mirror of
https://github.com/k2-fsa/icefall.git
synced 2025-08-27 02:34:21 +00:00
Merge branch 'master' into rs
This commit is contained in:
commit
ced8a53cdc
343
.github/scripts/aishell/ASR/run.sh
vendored
Executable file
343
.github/scripts/aishell/ASR/run.sh
vendored
Executable 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 $repo
|
||||
}
|
||||
|
||||
download_test_dev_manifests
|
||||
test_transducer_stateless3_2022_06_20
|
||||
test_zipformer_large_2023_10_24
|
||||
test_zipformer_2023_10_24
|
||||
test_zipformer_small_2023_10_24
|
||||
test_transducer_stateless_modified_2022_03_01
|
||||
test_transducer_stateless_modified_2_2022_03_01
|
||||
# test_conformer_ctc # fails for torch 1.13.x and torch 2.0.x
|
63
.github/scripts/docker/Dockerfile
vendored
Normal file
63
.github/scripts/docker/Dockerfile
vendored
Normal file
@ -0,0 +1,63 @@
|
||||
ARG PYTHON_VERSION=3.8
|
||||
FROM python:${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 \
|
||||
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 \
|
||||
dill \
|
||||
graphviz \
|
||||
kaldi-decoder \
|
||||
kaldi_native_io \
|
||||
kaldialign \
|
||||
kaldifst \
|
||||
kaldilm \
|
||||
matplotlib \
|
||||
multi_quantization \
|
||||
numpy \
|
||||
onnx \
|
||||
onnxmltools \
|
||||
onnxruntime \
|
||||
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
|
79
.github/scripts/docker/generate_build_matrix.py
vendored
Executable file
79
.github/scripts/docker/generate_build_matrix.py
vendored
Executable file
@ -0,0 +1,79 @@
|
||||
#!/usr/bin/env python3
|
||||
# Copyright 2023 Xiaomi Corp. (authors: Fangjun Kuang)
|
||||
|
||||
|
||||
import json
|
||||
|
||||
|
||||
def version_gt(a, b):
|
||||
a_major, a_minor = a.split(".")[:2]
|
||||
b_major, b_minor = 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 = a.split(".")[:2]
|
||||
b_major, b_minor = 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.dev20231220"
|
||||
kaldifeat_version = "1.25.3.dev20231221"
|
||||
version = "1.2"
|
||||
python_version = ["3.8", "3.9", "3.10", "3.11"]
|
||||
torch_version = ["1.13.0", "1.13.1", "2.0.0", "2.0.1", "2.1.0", "2.1.1", "2.1.2"]
|
||||
|
||||
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
|
||||
|
||||
matrix.append(
|
||||
{
|
||||
"k2-version": k2_version,
|
||||
"kaldifeat-version": kaldifeat_version,
|
||||
"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()
|
1603
.github/scripts/librispeech/ASR/run.sh
vendored
Executable file
1603
.github/scripts/librispeech/ASR/run.sh
vendored
Executable file
File diff suppressed because it is too large
Load Diff
@ -1,87 +0,0 @@
|
||||
#!/usr/bin/env bash
|
||||
|
||||
set -e
|
||||
|
||||
log() {
|
||||
# This function is from espnet
|
||||
local fname=${BASH_SOURCE[1]##*/}
|
||||
echo -e "$(date '+%Y-%m-%d %H:%M:%S') (${fname}:${BASH_LINENO[0]}:${FUNCNAME[1]}) $*"
|
||||
}
|
||||
|
||||
cd egs/aishell/ASR
|
||||
|
||||
git lfs install
|
||||
|
||||
fbank_url=https://huggingface.co/csukuangfj/aishell-test-dev-manifests
|
||||
log "Downloading pre-commputed fbank from $fbank_url"
|
||||
|
||||
git clone https://huggingface.co/csukuangfj/aishell-test-dev-manifests
|
||||
ln -s $PWD/aishell-test-dev-manifests/data .
|
||||
|
||||
repo_url=https://huggingface.co/csukuangfj/icefall-aishell-pruned-transducer-stateless3-2022-06-20
|
||||
log "Downloading pre-trained model from $repo_url"
|
||||
git clone $repo_url
|
||||
repo=$(basename $repo_url)
|
||||
|
||||
log "Display test files"
|
||||
tree $repo/
|
||||
ls -lh $repo/test_wavs/*.wav
|
||||
|
||||
pushd $repo/exp
|
||||
ln -s pretrained-epoch-29-avg-5-torch-1.10.0.pt pretrained.pt
|
||||
popd
|
||||
|
||||
for sym in 1 2 3; do
|
||||
log "Greedy search with --max-sym-per-frame $sym"
|
||||
|
||||
./pruned_transducer_stateless3/pretrained.py \
|
||||
--method greedy_search \
|
||||
--max-sym-per-frame $sym \
|
||||
--checkpoint $repo/exp/pretrained.pt \
|
||||
--lang-dir $repo/data/lang_char \
|
||||
$repo/test_wavs/BAC009S0764W0121.wav \
|
||||
$repo/test_wavs/BAC009S0764W0122.wav \
|
||||
$repo/test_wavs/BAC009S0764W0123.wav
|
||||
done
|
||||
|
||||
for method in modified_beam_search beam_search fast_beam_search; do
|
||||
log "$method"
|
||||
|
||||
./pruned_transducer_stateless3/pretrained.py \
|
||||
--method $method \
|
||||
--beam-size 4 \
|
||||
--checkpoint $repo/exp/pretrained.pt \
|
||||
--lang-dir $repo/data/lang_char \
|
||||
$repo/test_wavs/BAC009S0764W0121.wav \
|
||||
$repo/test_wavs/BAC009S0764W0122.wav \
|
||||
$repo/test_wavs/BAC009S0764W0123.wav
|
||||
done
|
||||
|
||||
echo "GITHUB_EVENT_NAME: ${GITHUB_EVENT_NAME}"
|
||||
echo "GITHUB_EVENT_LABEL_NAME: ${GITHUB_EVENT_LABEL_NAME}"
|
||||
if [[ x"${GITHUB_EVENT_NAME}" == x"schedule" || x"${GITHUB_EVENT_LABEL_NAME}" == x"run-decode" ]]; then
|
||||
mkdir -p pruned_transducer_stateless3/exp
|
||||
ln -s $PWD/$repo/exp/pretrained.pt pruned_transducer_stateless3/exp/epoch-999.pt
|
||||
ln -s $PWD/$repo/data/lang_char data/
|
||||
|
||||
ls -lh data
|
||||
ls -lh pruned_transducer_stateless3/exp
|
||||
|
||||
log "Decoding test and dev"
|
||||
|
||||
# use a small value for decoding with CPU
|
||||
max_duration=100
|
||||
|
||||
for method in greedy_search fast_beam_search modified_beam_search; do
|
||||
log "Decoding with $method"
|
||||
|
||||
./pruned_transducer_stateless3/decode.py \
|
||||
--decoding-method $method \
|
||||
--epoch 999 \
|
||||
--avg 1 \
|
||||
--max-duration $max_duration \
|
||||
--exp-dir pruned_transducer_stateless3/exp
|
||||
done
|
||||
|
||||
rm pruned_transducer_stateless3/exp/*.pt
|
||||
fi
|
103
.github/scripts/run-aishell-zipformer-2023-10-24.sh
vendored
103
.github/scripts/run-aishell-zipformer-2023-10-24.sh
vendored
@ -1,103 +0,0 @@
|
||||
#!/usr/bin/env bash
|
||||
|
||||
set -e
|
||||
|
||||
log() {
|
||||
# This function is from espnet
|
||||
local fname=${BASH_SOURCE[1]##*/}
|
||||
echo -e "$(date '+%Y-%m-%d %H:%M:%S') (${fname}:${BASH_LINENO[0]}:${FUNCNAME[1]}) $*"
|
||||
}
|
||||
|
||||
cd egs/aishell/ASR
|
||||
|
||||
git lfs install
|
||||
|
||||
fbank_url=https://huggingface.co/csukuangfj/aishell-test-dev-manifests
|
||||
log "Downloading pre-commputed fbank from $fbank_url"
|
||||
|
||||
git clone https://huggingface.co/csukuangfj/aishell-test-dev-manifests
|
||||
ln -s $PWD/aishell-test-dev-manifests/data .
