipformer
pytest -v -s
cd ../transducer_stateless
pytest -v -s
# cd ../transducer
# pytest -v -s
cd ../transducer_stateless2
pytest -v -s
cd ../transducer_lstm
pytest -v -s
- name: Run tests
if: startsWith(matrix.os, 'macos')
run: |
ls -lh
export PYTHONPATH=$PWD:$PWD/lhotse:$PYTHONPATH
lib_path=$(python -c "from distutils.sysconfig import get_python_lib; print(get_python_lib())")
echo "lib_path: $lib_path"
export DYLD_LIBRARY_PATH=$lib_path:$DYLD_LIBRARY_PATH
pytest -v -s ./test
# run tests for conformer ctc
cd egs/librispeech/ASR/conformer_ctc
pytest -v -s
cd ../pruned_transducer_stateless
pytest -v -s
cd ../pruned_transducer_stateless2
pytest -v -s
cd ../pruned_transducer_stateless3
pytest -v -s
cd ../pruned_transducer_stateless4
pytest -v -s
cd ../transducer_stateless
pytest -v -s
# cd ../transducer
# pytest -v -s
cd ../transducer_stateless2
pytest -v -s
cd ../transducer_lstm
pytest -v -s
- uses: actions/upload-artifact@v2
with:
path: egs/librispeech/ASR/zipformer/swoosh.pdf
name: swoosh.pdf

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
.gitignore vendored
View File

@ -34,3 +34,5 @@ node_modules
*.param
*.bin
.DS_Store
*.fst
*.arpa

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: 5.10.1
rev: 5.12.0
hooks:
- id: isort
args: ["--profile=black"]