|
||||
|
||||
log "======================="
|
||||
log "CI testing large model"
|
||||
repo_url=https://huggingface.co/zrjin/icefall-asr-aishell-zipformer-large-2023-10-24/
|
||||
log "Downloading pre-trained model from $repo_url"
|
||||
git clone $repo_url
|
||||
repo=$(basename $repo_url)
|
||||
|
||||
log "Display test files"
|
||||
tree $repo/
|
||||
ls -lh $repo/test_wavs/*.wav
|
||||
|
||||
for method in modified_beam_search greedy_search fast_beam_search; do
|
||||
log "$method"
|
||||
|
||||
./zipformer/pretrained.py \
|
||||
--method $method \
|
||||
--context-size 1 \
|
||||
--checkpoint $repo/exp/pretrained.pt \
|
||||
--tokens $repo/data/lang_char/tokens.txt \
|
||||
--num-encoder-layers 2,2,4,5,4,2 \
|
||||
--feedforward-dim 512,768,1536,2048,1536,768 \
|
||||
--encoder-dim 192,256,512,768,512,256 \
|
||||
--encoder-unmasked-dim 192,192,256,320,256,192 \
|
||||
$repo/test_wavs/BAC009S0764W0121.wav \
|
||||
$repo/test_wavs/BAC009S0764W0122.wav \
|
||||
$repo/test_wavs/BAC009S0764W0123.wav
|
||||
done
|
||||
|
||||
log "======================="
|
||||
log "CI testing medium model"
|
||||
repo_url=https://huggingface.co/zrjin/icefall-asr-aishell-zipformer-2023-10-24/
|
||||
log "Downloading pre-trained model from $repo_url"
|
||||
git clone $repo_url
|
||||
repo=$(basename $repo_url)
|
||||
|
||||
log "Display test files"
|
||||
tree $repo/
|
||||
ls -lh $repo/test_wavs/*.wav
|
||||
|
||||
|
||||
for method in modified_beam_search greedy_search fast_beam_search; do
|
||||
log "$method"
|
||||
|
||||
./zipformer/pretrained.py \
|
||||
--method $method \
|
||||
--context-size 1 \
|
||||
--checkpoint $repo/exp/pretrained.pt \
|
||||
--tokens $repo/data/lang_char/tokens.txt \
|
||||
$repo/test_wavs/BAC009S0764W0121.wav \
|
||||
$repo/test_wavs/BAC009S0764W0122.wav \
|
||||
$repo/test_wavs/BAC009S0764W0123.wav
|
||||
done
|
||||
|
||||
|
||||
log "======================="
|
||||
log "CI testing small model"
|
||||
repo_url=https://huggingface.co/zrjin/icefall-asr-aishell-zipformer-small-2023-10-24/
|
||||
log "Downloading pre-trained model from $repo_url"
|
||||
git clone $repo_url
|
||||
repo=$(basename $repo_url)
|
||||
|
||||
log "Display test files"
|
||||
tree $repo/
|
||||
ls -lh $repo/test_wavs/*.wav
|
||||
|
||||
|
||||
for method in modified_beam_search greedy_search fast_beam_search; do
|
||||
log "$method"
|
||||
|
||||
./zipformer/pretrained.py \
|
||||
--method $method \
|
||||
--context-size 1 \
|
||||
--checkpoint $repo/exp/pretrained.pt \
|
||||
--tokens $repo/data/lang_char/tokens.txt \
|
||||
--num-encoder-layers 2,2,2,2,2,2 \
|
||||
--feedforward-dim 512,768,768,768,768,768 \
|
||||
--encoder-dim 192,256,256,256,256,256 \
|
||||
--encoder-unmasked-dim 192,192,192,192,192,192 \
|
||||
$repo/test_wavs/BAC009S0764W0121.wav \
|
||||
$repo/test_wavs/BAC009S0764W0122.wav \
|
||||
$repo/test_wavs/BAC009S0764W0123.wav
|
||||
done
|
||||
|
@ -1,122 +0,0 @@
|
||||
#!/usr/bin/env bash
|
||||
|
||||
set -e
|
||||
|
||||
log() {
|
||||
# This function is from espnet
|
||||
local fname=${BASH_SOURCE[1]##*/}
|
||||
echo -e "$(date '+%Y-%m-%d %H:%M:%S') (${fname}:${BASH_LINENO[0]}:${FUNCNAME[1]}) $*"
|
||||
}
|
||||
|
||||
cd egs/librispeech/ASR
|
||||
|
||||
repo_url=https://huggingface.co/Zengwei/icefall-asr-librispeech-conformer-ctc3-2022-11-27
|
||||
|
||||
log "Downloading pre-trained model from $repo_url"
|
||||
GIT_LFS_SKIP_SMUDGE=1 git clone $repo_url
|
||||
repo=$(basename $repo_url)
|
||||
|
||||
log "Display test files"
|
||||
tree $repo/
|
||||
ls -lh $repo/test_wavs/*.wav
|
||||
|
||||
pushd $repo/exp
|
||||
git lfs pull --include "data/lang_bpe_500/HLG.pt"
|
||||
git lfs pull --include "data/lang_bpe_500/L.pt"
|
||||
git lfs pull --include "data/lang_bpe_500/LG.pt"
|
||||
git lfs pull --include "data/lang_bpe_500/Linv.pt"
|
||||
git lfs pull --include "data/lang_bpe_500/bpe.model"
|
||||
git lfs pull --include "data/lm/G_4_gram.pt"
|
||||
git lfs pull --include "exp/jit_trace.pt"
|
||||
git lfs pull --include "exp/pretrained.pt"
|
||||
ln -s pretrained.pt epoch-99.pt
|
||||
ls -lh *.pt
|
||||
popd
|
||||
|
||||
log "Decode with models exported by torch.jit.trace()"
|
||||
|
||||
for m in ctc-decoding 1best; do
|
||||
./conformer_ctc3/jit_pretrained.py \
|
||||
--model-filename $repo/exp/jit_trace.pt \
|
||||
--words-file $repo/data/lang_bpe_500/words.txt \
|
||||
--HLG $repo/data/lang_bpe_500/HLG.pt \
|
||||
--bpe-model $repo/data/lang_bpe_500/bpe.model \
|
||||
--G $repo/data/lm/G_4_gram.pt \
|
||||
--method $m \
|
||||
--sample-rate 16000 \
|
||||
$repo/test_wavs/1089-134686-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0002.wav
|
||||
done
|
||||
|
||||
log "Export to torchscript model"
|
||||
|
||||
./conformer_ctc3/export.py \
|
||||
--exp-dir $repo/exp \
|
||||
--tokens $repo/data/lang_bpe_500/tokens.txt \
|
||||
--jit-trace 1 \
|
||||
--epoch 99 \
|
||||
--avg 1 \
|
||||
--use-averaged-model 0
|
||||
|
||||
ls -lh $repo/exp/*.pt
|
||||
|
||||
log "Decode with models exported by torch.jit.trace()"
|
||||
|
||||
for m in ctc-decoding 1best; do
|
||||
./conformer_ctc3/jit_pretrained.py \
|
||||
--model-filename $repo/exp/jit_trace.pt \
|
||||
--words-file $repo/data/lang_bpe_500/words.txt \
|
||||
--HLG $repo/data/lang_bpe_500/HLG.pt \
|
||||
--bpe-model $repo/data/lang_bpe_500/bpe.model \
|
||||
--G $repo/data/lm/G_4_gram.pt \
|
||||
--method $m \
|
||||
--sample-rate 16000 \
|
||||
$repo/test_wavs/1089-134686-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0002.wav
|
||||
done
|
||||
|
||||
for m in ctc-decoding 1best; do
|
||||
./conformer_ctc3/pretrained.py \
|
||||
--checkpoint $repo/exp/pretrained.pt \
|
||||
--words-file $repo/data/lang_bpe_500/words.txt \
|
||||
--HLG $repo/data/lang_bpe_500/HLG.pt \
|
||||
--tokens $repo/data/lang_bpe_500/tokens.txt \
|
||||
--G $repo/data/lm/G_4_gram.pt \
|
||||
--method $m \
|
||||
--sample-rate 16000 \
|
||||
$repo/test_wavs/1089-134686-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0002.wav
|
||||
done
|
||||
|
||||
echo "GITHUB_EVENT_NAME: ${GITHUB_EVENT_NAME}"
|
||||
echo "GITHUB_EVENT_LABEL_NAME: ${GITHUB_EVENT_LABEL_NAME}"
|
||||
if [[ x"${GITHUB_EVENT_NAME}" == x"schedule" || x"${GITHUB_EVENT_LABEL_NAME}" == x"run-decode" ]]; then
|
||||
mkdir -p conformer_ctc3/exp
|
||||
ln -s $PWD/$repo/exp/pretrained.pt conformer_ctc3/exp/epoch-999.pt
|
||||
ln -s $PWD/$repo/data/lang_bpe_500 data/
|
||||
|
||||
ls -lh data
|
||||
ls -lh conformer_ctc3/exp
|
||||
|
||||
log "Decoding test-clean and test-other"
|
||||
|
||||
# use a small value for decoding with CPU
|
||||
max_duration=100
|
||||
|
||||
for method in ctc-decoding 1best; do
|
||||
log "Decoding with $method"
|
||||
./conformer_ctc3/decode.py \
|
||||
--epoch 999 \
|
||||
--avg 1 \
|
||||
--use-averaged-model 0 \
|
||||
--exp-dir conformer_ctc3/exp/ \
|
||||
--max-duration $max_duration \
|
||||
--decoding-method $method \
|
||||
--lm-dir data/lm
|
||||
done
|
||||
|
||||
rm conformer_ctc3/exp/*.pt
|
||||
fi
|
@ -1,77 +0,0 @@
|
||||
#!/usr/bin/env bash
|
||||
|
||||
set -e
|
||||
|
||||
log() {
|
||||
# This function is from espnet
|
||||
local fname=${BASH_SOURCE[1]##*/}
|
||||
echo -e "$(date '+%Y-%m-%d %H:%M:%S') (${fname}:${BASH_LINENO[0]}:${FUNCNAME[1]}) $*"
|
||||
}
|
||||
|
||||
cd egs/librispeech/ASR
|
||||
|
||||
repo_url=https://huggingface.co/csukuangfj/icefall-asr-librispeech-pruned-transducer-stateless-2022-03-12
|
||||
|
||||
log "Downloading pre-trained model from $repo_url"
|
||||
git lfs install
|
||||
git clone $repo_url
|
||||
repo=$(basename $repo_url)
|
||||
|
||||
log "Display test files"
|
||||
tree $repo/
|
||||
ls -lh $repo/test_wavs/*.wav
|
||||
|
||||
for sym in 1 2 3; do
|
||||
log "Greedy search with --max-sym-per-frame $sym"
|
||||
|
||||
./pruned_transducer_stateless/pretrained.py \
|
||||
--method greedy_search \
|
||||
--max-sym-per-frame $sym \
|
||||
--checkpoint $repo/exp/pretrained.pt \
|
||||
--tokens $repo/data/lang_bpe_500/tokens.txt \
|
||||
$repo/test_wavs/1089-134686-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0002.wav
|
||||
done
|
||||
|
||||
for method in fast_beam_search modified_beam_search beam_search; do
|
||||
log "$method"
|
||||
|
||||
./pruned_transducer_stateless/pretrained.py \
|
||||
--method $method \
|
||||
--beam-size 4 \
|
||||
--checkpoint $repo/exp/pretrained.pt \
|
||||
--tokens $repo/data/lang_bpe_500/tokens.txt \
|
||||
$repo/test_wavs/1089-134686-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0002.wav
|
||||
done
|
||||
|
||||
echo "GITHUB_EVENT_NAME: ${GITHUB_EVENT_NAME}"
|
||||
echo "GITHUB_EVENT_LABEL_NAME: ${GITHUB_EVENT_LABEL_NAME}"
|
||||
if [[ x"${GITHUB_EVENT_NAME}" == x"schedule" || x"${GITHUB_EVENT_LABEL_NAME}" == x"run-decode" ]]; then
|
||||
mkdir -p pruned_transducer_stateless/exp
|
||||
ln -s $PWD/$repo/exp/pretrained.pt pruned_transducer_stateless/exp/epoch-999.pt
|
||||
ln -s $PWD/$repo/data/lang_bpe_500 data/
|
||||
|
||||
ls -lh data
|
||||
ls -lh pruned_transducer_stateless/exp
|
||||
|
||||
log "Decoding test-clean and test-other"
|
||||
|
||||
# use a small value for decoding with CPU
|
||||
max_duration=100
|
||||
|
||||
for method in greedy_search fast_beam_search modified_beam_search; do
|
||||
log "Decoding with $method"
|
||||
|
||||
./pruned_transducer_stateless/decode.py \
|
||||
--decoding-method $method \
|
||||
--epoch 999 \
|
||||
--avg 1 \
|
||||
--max-duration $max_duration \
|
||||
--exp-dir pruned_transducer_stateless/exp
|
||||
done
|
||||
|
||||
rm pruned_transducer_stateless/exp/*.pt
|
||||
fi
|
@ -1,86 +0,0 @@
|
||||
#!/usr/bin/env bash
|
||||
|
||||
set -e
|
||||
|
||||
log() {
|
||||
# This function is from espnet
|
||||
local fname=${BASH_SOURCE[1]##*/}
|
||||
echo -e "$(date '+%Y-%m-%d %H:%M:%S') (${fname}:${BASH_LINENO[0]}:${FUNCNAME[1]}) $*"
|
||||
}
|
||||
|
||||
cd egs/librispeech/ASR
|
||||
|
||||
repo_url=https://huggingface.co/csukuangfj/icefall-asr-librispeech-pruned-transducer-stateless2-2022-04-29
|
||||
|
||||
log "Downloading pre-trained model from $repo_url"
|
||||
GIT_LFS_SKIP_SMUDGE=1 git clone $repo_url
|
||||
repo=$(basename $repo_url)
|
||||
|
||||
pushd $repo
|
||||
git lfs pull --include "data/lang_bpe_500/bpe.model"
|
||||
git lfs pull --include "exp/pretrained-epoch-38-avg-10.pt"
|
||||
popd
|
||||
|
||||
log "Display test files"
|
||||
tree $repo/
|
||||
ls -lh $repo/test_wavs/*.wav
|
||||
|
||||
pushd $repo/exp
|
||||
ln -s pretrained-epoch-38-avg-10.pt pretrained.pt
|
||||
popd
|
||||
|
||||
for sym in 1 2 3; do
|
||||
log "Greedy search with --max-sym-per-frame $sym"
|
||||
|
||||
./pruned_transducer_stateless2/pretrained.py \
|
||||
--method greedy_search \
|
||||
--max-sym-per-frame $sym \
|
||||
--checkpoint $repo/exp/pretrained.pt \
|
||||
--tokens $repo/data/lang_bpe_500/tokens.txt \
|
||||
$repo/test_wavs/1089-134686-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0002.wav
|
||||
done
|
||||
|
||||
for method in modified_beam_search beam_search fast_beam_search; do
|
||||
log "$method"
|
||||
|
||||
./pruned_transducer_stateless2/pretrained.py \
|
||||
--method $method \
|
||||
--beam-size 4 \
|
||||
--checkpoint $repo/exp/pretrained.pt \
|
||||
--tokens $repo/data/lang_bpe_500/tokens.txt \
|
||||
$repo/test_wavs/1089-134686-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0002.wav
|
||||
done
|
||||
|
||||
echo "GITHUB_EVENT_NAME: ${GITHUB_EVENT_NAME}"
|
||||
echo "GITHUB_EVENT_LABEL_NAME: ${GITHUB_EVENT_LABEL_NAME}"
|
||||
if [[ x"${GITHUB_EVENT_NAME}" == x"schedule" || x"${GITHUB_EVENT_LABEL_NAME}" == x"run-decode" ]]; then
|
||||
mkdir -p pruned_transducer_stateless2/exp
|
||||
ln -s $PWD/$repo/exp/pretrained.pt pruned_transducer_stateless2/exp/epoch-999.pt
|
||||
ln -s $PWD/$repo/data/lang_bpe_500 data/
|
||||
|
||||
ls -lh data
|
||||
ls -lh pruned_transducer_stateless2/exp
|
||||
|
||||
log "Decoding test-clean and test-other"
|
||||
|
||||
# use a small value for decoding with CPU
|
||||
max_duration=100
|
||||
|
||||
for method in greedy_search fast_beam_search modified_beam_search; do
|
||||
log "Decoding with $method"
|
||||
|
||||
./pruned_transducer_stateless2/decode.py \
|
||||
--decoding-method $method \
|
||||
--epoch 999 \
|
||||
--avg 1 \
|
||||
--max-duration $max_duration \
|
||||
--exp-dir pruned_transducer_stateless2/exp
|
||||
done
|
||||
|
||||
rm pruned_transducer_stateless2/exp/*.pt
|
||||
rm -r data/lang_bpe_500
|
||||
fi
|
@ -1,85 +0,0 @@
|
||||
#!/usr/bin/env bash
|
||||
|
||||
set -e
|
||||
|
||||
log() {
|
||||
# This function is from espnet
|
||||