486
README.md
View File

@ -2,43 +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]
- [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]
- [GigaSpeech][gigaspeech]
- [Aidatatang_200zh][aidatatang_200zh]
- [WenetSpeech][wenetspeech]
- [Alimeeting][alimeeting]
- [Aishell4][aishell4]
- [TAL_CSASR][tal_csasr]
- [Voxpopuli][voxpopuli]
- [XBMU-AMDO31][xbmu-amdo31]
- [WenetSpeech][wenetspeech]
More datasets will be added in the future.
### yesno
### 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:
@ -46,314 +89,291 @@ Training takes less than 30 seconds and gives you the following WER:
```
[test_set] %WER 0.42% [1 / 240, 0 ins, 1 del, 0 sub ]
```
We do 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)
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 4 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]
#### 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
#### [Transducer (Conformer Encoder + LSTM Predictor)](https://github.com/k2-fsa/icefall/tree/master/egs/librispeech/ASR/transducer)
Using Conformer as encoder and LSTM as decoder.
| | test-clean | test-other |
|---------------|------------|------------|
| greedy_search | 3.07 | 7.51 |
The best WER with greedy search is:
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)
| | test-clean | test-other |
|-----|------------|------------|
| WER | 3.07 | 7.51 |
#### [Transducer (Conformer Encoder + Stateless Predictor)](https://github.com/k2-fsa/icefall/tree/master/egs/librispeech/ASR/transducer)
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)
#### 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:
| | 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)
| | test-clean | test-other |
|---------------------------------------|------------|------------|
| modified_beam_search (`beam_size=4`) | 2.56 | 6.27 |
#### k2 pruned RNN-T
| | test-clean | test-other |
|-----|------------|------------|
| WER | 2.57 | 5.95 |
#### k2 pruned RNN-T + GigaSpeech
| | test-clean | test-other |
|-----|------------|------------|
| WER | 2.00 | 4.63 |
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)
### Aishell
#### [Transducer (Zipformer Encoder + Stateless Predictor)](https://github.com/k2-fsa/icefall/tree/master/egs/librispeech/ASR/zipformer)
We provide two models for this recipe: [conformer CTC model][Aishell_conformer_ctc]
and [TDNN LSTM CTC model][Aishell_tdnn_lstm_ctc].
WER (modified_beam_search `beam_size=4` unless further stated)
#### Conformer CTC Model
1. LibriSpeech-960hr
The best CER we currently have is:
| 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 |
| | test |
|-----|------|
| CER | 4.26 |
2. LibriSpeech-960hr + GigaSpeech
| Encoder | Params | test-clean | test-other |
|-----------------|--------|------------|------------|
| Zipformer | 65.5M | 1.78 | 4.08 |
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/1WnG17io5HEZ0Gn_cnh_VzK5QYOoiiklC?usp=sharing)
3. LibriSpeech-960hr + GigaSpeech + CommonVoice
#### Transducer Stateless Model
The best CER we currently have is:
| | test |
|-----|------|
| CER | 4.68 |
| Encoder | Params | test-clean | test-other |
|-----------------|--------|------------|------------|
| Zipformer | 65.5M | 1.90 | 3.98 |
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)
### [GigaSpeech][gigaspeech]
#### TDNN LSTM CTC Model
The CER for this model is:
| | 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)
### 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 Model
The best PER we currently have is:
||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)
#### TDNN LiGRU CTC Model
The PER for this model is:
||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/11IT-k4HQIgQngXz1uvWsEYktjqQt7Tmb?usp=sharing)
### TED-LIUM3
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:
| | dev | test |
|-----|-------|--------|
| WER | 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)
#### Pruned Transducer Stateless: Conformer encoder + Embedding decoder + k2 pruned RNN-T loss
The best WER using modified beam search with beam size 4 is:
| | dev | test |
|-----|-------|--------|
| WER | 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)
### GigaSpeech
We provide two models for this recipe: [Conformer CTC model][GigaSpeech_conformer_ctc]
and [Pruned stateless RNN-T: Conformer encoder + Embedding decoder + k2 pruned RNN-T loss][GigaSpeech_pruned_transducer_stateless2].
#### Conformer CTC
#### [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 |
### 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 (Zipformer Encoder + Stateless Predictor)](https://github.com/k2-fsa/icefall/tree/master/egs/gigaspeech/ASR/zipformer)
| | Dev | Test |
|----------------------|-------|-------|
| greedy search | 5.53 | 6.59 |
| fast beam search | 5.30 | 6.34 |
| modified beam search | 5.27 | 6.33 |
| greedy_search | 10.31 | 10.50 |
| fast_beam_search | 10.26 | 10.48 |
| modified_beam_search | 10.25 | 10.38 |
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)
### WenetSpeech
### [Aishell][aishell]
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].
#### [TDNN LSTM CTC](https://github.com/k2-fsa/icefall/tree/master/egs/aishell/ASR/tdnn_lstm_ctc)
#### Pruned stateless RNN-T_2: Conformer encoder + Embedding decoder + k2 pruned RNN-T loss (trained with L subset, offline ASR)
| | test |
|-----|-------|
| CER | 10.16 |
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 (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 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 |
### [Aishell4][aishell4]
#### [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 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]
#### [TDNN LSTM CTC](https://github.com/k2-fsa/icefall/tree/master/egs/timit/ASR/tdnn_lstm_ctc)
| |TEST|
|---|----|
|PER| 19.71% |
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](https://github.com/k2-fsa/icefall/tree/master/egs/timit/ASR/tdnn_ligru_ctc)
| |TEST|
|---|----|
|PER| 17.66% |
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][tedlium3]
#### [Transducer (Conformer Encoder + Stateless Predictor)](https://github.com/k2-fsa/icefall/tree/master/egs/tedlium3/ASR/transducer_stateless)
| | dev | test |
|--------------------------------------|-------|--------|
| modified_beam_search (`beam_size=4`) | 6.91 | 6.33 |
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)
#### [Transducer (pruned_transducer_stateless)](https://github.com/k2-fsa/icefall/tree/master/egs/tedlium3/ASR/pruned_transducer_stateless)
| | dev | test |
|--------------------------------------|-------|--------|
| modified_beam_search (`beam_size=4`) | 6.77 | 6.14 |
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]
#### [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 |
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]
#### [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)
### 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].
### [TAL_CSASR][tal_csasr]
#### Pruned stateless RNN-T: Conformer encoder + Embedding decoder + k2 pruned RNN-T loss (trained with all subsets)
The best CER(%) results:
| | test |
|----------------------|--------|
| greedy search | 29.89 |
| fast beam search | 28.91 |
| modified beam search | 29.08 |
#### [Transducer (pruned_transducer_stateless5)](https://github.com/k2-fsa/icefall/tree/master/egs/tal_csasr/ASR/pruned_transducer_stateless5)
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)
### 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
The best results for Chinese CER(%) and English WER(%) respectivly (zh: Chinese, en: English):
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/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
[Aishell_tdnn_lstm_ctc]: egs/aishell/ASR/tdnn_lstm_ctc
[Aishell_conformer_ctc]: egs/aishell/ASR/conformer_ctc
[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
[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
[Aishell4_pruned_transducer_stateless5]: egs/aishell4/ASR/pruned_transducer_stateless5
[TAL_CSASR_pruned_transducer_stateless5]: egs/tal_csasr/ASR/pruned_transducer_stateless5
[yesno]: egs/yesno/ASR
[librispeech]: egs/librispeech/ASR
[aishell]: egs/aishell/ASR
[aishell2]: egs/aishell2/ASR
[aishell4]: egs/aishell4/ASR
[timit]: egs/timit/ASR
[tedlium3]: egs/tedlium3/ASR
[gigaspeech]: egs/gigaspeech/ASR
[aidatatang_200zh]: egs/aidatatang_200zh/ASR
[wenetspeech]: egs/wenetspeech/ASR
[alimeeting]: egs/alimeeting/ASR
[aishell4]: egs/aishell4/ASR
[tal_csasr]: egs/tal_csasr/ASR
[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