local fname=${BASH_SOURCE[1]##*/}
|
||||
echo -e "$(date '+%Y-%m-%d %H:%M:%S') (${fname}:${BASH_LINENO[0]}:${FUNCNAME[1]}) $*"
|
||||
}
|
||||
|
||||
cd egs/librispeech/ASR
|
||||
|
||||
repo_url=https://huggingface.co/csukuangfj/icefall-asr-librispeech-pruned-transducer-stateless3-2022-04-29
|
||||
|
||||
log "Downloading pre-trained model from $repo_url"
|
||||
GIT_LFS_SKIP_SMUDGE=1 git clone $repo_url
|
||||
repo=$(basename $repo_url)
|
||||
pushd $repo
|
||||
git lfs pull --include "data/lang_bpe_500/bpe.model"
|
||||
git lfs pull --include "exp/pretrained-epoch-25-avg-6.pt"
|
||||
popd
|
||||
|
||||
log "Display test files"
|
||||
tree $repo/
|
||||
ls -lh $repo/test_wavs/*.wav
|
||||
|
||||
pushd $repo/exp
|
||||
ln -s pretrained-epoch-25-avg-6.pt pretrained.pt
|
||||
popd
|
||||
|
||||
for sym in 1 2 3; do
|
||||
log "Greedy search with --max-sym-per-frame $sym"
|
||||
|
||||
./pruned_transducer_stateless3/pretrained.py \
|
||||
--method greedy_search \
|
||||
--max-sym-per-frame $sym \
|
||||
--checkpoint $repo/exp/pretrained.pt \
|
||||
--tokens $repo/data/lang_bpe_500/tokens.txt \
|
||||
$repo/test_wavs/1089-134686-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0002.wav
|
||||
done
|
||||
|
||||
for method in modified_beam_search beam_search fast_beam_search; do
|
||||
log "$method"
|
||||
|
||||
./pruned_transducer_stateless3/pretrained.py \
|
||||
--method $method \
|
||||
--beam-size 4 \
|
||||
--checkpoint $repo/exp/pretrained.pt \
|
||||
--tokens $repo/data/lang_bpe_500/tokens.txt \
|
||||
$repo/test_wavs/1089-134686-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0002.wav
|
||||
done
|
||||
|
||||
echo "GITHUB_EVENT_NAME: ${GITHUB_EVENT_NAME}"
|
||||
echo "GITHUB_EVENT_LABEL_NAME: ${GITHUB_EVENT_LABEL_NAME}"
|
||||
if [[ x"${GITHUB_EVENT_NAME}" == x"schedule" || x"${GITHUB_EVENT_LABEL_NAME}" == x"run-decode" ]]; then
|
||||
mkdir -p pruned_transducer_stateless3/exp
|
||||
ln -s $PWD/$repo/exp/pretrained.pt pruned_transducer_stateless3/exp/epoch-999.pt
|
||||
ln -s $PWD/$repo/data/lang_bpe_500 data/
|
||||
|
||||
ls -lh data
|
||||
ls -lh pruned_transducer_stateless3/exp
|
||||
|
||||
log "Decoding test-clean and test-other"
|
||||
|
||||
# use a small value for decoding with CPU
|
||||
max_duration=100
|
||||
|
||||
for method in greedy_search fast_beam_search modified_beam_search; do
|
||||
log "Decoding with $method"
|
||||
|
||||
./pruned_transducer_stateless3/decode.py \
|
||||
--decoding-method $method \
|
||||
--epoch 999 \
|
||||
--avg 1 \
|
||||
--max-duration $max_duration \
|
||||
--exp-dir pruned_transducer_stateless3/exp
|
||||
done
|
||||
|
||||
rm pruned_transducer_stateless3/exp/*.pt
|
||||
rm -r data/lang_bpe_500
|
||||
fi
|
@ -1,123 +0,0 @@
|
||||
#!/usr/bin/env bash
|
||||
|
||||
set -e
|
||||
|
||||
log() {
|
||||
# This function is from espnet
|
||||
local fname=${BASH_SOURCE[1]##*/}
|
||||
echo -e "$(date '+%Y-%m-%d %H:%M:%S') (${fname}:${BASH_LINENO[0]}:${FUNCNAME[1]}) $*"
|
||||
}
|
||||
|
||||
cd egs/librispeech/ASR
|
||||
|
||||
repo_url=https://huggingface.co/csukuangfj/icefall-asr-librispeech-pruned-transducer-stateless3-2022-05-13
|
||||
|
||||
log "Downloading pre-trained model from $repo_url"
|
||||
git lfs install
|
||||
git clone $repo_url
|
||||
repo=$(basename $repo_url)
|
||||
|
||||
log "Display test files"
|
||||
tree $repo/
|
||||
ls -lh $repo/test_wavs/*.wav
|
||||
|
||||
pushd $repo/exp
|
||||
ln -s pretrained-iter-1224000-avg-14.pt pretrained.pt
|
||||
ln -s pretrained-iter-1224000-avg-14.pt epoch-99.pt
|
||||
popd
|
||||
|
||||
|
||||
log "Export to torchscript model"
|
||||
./pruned_transducer_stateless3/export.py \
|
||||
--exp-dir $repo/exp \
|
||||
--tokens $repo/data/lang_bpe_500/tokens.txt \
|
||||
--epoch 99 \
|
||||
--avg 1 \
|
||||
--jit 1
|
||||
|
||||
./pruned_transducer_stateless3/export.py \
|
||||
--exp-dir $repo/exp \
|
||||
--tokens $repo/data/lang_bpe_500/tokens.txt \
|
||||
--epoch 99 \
|
||||
--avg 1 \
|
||||
--jit-trace 1
|
||||
|
||||
ls -lh $repo/exp/*.pt
|
||||
|
||||
log "Decode with models exported by torch.jit.trace()"
|
||||
|
||||
./pruned_transducer_stateless3/jit_pretrained.py \
|
||||
--bpe-model $repo/data/lang_bpe_500/bpe.model \
|
||||
--encoder-model-filename $repo/exp/encoder_jit_trace.pt \
|
||||
--decoder-model-filename $repo/exp/decoder_jit_trace.pt \
|
||||
--joiner-model-filename $repo/exp/joiner_jit_trace.pt \
|
||||
$repo/test_wavs/1089-134686-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0002.wav
|
||||
|
||||
log "Decode with models exported by torch.jit.script()"
|
||||
|
||||
./pruned_transducer_stateless3/jit_pretrained.py \
|
||||
--bpe-model $repo/data/lang_bpe_500/bpe.model \
|
||||
--encoder-model-filename $repo/exp/encoder_jit_script.pt \
|
||||
--decoder-model-filename $repo/exp/decoder_jit_script.pt \
|
||||
--joiner-model-filename $repo/exp/joiner_jit_script.pt \
|
||||
$repo/test_wavs/1089-134686-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0002.wav
|
||||
|
||||
|
||||
for sym in 1 2 3; do
|
||||
log "Greedy search with --max-sym-per-frame $sym"
|
||||
|
||||
./pruned_transducer_stateless3/pretrained.py \
|
||||
--method greedy_search \
|
||||
--max-sym-per-frame $sym \
|
||||
--checkpoint $repo/exp/pretrained.pt \
|
||||
--tokens $repo/data/lang_bpe_500/tokens.txt \
|
||||
$repo/test_wavs/1089-134686-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0002.wav
|
||||
done
|
||||
|
||||
for method in modified_beam_search beam_search fast_beam_search; do
|
||||
log "$method"
|
||||
|
||||
./pruned_transducer_stateless3/pretrained.py \
|
||||
--method $method \
|
||||
--beam-size 4 \
|
||||
--checkpoint $repo/exp/pretrained.pt \
|
||||
--tokens $repo/data/lang_bpe_500/tokens.txt \
|
||||
$repo/test_wavs/1089-134686-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0002.wav
|
||||
done
|
||||
|
||||
echo "GITHUB_EVENT_NAME: ${GITHUB_EVENT_NAME}"
|
||||
echo "GITHUB_EVENT_LABEL_NAME: ${GITHUB_EVENT_LABEL_NAME}"
|
||||
if [[ x"${GITHUB_EVENT_NAME}" == x"schedule" || x"${GITHUB_EVENT_LABEL_NAME}" == x"run-decode" ]]; then
|
||||
mkdir -p pruned_transducer_stateless3/exp
|
||||
ln -s $PWD/$repo/exp/pretrained.pt pruned_transducer_stateless3/exp/epoch-999.pt
|
||||
ln -s $PWD/$repo/data/lang_bpe_500 data/
|
||||
|
||||
ls -lh data
|
||||
ls -lh pruned_transducer_stateless3/exp
|
||||
|
||||
log "Decoding test-clean and test-other"
|
||||
|
||||
# use a small value for decoding with CPU
|
||||
max_duration=100
|
||||
|
||||
for method in greedy_search fast_beam_search modified_beam_search; do
|
||||
log "Decoding with $method"
|
||||
|
||||
./pruned_transducer_stateless3/decode.py \
|
||||
--decoding-method $method \
|
||||
--epoch 999 \
|
||||
--avg 1 \
|
||||
--max-duration $max_duration \
|
||||
--exp-dir pruned_transducer_stateless3/exp
|
||||
done
|
||||
|
||||
rm pruned_transducer_stateless3/exp/*.pt
|
||||
fi
|
@ -1,100 +0,0 @@
|
||||
#!/usr/bin/env bash
|
||||
|
||||
set -e
|
||||
|
||||
log() {
|
||||
# This function is from espnet
|
||||
local fname=${BASH_SOURCE[1]##*/}
|
||||
echo -e "$(date '+%Y-%m-%d %H:%M:%S') (${fname}:${BASH_LINENO[0]}:${FUNCNAME[1]}) $*"
|
||||
}
|
||||
|
||||
cd egs/librispeech/ASR
|
||||
|
||||
repo_url=https://huggingface.co/csukuangfj/icefall-asr-librispeech-pruned-transducer-stateless5-2022-05-13
|
||||
|
||||
log "Downloading pre-trained model from $repo_url"
|
||||
git lfs install
|
||||
git clone $repo_url
|
||||
repo=$(basename $repo_url)
|
||||
|
||||
log "Display test files"
|
||||
tree $repo/
|
||||
ls -lh $repo/test_wavs/*.wav
|
||||
|
||||
pushd $repo/exp
|
||||
ln -s pretrained-epoch-39-avg-7.pt pretrained.pt
|
||||
popd
|
||||
|
||||
for sym in 1 2 3; do
|
||||
log "Greedy search with --max-sym-per-frame $sym"
|
||||
|
||||
./pruned_transducer_stateless5/pretrained.py \
|
||||
--method greedy_search \
|
||||
--max-sym-per-frame $sym \
|
||||
--checkpoint $repo/exp/pretrained.pt \
|
||||
--tokens $repo/data/lang_bpe_500/tokens.txt \
|
||||
--num-encoder-layers 18 \
|
||||
--dim-feedforward 2048 \
|
||||
--nhead 8 \
|
||||
--encoder-dim 512 \
|
||||
--decoder-dim 512 \
|
||||
--joiner-dim 512 \
|
||||
$repo/test_wavs/1089-134686-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0002.wav
|
||||
done
|
||||
|
||||
for method in modified_beam_search beam_search fast_beam_search; do
|
||||
log "$method"
|
||||
|
||||
./pruned_transducer_stateless5/pretrained.py \
|
||||
--method $method \
|
||||
--beam-size 4 \
|
||||
--checkpoint $repo/exp/pretrained.pt \
|
||||
--tokens $repo/data/lang_bpe_500/tokens.txt \
|
||||
$repo/test_wavs/1089-134686-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0002.wav \
|
||||
--num-encoder-layers 18 \
|
||||
--dim-feedforward 2048 \
|
||||
--nhead 8 \
|
||||
--encoder-dim 512 \
|
||||
--decoder-dim 512 \
|
||||
--joiner-dim 512
|
||||
done
|
||||
|
||||
echo "GITHUB_EVENT_NAME: ${GITHUB_EVENT_NAME}"
|
||||
echo "GITHUB_EVENT_LABEL_NAME: ${GITHUB_EVENT_LABEL_NAME}"
|
||||
if [[ x"${GITHUB_EVENT_NAME}" == x"schedule" || x"${GITHUB_EVENT_LABEL_NAME}" == x"run-decode" ]]; then
|
||||
mkdir -p pruned_transducer_stateless5/exp
|
||||
ln -s $PWD/$repo/exp/pretrained-epoch-39-avg-7.pt pruned_transducer_stateless5/exp/epoch-999.pt
|
||||
ln -s $PWD/$repo/data/lang_bpe_500 data/
|
||||
|
||||
ls -lh data
|
||||
ls -lh pruned_transducer_stateless5/exp
|
||||
|
||||
log "Decoding test-clean and test-other"
|
||||
|
||||
# use a small value for decoding with CPU
|
||||
max_duration=100
|
||||
|
||||
for method in greedy_search fast_beam_search modified_beam_search; do
|
||||
log "Decoding with $method"
|
||||
|
||||
./pruned_transducer_stateless5/decode.py \
|
||||
--decoding-method $method \
|
||||
--use-averaged-model 0 \
|
||||
--epoch 999 \
|
||||
--avg 1 \
|
||||
--max-duration $max_duration \
|
||||
--exp-dir pruned_transducer_stateless5/exp \
|
||||
--num-encoder-layers 18 \
|
||||
--dim-feedforward 2048 \
|
||||
--nhead 8 \
|
||||
--encoder-dim 512 \
|
||||
--decoder-dim 512 \
|
||||
--joiner-dim 512
|
||||
done
|
||||
|
||||
rm pruned_transducer_stateless5/exp/*.pt
|
||||
fi
|
@ -1,106 +0,0 @@
|
||||
#!/usr/bin/env bash
|
||||
|
||||
set -e
|
||||
|
||||
log() {
|
||||
# This function is from espnet
|
||||
local fname=${BASH_SOURCE[1]##*/}
|
||||
echo -e "$(date '+%Y-%m-%d %H:%M:%S') (${fname}:${BASH_LINENO[0]}:${FUNCNAME[1]}) $*"
|
||||
}
|
||||
|
||||
cd egs/librispeech/ASR
|
||||
|
||||
repo_url=https://huggingface.co/csukuangfj/icefall-asr-librispeech-pruned-transducer-stateless7-2022-11-11
|
||||
|
||||
log "Downloading pre-trained model from $repo_url"
|
||||
git lfs install
|
||||
GIT_LFS_SKIP_SMUDGE=1 git clone $repo_url
|
||||
repo=$(basename $repo_url)
|
||||
|
||||
log "Display test files"
|
||||
tree $repo/
|
||||
ls -lh $repo/test_wavs/*.wav
|
||||
|
||||
pushd $repo/exp
|
||||
git lfs pull --include "data/lang_bpe_500/bpe.model"
|
||||
git lfs pull --include "exp/cpu_jit.pt"
|
||||
git lfs pull --include "exp/pretrained.pt"
|
||||
ln -s pretrained.pt epoch-99.pt
|
||||
ls -lh *.pt
|
||||
popd
|
||||
|
||||
log "Export to torchscript model"
|
||||
./pruned_transducer_stateless7/export.py \
|
||||
--exp-dir $repo/exp \
|
||||
--use-averaged-model false \
|
||||
--tokens $repo/data/lang_bpe_500/tokens.txt \
|
||||
--epoch 99 \
|
||||
--avg 1 \
|
||||
--jit 1
|
||||
|
||||
ls -lh $repo/exp/*.pt
|
||||
|
||||
log "Decode with models exported by torch.jit.script()"
|
||||
|
||||
./pruned_transducer_stateless7/jit_pretrained.py \
|
||||
--bpe-model $repo/data/lang_bpe_500/bpe.model \
|
||||
--nn-model-filename $repo/exp/cpu_jit.pt \
|
||||
$repo/test_wavs/1089-134686-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0002.wav
|
||||
|
||||
for sym in 1 2 3; do
|
||||
log "Greedy search with --max-sym-per-frame $sym"
|
||||
|
||||
./pruned_transducer_stateless7/pretrained.py \
|
||||
--method greedy_search \
|
||||
--max-sym-per-frame $sym \
|
||||
--checkpoint $repo/exp/pretrained.pt \
|
||||
--tokens $repo/data/lang_bpe_500/tokens.txt \
|
||||
$repo/test_wavs/1089-134686-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0002.wav
|
||||
done
|
||||
|
||||
for method in modified_beam_search beam_search fast_beam_search; do
|
||||
log "$method"
|
||||
|
||||
./pruned_transducer_stateless7/pretrained.py \
|
||||
--method $method \
|
||||
--beam-size 4 \
|
||||
--checkpoint $repo/exp/pretrained.pt \
|
||||
--tokens $repo/data/lang_bpe_500/tokens.txt \
|
||||
$repo/test_wavs/1089-134686-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0002.wav
|
||||
done
|
||||
|
||||
echo "GITHUB_EVENT_NAME: ${GITHUB_EVENT_NAME}"