58
contributing.md
View File

@ -1,39 +1,37 @@
# Contributing to Our Project
## Pre-commit hooks
Thank you for your interest in contributing to our project! We use Git pre-commit hooks to ensure code quality and consistency. Before contributing, please follow these guidelines to enable and use the pre-commit hooks.
We use [git][git] [pre-commit][pre-commit] [hooks][hooks] to check that files
going to be committed:
## Pre-Commit Hooks
- contain no trailing spaces
- are formatted with [black][black]
- are compatible to [PEP8][PEP8] (checked by [flake8][flake8])
- end in a newline and only a newline
- contain sorted `imports` (checked by [isort][isort])
We have set up pre-commit hooks to check that the files you're committing meet our coding and formatting standards. These checks include:
These hooks are disabled by default. Please use the following commands to enable them:
- Ensuring there are no trailing spaces.
- Formatting code with [black](https://github.com/psf/black).
- Checking compliance with PEP8 using [flake8](https://flake8.pycqa.org/).
- Verifying that files end with a newline character (and only a newline).
- Sorting imports using [isort](https://pycqa.github.io/isort/).
```bash
pip install pre-commit # run it only once
pre-commit install # run it only once, it will install all hooks
Please note that these hooks are disabled by default. To enable them, follow these steps:
# modify some files
git add <some files>
git commit # It runs all hooks automatically.
### Installation (Run only once)
# If all hooks run successfully, you can write the commit message now. Done!
#
# If any hook failed, your commit was not successful.
# Please read the error messages and make changes accordingly.
# And rerun
1. Install the `pre-commit` package using pip:
```bash
pip install pre-commit
```
1. Install the Git hooks using:
```bash
pre-commit install
```
### Making a Commit
Once you have enabled the pre-commit hooks, follow these steps when making a commit:
1. Make your changes to the codebase.
2. Stage your changes by using git add for the files you modified.
3. Commit your changes using git commit. The pre-commit hooks will run automatically at this point.
4. If all hooks run successfully, you can write your commit message, and your changes will be successfully committed.
5. If any hook fails, your commit will not be successful. Please read and follow the error messages provided, make the necessary changes, and then re-run git add and git commit.
git add <some files>
git commit
```
### Your Contribution
Your contributions are valuable to us, and by following these guidelines, you help maintain code consistency and quality in our project. We appreciate your dedication to ensuring high-quality code. If you have questions or need assistance, feel free to reach out to us. Thank you for being part of our open-source community!
[git]: https://git-scm.com/book/en/v2/Customizing-Git-Git-Hooks
[flake8]: https://github.com/PyCQA/flake8
[PEP8]: https://www.python.org/dev/peps/pep-0008/
[black]: https://github.com/psf/black
[hooks]: https://github.com/pre-commit/pre-commit-hooks
[pre-commit]: https://github.com/pre-commit/pre-commit
[isort]: https://github.com/PyCQA/isort

15
docker/README.md
View File

@ -1,5 +1,20 @@
# icefall dockerfile
## Download from dockerhub
You can find pre-built docker image for icefall at the following address:
<https://hub.docker.com/r/k2fsa/icefall/tags>
Example usage:
```bash
docker run --gpus all --rm -it k2fsa/icefall:torch1.13.0-cuda11.6 /bin/bash
```
## Build from dockerfile
2 sets of configuration are provided - (a) Ubuntu18.04-pytorch1.12.1-cuda11.3-cudnn8, and (b) Ubuntu18.04-pytorch1.7.1-cuda11.0-cudnn8.
If your NVIDIA driver supports CUDA Version: 11.3, please go for case (a) Ubuntu18.04-pytorch1.12.1-cuda11.3-cudnn8.