|
||||
echo "GITHUB_EVENT_LABEL_NAME: ${GITHUB_EVENT_LABEL_NAME}"
|
||||
if [[ x"${GITHUB_EVENT_NAME}" == x"schedule" || x"${GITHUB_EVENT_LABEL_NAME}" == x"run-decode" ]]; then
|
||||
mkdir -p pruned_transducer_stateless7/exp
|
||||
ln -s $PWD/$repo/exp/pretrained.pt pruned_transducer_stateless7/exp/epoch-999.pt
|
||||
ln -s $PWD/$repo/data/lang_bpe_500 data/
|
||||
|
||||
ls -lh data
|
||||
ls -lh pruned_transducer_stateless7/exp
|
||||
|
||||
log "Decoding test-clean and test-other"
|
||||
|
||||
# use a small value for decoding with CPU
|
||||
max_duration=100
|
||||
|
||||
for method in greedy_search fast_beam_search modified_beam_search; do
|
||||
log "Decoding with $method"
|
||||
|
||||
./pruned_transducer_stateless7/decode.py \
|
||||
--decoding-method $method \
|
||||
--epoch 999 \
|
||||
--avg 1 \
|
||||
--use-averaged-model 0 \
|
||||
--max-duration $max_duration \
|
||||
--exp-dir pruned_transducer_stateless7/exp
|
||||
done
|
||||
|
||||
rm pruned_transducer_stateless7/exp/*.pt
|
||||
fi
|
@ -1,150 +0,0 @@
|
||||
#!/usr/bin/env bash
|
||||
|
||||
set -e
|
||||
|
||||
log() {
|
||||
# This function is from espnet
|
||||
local fname=${BASH_SOURCE[1]##*/}
|
||||
echo -e "$(date '+%Y-%m-%d %H:%M:%S') (${fname}:${BASH_LINENO[0]}:${FUNCNAME[1]}) $*"
|
||||
}
|
||||
|
||||
cd egs/librispeech/ASR
|
||||
|
||||
repo_url=https://huggingface.co/Zengwei/icefall-asr-librispeech-pruned-transducer-stateless7-ctc-2022-12-01
|
||||
|
||||
log "Downloading pre-trained model from $repo_url"
|
||||
GIT_LFS_SKIP_SMUDGE=1 git clone $repo_url
|
||||
repo=$(basename $repo_url)
|
||||
|
||||
log "Display test files"
|
||||
tree $repo/
|
||||
ls -lh $repo/test_wavs/*.wav
|
||||
|
||||
pushd $repo/exp
|
||||
git lfs pull --include "data/lang_bpe_500/HLG.pt"
|
||||
git lfs pull --include "data/lang_bpe_500/L.pt"
|
||||
git lfs pull --include "data/lang_bpe_500/LG.pt"
|
||||
git lfs pull --include "data/lang_bpe_500/Linv.pt"
|
||||
git lfs pull --include "data/lang_bpe_500/bpe.model"
|
||||
git lfs pull --include "data/lm/G_4_gram.pt"
|
||||
git lfs pull --include "exp/cpu_jit.pt"
|
||||
git lfs pull --include "exp/pretrained.pt"
|
||||
ln -s pretrained.pt epoch-99.pt
|
||||
ls -lh *.pt
|
||||
popd
|
||||
|
||||
log "Export to torchscript model"
|
||||
./pruned_transducer_stateless7_ctc/export.py \
|
||||
--exp-dir $repo/exp \
|
||||
--use-averaged-model false \
|
||||
--tokens $repo/data/lang_bpe_500/tokens.txt \
|
||||
--epoch 99 \
|
||||
--avg 1 \
|
||||
--jit 1
|
||||
|
||||
ls -lh $repo/exp/*.pt
|
||||
|
||||
log "Decode with models exported by torch.jit.script()"
|
||||
|
||||
./pruned_transducer_stateless7_ctc/jit_pretrained.py \
|
||||
--bpe-model $repo/data/lang_bpe_500/bpe.model \
|
||||
--nn-model-filename $repo/exp/cpu_jit.pt \
|
||||
$repo/test_wavs/1089-134686-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0002.wav
|
||||
|
||||
for m in ctc-decoding 1best; do
|
||||
./pruned_transducer_stateless7_ctc/jit_pretrained_ctc.py \
|
||||
--model-filename $repo/exp/cpu_jit.pt \
|
||||
--words-file $repo/data/lang_bpe_500/words.txt \
|
||||
--HLG $repo/data/lang_bpe_500/HLG.pt \
|
||||
--bpe-model $repo/data/lang_bpe_500/bpe.model \
|
||||
--G $repo/data/lm/G_4_gram.pt \
|
||||
--method $m \
|
||||
--sample-rate 16000 \
|
||||
$repo/test_wavs/1089-134686-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0002.wav
|
||||
done
|
||||
|
||||
for sym in 1 2 3; do
|
||||
log "Greedy search with --max-sym-per-frame $sym"
|
||||
|
||||
./pruned_transducer_stateless7_ctc/pretrained.py \
|
||||
--method greedy_search \
|
||||
--max-sym-per-frame $sym \
|
||||
--checkpoint $repo/exp/pretrained.pt \
|
||||
--tokens $repo/data/lang_bpe_500/tokens.txt \
|
||||
$repo/test_wavs/1089-134686-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0002.wav
|
||||
done
|
||||
|
||||
for method in modified_beam_search beam_search fast_beam_search; do
|
||||
log "$method"
|
||||
|
||||
./pruned_transducer_stateless7_ctc/pretrained.py \
|
||||
--method $method \
|
||||
--beam-size 4 \
|
||||
--checkpoint $repo/exp/pretrained.pt \
|
||||
--tokens $repo/data/lang_bpe_500/tokens.txt \
|
||||
$repo/test_wavs/1089-134686-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0002.wav
|
||||
done
|
||||
|
||||
for m in ctc-decoding 1best; do
|
||||
./pruned_transducer_stateless7_ctc/pretrained_ctc.py \
|
||||
--checkpoint $repo/exp/pretrained.pt \
|
||||
--words-file $repo/data/lang_bpe_500/words.txt \
|
||||
--HLG $repo/data/lang_bpe_500/HLG.pt \
|
||||
--bpe-model $repo/data/lang_bpe_500/bpe.model \
|
||||
--G $repo/data/lm/G_4_gram.pt \
|
||||
--method $m \
|
||||
--sample-rate 16000 \
|
||||
$repo/test_wavs/1089-134686-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0002.wav
|
||||
done
|
||||
|
||||
echo "GITHUB_EVENT_NAME: ${GITHUB_EVENT_NAME}"
|
||||
echo "GITHUB_EVENT_LABEL_NAME: ${GITHUB_EVENT_LABEL_NAME}"
|
||||
if [[ x"${GITHUB_EVENT_NAME}" == x"schedule" || x"${GITHUB_EVENT_LABEL_NAME}" == x"run-decode" ]]; then
|
||||
mkdir -p pruned_transducer_stateless7_ctc/exp
|
||||
ln -s $PWD/$repo/exp/pretrained.pt pruned_transducer_stateless7_ctc/exp/epoch-999.pt
|
||||
ln -s $PWD/$repo/data/lang_bpe_500 data/
|
||||
|
||||
ls -lh data
|
||||
ls -lh pruned_transducer_stateless7_ctc/exp
|
||||
|
||||
log "Decoding test-clean and test-other"
|
||||
|
||||
# use a small value for decoding with CPU
|
||||
max_duration=100
|
||||
|
||||
for method in greedy_search fast_beam_search modified_beam_search; do
|
||||
log "Decoding with $method"
|
||||
|
||||
./pruned_transducer_stateless7_ctc/decode.py \
|
||||
--decoding-method $method \
|
||||
--epoch 999 \
|
||||
--avg 1 \
|
||||
--use-averaged-model 0 \
|
||||
--max-duration $max_duration \
|
||||
--exp-dir pruned_transducer_stateless7_ctc/exp
|
||||
done
|
||||
|
||||
for m in ctc-decoding 1best; do
|
||||
./pruned_transducer_stateless7_ctc/ctc_decode.py \
|
||||
--epoch 999 \
|
||||
--avg 1 \
|
||||
--exp-dir ./pruned_transducer_stateless7_ctc/exp \
|
||||
--max-duration $max_duration \
|
||||
--use-averaged-model 0 \
|
||||
--decoding-method $m \
|
||||
--hlg-scale 0.6 \
|
||||
--lm-dir data/lm
|
||||
done
|
||||
|
||||
rm pruned_transducer_stateless7_ctc/exp/*.pt
|
||||
fi
|
@ -1,147 +0,0 @@
|
||||
#!/usr/bin/env bash