72
docker/torch1.12.1-cuda11.3.dockerfile Normal file
View File

@ -0,0 +1,72 @@
FROM pytorch/pytorch:1.12.1-cuda11.3-cudnn8-devel
ENV LC_ALL C.UTF-8
ARG DEBIAN_FRONTEND=noninteractive
# python 3.7
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>"
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

74
docker/torch1.13.0-cuda11.6.dockerfile Normal file
View File

@ -0,0 +1,74 @@
FROM pytorch/pytorch:1.13.0-cuda11.6-cudnn8-runtime
ENV LC_ALL C.UTF-8
ARG DEBIAN_FRONTEND=noninteractive
# python 3.9
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>"
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
ENV LD_LIBRARY_PATH /opt/conda/lib/stubs:$LD_LIBRARY_PATH
WORKDIR /workspace/icefall

88
docker/torch1.9.0-cuda10.2.dockerfile Normal file
View File

@ -0,0 +1,88 @@
FROM pytorch/pytorch:1.9.0-cuda10.2-cudnn7-devel
ENV LC_ALL C.UTF-8
ARG DEBIAN_FRONTEND=noninteractive
# python 3.7
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>"
LABEL k2_version=${K2_VERSION}
LABEL kaldifeat_version=${KALDIFEAT_VERSION}
LABEL github_repo="https://github.com/k2-fsa/icefall"
# see https://developer.nvidia.com/blog/updating-the-cuda-linux-gpg-repository-key/
RUN rm /etc/apt/sources.list.d/cuda.list && \
rm /etc/apt/sources.list.d/nvidia-ml.list && \
apt-key del 7fa2af80
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/*
RUN wget https://developer.download.nvidia.com/compute/cuda/repos/ubuntu1804/x86_64/cuda-keyring_1.0-1_all.deb && \
dpkg -i cuda-keyring_1.0-1_all.deb && \
rm -v cuda-keyring_1.0-1_all.deb && \
apt-get update && \
rm -rf /var/lib/apt/lists/*
# Install dependencies
RUN pip uninstall -y tqdm && \
pip install -U --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 \
kaldifeat==${KALDIFEAT_VERSION} -f https://csukuangfj.github.io/kaldifeat/cuda.html \
git+https://github.com/lhotse-speech/lhotse \
kaldi_native_io \
kaldialign \
kaldifst \
kaldilm \
sentencepiece>=0.1.96 \
tensorboard \
typeguard \
dill \
onnx \
onnxruntime \
onnxmltools \
onnxoptimizer \
onnxsim \
multi_quantization \
typeguard \
numpy \
pytest \
graphviz \
tqdm>=4.63.0
RUN git clone https://github.com/k2-fsa/icefall /workspace/icefall && \
cd /workspace/icefall && \
pip install --no-cache-dir -r requirements.txt
ENV PYTHONPATH /workspace/icefall:$PYTHONPATH
WORKDIR /workspace/icefall

73
docker/torch2.0.0-cuda11.7.dockerfile Normal file
View File

@ -0,0 +1,73 @@
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.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>"
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/torch2.1.0-cuda11.8.dockerfile Normal file
View File

@ -0,0 +1,73 @@
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.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>"
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/torch2.1.0-cuda12.1.dockerfile Normal file
View File

@ -0,0 +1,73 @@
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.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>"
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-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

9
docs/source/conf.py
View File

@ -86,7 +86,16 @@ rst_epilog = """
.. _git-lfs: https://git-lfs.com/
.. _ncnn: https://github.com/tencent/ncnn
.. _LibriSpeech: https://www.openslr.org/12
.. _Gigaspeech: https://github.com/SpeechColab/GigaSpeech
.. _musan: http://www.openslr.org/17/
.. _ONNX: https://github.com/onnx/onnx
.. _onnxruntime: https://github.com/microsoft/onnxruntime
.. _torch: https://github.com/pytorch/pytorch
.. _torchaudio: https://github.com/pytorch/audio
.. _k2: https://github.com/k2-fsa/k2
.. _lhotse: https://github.com/lhotse-speech/lhotse
.. _yesno: https://www.openslr.org/1/
.. _Next-gen Kaldi: https://github.com/k2-fsa
.. _Kaldi: https://github.com/kaldi-asr/kaldi
.. _lilcom: https://github.com/danpovey/lilcom
"""