|
||||
|
||||
set -e
|
||||
|
||||
log() {
|
||||
# This function is from espnet
|
||||
local fname=${BASH_SOURCE[1]##*/}
|
||||
echo -e "$(date '+%Y-%m-%d %H:%M:%S') (${fname}:${BASH_LINENO[0]}:${FUNCNAME[1]}) $*"
|
||||
}
|
||||
|
||||
cd egs/librispeech/ASR
|
||||
|
||||
repo_url=https://huggingface.co/yfyeung/icefall-asr-librispeech-pruned_transducer_stateless7_ctc_bs-2023-01-29
|
||||
|
||||
log "Downloading pre-trained model from $repo_url"
|
||||
GIT_LFS_SKIP_SMUDGE=1 git clone $repo_url
|
||||
repo=$(basename $repo_url)
|
||||
|
||||
log "Display test files"
|
||||
tree $repo/
|
||||
ls -lh $repo/test_wavs/*.wav
|
||||
|
||||
pushd $repo/exp
|
||||
git lfs pull --include "data/lang_bpe_500/L.pt"
|
||||
git lfs pull --include "data/lang_bpe_500/LG.pt"
|
||||
git lfs pull --include "data/lang_bpe_500/HLG.pt"
|
||||
git lfs pull --include "data/lang_bpe_500/Linv.pt"
|
||||
git lfs pull --include "data/lang_bpe_500/bpe.model"
|
||||
git lfs pull --include "exp/cpu_jit.pt"
|
||||
git lfs pull --include "exp/pretrained.pt"
|
||||
ln -s pretrained.pt epoch-99.pt
|
||||
ls -lh *.pt
|
||||
popd
|
||||
|
||||
log "Export to torchscript model"
|
||||
./pruned_transducer_stateless7_ctc_bs/export.py \
|
||||
--exp-dir $repo/exp \
|
||||
--use-averaged-model false \
|
||||
--tokens $repo/data/lang_bpe_500/tokens.txt \
|
||||
--epoch 99 \
|
||||
--avg 1 \
|
||||
--jit 1
|
||||
|
||||
ls -lh $repo/exp/*.pt
|
||||
|
||||
log "Decode with models exported by torch.jit.script()"
|
||||
|
||||
./pruned_transducer_stateless7_ctc_bs/jit_pretrained.py \
|
||||
--bpe-model $repo/data/lang_bpe_500/bpe.model \
|
||||
--nn-model-filename $repo/exp/cpu_jit.pt \
|
||||
$repo/test_wavs/1089-134686-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0002.wav
|
||||
|
||||
for m in ctc-decoding 1best; do
|
||||
./pruned_transducer_stateless7_ctc_bs/jit_pretrained_ctc.py \
|
||||
--model-filename $repo/exp/cpu_jit.pt \
|
||||
--words-file $repo/data/lang_bpe_500/words.txt \
|
||||
--HLG $repo/data/lang_bpe_500/HLG.pt \
|
||||
--bpe-model $repo/data/lang_bpe_500/bpe.model \
|
||||
--method $m \
|
||||
--sample-rate 16000 \
|
||||
$repo/test_wavs/1089-134686-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0002.wav
|
||||
done
|
||||
|
||||
for sym in 1 2 3; do
|
||||
log "Greedy search with --max-sym-per-frame $sym"
|
||||
|
||||
./pruned_transducer_stateless7_ctc_bs/pretrained.py \
|
||||
--method greedy_search \
|
||||
--max-sym-per-frame $sym \
|
||||
--checkpoint $repo/exp/pretrained.pt \
|
||||
--tokens $repo/data/lang_bpe_500/tokens.txt \
|
||||
$repo/test_wavs/1089-134686-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0002.wav
|
||||
done
|
||||
|
||||
for method in modified_beam_search beam_search fast_beam_search; do
|
||||
log "$method"
|
||||
|
||||
./pruned_transducer_stateless7_ctc_bs/pretrained.py \
|
||||
--method $method \
|
||||
--beam-size 4 \
|
||||
--checkpoint $repo/exp/pretrained.pt \
|
||||
--tokens $repo/data/lang_bpe_500/tokens.txt \
|
||||
$repo/test_wavs/1089-134686-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0002.wav
|
||||
done
|
||||
|
||||
for m in ctc-decoding 1best; do
|
||||
./pruned_transducer_stateless7_ctc_bs/pretrained_ctc.py \
|
||||
--checkpoint $repo/exp/pretrained.pt \
|
||||
--words-file $repo/data/lang_bpe_500/words.txt \
|
||||
--HLG $repo/data/lang_bpe_500/HLG.pt \
|
||||
--bpe-model $repo/data/lang_bpe_500/bpe.model \
|
||||
--method $m \
|
||||
--sample-rate 16000 \
|
||||
$repo/test_wavs/1089-134686-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0002.wav
|
||||
done
|
||||
|
||||
echo "GITHUB_EVENT_NAME: ${GITHUB_EVENT_NAME}"
|
||||
echo "GITHUB_EVENT_LABEL_NAME: ${GITHUB_EVENT_LABEL_NAME}"
|
||||
|
||||
if [[ x"${GITHUB_EVENT_NAME}" == x"schedule" || x"${GITHUB_EVENT_LABEL_NAME}" == x"run-decode" ]]; then
|
||||
mkdir -p pruned_transducer_stateless7_ctc_bs/exp
|
||||
ln -s $PWD/$repo/exp/pretrained.pt pruned_transducer_stateless7_ctc_bs/exp/epoch-999.pt
|
||||
ln -s $PWD/$repo/data/lang_bpe_500 data/
|
||||
|
||||
ls -lh data
|
||||
ls -lh pruned_transducer_stateless7_ctc_bs/exp
|
||||
|
||||
log "Decoding test-clean and test-other"
|
||||
|
||||
# use a small value for decoding with CPU
|
||||
max_duration=100
|
||||
|
||||
for method in greedy_search fast_beam_search modified_beam_search; do
|
||||
log "Decoding with $method"
|
||||
|
||||
./pruned_transducer_stateless7_ctc_bs/decode.py \
|
||||
--decoding-method $method \
|
||||
--epoch 999 \
|
||||
--avg 1 \
|
||||
--use-averaged-model 0 \
|
||||
--max-duration $max_duration \
|
||||
--exp-dir pruned_transducer_stateless7_ctc_bs/exp
|
||||
done
|
||||
|
||||
for m in ctc-decoding 1best; do
|
||||
./pruned_transducer_stateless7_ctc_bs/ctc_decode.py \
|
||||
--epoch 999 \
|
||||
--avg 1 \
|
||||
--exp-dir ./pruned_transducer_stateless7_ctc_bs/exp \
|
||||
--max-duration $max_duration \
|
||||
--use-averaged-model 0 \
|
||||
--decoding-method $m \
|
||||
--hlg-scale 0.6
|
||||
done
|
||||
|
||||
rm pruned_transducer_stateless7_ctc_bs/exp/*.pt
|
||||
fi
|
@ -1,148 +0,0 @@
|
||||
#!/usr/bin/env bash
|
||||
|
||||
set -e
|
||||
|
||||
log() {
|
||||
# This function is from espnet
|
||||
local fname=${BASH_SOURCE[1]##*/}
|
||||
echo -e "$(date '+%Y-%m-%d %H:%M:%S') (${fname}:${BASH_LINENO[0]}:${FUNCNAME[1]}) $*"
|
||||
}
|
||||
|
||||
cd egs/librispeech/ASR
|
||||
|
||||
repo_url=https://huggingface.co/Zengwei/icefall-asr-librispeech-pruned-transducer-stateless7-streaming-2022-12-29
|
||||
|
||||
log "Downloading pre-trained model from $repo_url"
|
||||
git lfs install
|
||||
GIT_LFS_SKIP_SMUDGE=1 git clone $repo_url
|
||||
repo=$(basename $repo_url)
|
||||
|
||||
log "Display test files"
|
||||
tree $repo/
|
||||
ls -lh $repo/test_wavs/*.wav
|
||||
|
||||
pushd $repo
|
||||
git lfs pull --include "data/lang_bpe_500/bpe.model"
|
||||
git lfs pull --include "exp/cpu_jit.pt"
|
||||
git lfs pull --include "exp/pretrained.pt"
|
||||
git lfs pull --include "exp/encoder_jit_trace.pt"
|
||||
git lfs pull --include "exp/decoder_jit_trace.pt"
|
||||
git lfs pull --include "exp/joiner_jit_trace.pt"
|
||||
cd exp
|
||||
ln -s pretrained.pt epoch-99.pt
|
||||
ls -lh *.pt
|
||||
popd
|
||||
|
||||
log "Export to torchscript model"
|
||||