2
docs/source/contributing/code-style.rst
View File

@ -38,7 +38,7 @@ Please fix any issues reported by the check tools.
.. HINT::
Some of the check tools, i.e., ``black`` and ``isort`` will modify
the files to be commited **in-place**. So please run ``git status``
the files to be committed **in-place**. So please run ``git status``
after failure to see which file has been modified by the tools
before you make any further changes.

2
docs/source/contributing/how-to-create-a-recipe.rst
View File

@ -3,7 +3,7 @@ How to create a recipe
.. HINT::
Please read :ref:`follow the code style` to adjust your code sytle.
Please read :ref:`follow the code style` to adjust your code style.
.. CAUTION::

187
docs/source/decoding-with-langugage-models/LODR.rst Normal file
View File

@ -0,0 +1,187 @@
.. _LODR:
LODR for RNN Transducer
=======================
As a type of E2E model, neural transducers are usually considered as having an internal
language model, which learns the language level information on the training corpus.
In real-life scenario, there is often a mismatch between the training corpus and the target corpus space.
This mismatch can be a problem when decoding for neural transducer models with language models as its internal
language can act "against" the external LM. In this tutorial, we show how to use
`Low-order Density Ratio <https://arxiv.org/abs/2203.16776>`_ to alleviate this effect to further improve the performance
of langugae model integration.
.. note::
This tutorial is based on the recipe
`pruned_transducer_stateless7_streaming <https://github.com/k2-fsa/icefall/tree/master/egs/librispeech/ASR/pruned_transducer_stateless7_streaming>`_,
which is a streaming transducer model trained on `LibriSpeech`_.
However, you can easily apply LODR to other recipes.
If you encounter any problems, please open an issue here `icefall <https://github.com/k2-fsa/icefall/issues>`__.
.. note::
For simplicity, the training and testing corpus in this tutorial are the same (`LibriSpeech`_). However,
you can change the testing set to any other domains (e.g `GigaSpeech`_) and prepare the language models
using that corpus.
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 formula for decoding with Bayes' theorem:
.. math::
\text{score}\left(y_u|\mathit{x},y\right) =
\log p\left(y_u|\mathit{x},y_{1:u-1}\right) +
\lambda_1 \log p_{\text{Target LM}}\left(y_u|\mathit{x},y_{1:u-1}\right) -
\lambda_2 \log p_{\text{Source LM}}\left(y_u|\mathit{x},y_{1:u-1}\right)
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 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
considered to be weak and can only capture low-level language information. Therefore, `LODR <https://arxiv.org/abs/2203.16776>`__ proposed to use
a low-order n-gram LM as an approximation of the ILM of the neural transducer. This leads to the following formula
during decoding for transducer model:
.. math::
\text{score}\left(y_u|\mathit{x},y\right) =
\log p_{rnnt}\left(y_u|\mathit{x},y_{1:u-1}\right) +
\lambda_1 \log p_{\text{Target LM}}\left(y_u|\mathit{x},y_{1:u-1}\right) -
\lambda_2 \log p_{\text{bi-gram}}\left(y_u|\mathit{x},y_{1:u-1}\right)
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 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``.
For illustration purpose, we will use a pre-trained ASR model from this `link <https://huggingface.co/Zengwei/icefall-asr-librispeech-pruned-transducer-stateless7-streaming-2022-12-29>`_.
If you want to train your model from scratch, please have a look at :ref:`non_streaming_librispeech_pruned_transducer_stateless`.
The testing scenario here is intra-domain (we decode the model trained on `LibriSpeech`_ on `LibriSpeech`_ testing sets).
As the initial step, let's download the pre-trained model.
.. code-block:: bash
$ GIT_LFS_SKIP_SMUDGE=1 git clone https://huggingface.co/Zengwei/icefall-asr-librispeech-pruned-transducer-stateless7-streaming-2022-12-29
$ cd icefall-asr-librispeech-pruned-transducer-stateless7-streaming-2022-12-29/exp
$ git lfs pull --include "pretrained.pt"
$ ln -s pretrained.pt epoch-99.pt # create a symbolic link so that the checkpoint can be loaded
$ cd ../data/lang_bpe_500
$ git lfs pull --include bpe.model
$ cd ../../..
To test the model, let's have a look at the decoding results **without** using LM. This can be done via the following command:
.. code-block:: bash
$ exp_dir=./icefall-asr-librispeech-pruned-transducer-stateless7-streaming-2022-12-29/exp/
$ ./pruned_transducer_stateless7_streaming/decode.py \
--epoch 99 \
--avg 1 \
--use-averaged-model False \
--exp-dir $exp_dir \
--bpe-model ./icefall-asr-librispeech-pruned-transducer-stateless7-streaming-2022-12-29/data/lang_bpe_500/bpe.model \
--max-duration 600 \
--decode-chunk-len 32 \
--decoding-method modified_beam_search
The following WERs are achieved on test-clean and test-other:
.. code-block:: text
$ For test-clean, WER of different settings are:
$ beam_size_4 3.11 best for test-clean
$ For test-other, WER of different settings are:
$ beam_size_4 7.93 best for test-other
Then, we download the external language model and bi-gram LM that are necessary for LODR.
Note that the bi-gram is estimated on the LibriSpeech 960 hours' text.
.. code-block:: bash
$ # download the external LM
$ GIT_LFS_SKIP_SMUDGE=1 git clone https://huggingface.co/ezerhouni/icefall-librispeech-rnn-lm
$ # create a symbolic link so that the checkpoint can be loaded
$ pushd icefall-librispeech-rnn-lm/exp
$ git lfs pull --include "pretrained.pt"
$ ln -s pretrained.pt epoch-99.pt
$ popd
$
$ # download the bi-gram
$ git lfs install
$ git clone https://huggingface.co/marcoyang/librispeech_bigram
$ pushd data/lang_bpe_500
$ ln -s ../../librispeech_bigram/2gram.fst.txt .
$ popd
Then, we perform LODR decoding by setting ``--decoding-method`` to ``modified_beam_search_lm_LODR``:
.. code-block:: bash
$ exp_dir=./icefall-asr-librispeech-pruned-transducer-stateless7-streaming-2022-12-29/exp
$ lm_dir=./icefall-librispeech-rnn-lm/exp
$ lm_scale=0.42
$ LODR_scale=-0.24
$ ./pruned_transducer_stateless7_streaming/decode.py \
--epoch 99 \
--avg 1 \
--use-averaged-model False \
--beam-size 4 \
--exp-dir $exp_dir \
--max-duration 600 \
--decode-chunk-len 32 \
--decoding-method modified_beam_search_LODR \
--bpe-model ./icefall-asr-librispeech-pruned-transducer-stateless7-streaming-2022-12-29/data/lang_bpe_500/bpe.model \
--use-shallow-fusion 1 \
--lm-type rnn \
--lm-exp-dir $lm_dir \
--lm-epoch 99 \
--lm-scale $lm_scale \
--lm-avg 1 \
--rnn-lm-embedding-dim 2048 \
--rnn-lm-hidden-dim 2048 \
--rnn-lm-num-layers 3 \
--lm-vocab-size 500 \
--tokens-ngram 2 \
--ngram-lm-scale $LODR_scale
There are two extra arguments that need to be given when doing LODR. ``--tokens-ngram`` specifies the order of n-gram. As we
are using a bi-gram, we set it to 2. ``--ngram-lm-scale`` is the scale of the bi-gram, it should be a negative number
as we are subtracting the bi-gram's score during decoding.
The decoding results obtained with the above command are shown below:
.. code-block:: text
$ For test-clean, WER of different settings are:
$ beam_size_4 2.61 best for test-clean
$ For test-other, WER of different settings are:
$ beam_size_4 6.74 best for test-other
Recall that the lowest WER we obtained in :ref:`shallow_fusion` with beam size of 4 is ``2.77/7.08``, LODR
indeed **further improves** the WER. We can do even better if we increase ``--beam-size``:
.. list-table:: WER of LODR with different beam sizes
:widths: 25 25 50
:header-rows: 1
* - Beam size
- test-clean
- test-other
* - 4
- 2.61
- 6.74
* - 8
- 2.45
- 6.38
* - 12
- 2.4
- 6.23

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