./pruned_transducer_stateless7_streaming/export.py \
|
||||
--exp-dir $repo/exp \
|
||||
--use-averaged-model false \
|
||||
--tokens $repo/data/lang_bpe_500/tokens.txt \
|
||||
--decode-chunk-len 32 \
|
||||
--epoch 99 \
|
||||
--avg 1 \
|
||||
--jit 1
|
||||
|
||||
ls -lh $repo/exp/*.pt
|
||||
|
||||
log "Decode with models exported by torch.jit.script()"
|
||||
|
||||
./pruned_transducer_stateless7_streaming/jit_pretrained.py \
|
||||
--bpe-model $repo/data/lang_bpe_500/bpe.model \
|
||||
--nn-model-filename $repo/exp/cpu_jit.pt \
|
||||
--decode-chunk-len 32 \
|
||||
$repo/test_wavs/1089-134686-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0002.wav
|
||||
|
||||
log "Export to torchscript model by torch.jit.trace()"
|
||||
./pruned_transducer_stateless7_streaming/jit_trace_export.py \
|
||||
--exp-dir $repo/exp \
|
||||
--use-averaged-model false \
|
||||
--bpe-model $repo/data/lang_bpe_500/bpe.model \
|
||||
--decode-chunk-len 32 \
|
||||
--epoch 99 \
|
||||
--avg 1
|
||||
|
||||
log "Decode with models exported by torch.jit.trace()"
|
||||
|
||||
./pruned_transducer_stateless7_streaming/jit_trace_pretrained.py \
|
||||
--bpe-model $repo/data/lang_bpe_500/bpe.model \
|
||||
--encoder-model-filename $repo/exp/encoder_jit_trace.pt \
|
||||
--decoder-model-filename $repo/exp/decoder_jit_trace.pt \
|
||||
--joiner-model-filename $repo/exp/joiner_jit_trace.pt \
|
||||
--decode-chunk-len 32 \
|
||||
$repo/test_wavs/1089-134686-0001.wav
|
||||
|
||||
for sym in 1 2 3; do
|
||||
log "Greedy search with --max-sym-per-frame $sym"
|
||||
|
||||
./pruned_transducer_stateless7_streaming/pretrained.py \
|
||||
--method greedy_search \
|
||||
--max-sym-per-frame $sym \
|
||||
--checkpoint $repo/exp/pretrained.pt \
|
||||
--tokens $repo/data/lang_bpe_500/tokens.txt \
|
||||
--decode-chunk-len 32 \
|
||||
$repo/test_wavs/1089-134686-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0002.wav
|
||||
done
|
||||
|
||||
for method in modified_beam_search beam_search fast_beam_search; do
|
||||
log "$method"
|
||||
|
||||
./pruned_transducer_stateless7_streaming/pretrained.py \
|
||||
--method $method \
|
||||
--beam-size 4 \
|
||||
--checkpoint $repo/exp/pretrained.pt \
|
||||
--tokens $repo/data/lang_bpe_500/tokens.txt \
|
||||
--decode-chunk-len 32 \
|
||||
$repo/test_wavs/1089-134686-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0002.wav
|
||||
done
|
||||
|
||||
echo "GITHUB_EVENT_NAME: ${GITHUB_EVENT_NAME}"
|
||||
echo "GITHUB_EVENT_LABEL_NAME: ${GITHUB_EVENT_LABEL_NAME}"
|
||||
if [[ x"${GITHUB_EVENT_NAME}" == x"schedule" || x"${GITHUB_EVENT_LABEL_NAME}" == x"run-decode" ]]; then
|
||||
mkdir -p pruned_transducer_stateless7_streaming/exp
|
||||
ln -s $PWD/$repo/exp/pretrained.pt pruned_transducer_stateless7_streaming/exp/epoch-999.pt
|
||||
ln -s $PWD/$repo/data/lang_bpe_500 data/
|
||||
|
||||
ls -lh data
|
||||
ls -lh pruned_transducer_stateless7_streaming/exp
|
||||
|
||||
log "Decoding test-clean and test-other"
|
||||
|
||||
# use a small value for decoding with CPU
|
||||
max_duration=100
|
||||
num_decode_stream=200
|
||||
|
||||
for method in greedy_search fast_beam_search modified_beam_search; do
|
||||
log "decoding with $method"
|
||||
|
||||
./pruned_transducer_stateless7_streaming/decode.py \
|
||||
--decoding-method $method \
|
||||
--epoch 999 \
|
||||
--avg 1 \
|
||||
--use-averaged-model 0 \
|
||||
--max-duration $max_duration \
|
||||
--decode-chunk-len 32 \
|
||||
--exp-dir pruned_transducer_stateless7_streaming/exp
|
||||
done
|
||||
|
||||
for method in greedy_search fast_beam_search modified_beam_search; do
|
||||
log "Decoding with $method"
|
||||
|
||||
./pruned_transducer_stateless7_streaming/streaming_decode.py \
|
||||
--decoding-method $method \
|
||||
--epoch 999 \
|
||||
--avg 1 \
|
||||
--use-averaged-model 0 \
|
||||
--decode-chunk-len 32 \
|
||||
--num-decode-streams $num_decode_stream
|
||||
--exp-dir pruned_transducer_stateless7_streaming/exp
|
||||
done
|
||||
|
||||
rm pruned_transducer_stateless7_streaming/exp/*.pt
|
||||
fi
|
@ -1,115 +0,0 @@
|
||||
#!/usr/bin/env bash
|
||||
|
||||
set -e
|
||||
|
||||
log() {
|
||||
# This function is from espnet
|
||||
local fname=${BASH_SOURCE[1]##*/}
|
||||
echo -e "$(date '+%Y-%m-%d %H:%M:%S') (${fname}:${BASH_LINENO[0]}:${FUNCNAME[1]}) $*"
|
||||
}
|
||||
|
||||
cd egs/librispeech/ASR
|
||||
|
||||
repo_url=https://huggingface.co/csukuangfj/icefall-asr-librispeech-pruned-transducer-stateless8-2022-11-14
|
||||
|
||||
log "Downloading pre-trained model from $repo_url"
|
||||
git lfs install
|
||||
GIT_LFS_SKIP_SMUDGE=1 git clone $repo_url
|
||||
repo=$(basename $repo_url)
|
||||
|
||||
log "Display test files"
|
||||
tree $repo/
|
||||
ls -lh $repo/test_wavs/*.wav
|
||||
|
||||
pushd $repo/exp
|
||||
git lfs pull --include "data/lang_bpe_500/bpe.model"
|
||||
git lfs pull --include "exp/cpu_jit.pt"
|
||||
git lfs pull --include "exp/pretrained.pt"
|
||||
ln -s pretrained.pt epoch-99.pt
|
||||
ls -lh *.pt
|
||||
popd
|
||||
|
||||
log "Decode with models exported by torch.jit.script()"
|
||||
|
||||
./pruned_transducer_stateless8/jit_pretrained.py \
|
||||
--bpe-model $repo/data/lang_bpe_500/bpe.model \
|
||||
--nn-model-filename $repo/exp/cpu_jit.pt \
|
||||
$repo/test_wavs/1089-134686-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0002.wav
|
||||
|
||||
log "Export to torchscript model"
|
||||
./pruned_transducer_stateless8/export.py \
|
||||
--exp-dir $repo/exp \
|
||||
--tokens $repo/data/lang_bpe_500/tokens.txt \
|
||||
--use-averaged-model false \
|
||||
--epoch 99 \
|
||||
--avg 1 \
|
||||
--jit 1
|
||||
|
||||
ls -lh $repo/exp/*.pt
|
||||
|
||||
log "Decode with models exported by torch.jit.script()"
|
||||
|
||||
./pruned_transducer_stateless8/jit_pretrained.py \
|
||||
--bpe-model $repo/data/lang_bpe_500/bpe.model \
|
||||
--nn-model-filename $repo/exp/cpu_jit.pt \
|
||||
$repo/test_wavs/1089-134686-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0002.wav
|
||||
|
||||
for sym in 1 2 3; do
|
||||
log "Greedy search with --max-sym-per-frame $sym"
|
||||
|
||||
./pruned_transducer_stateless8/pretrained.py \
|
||||
--method greedy_search \
|
||||
--max-sym-per-frame $sym \
|
||||
--checkpoint $repo/exp/pretrained.pt \
|
||||
--tokens $repo/data/lang_bpe_500/tokens.txt \
|
||||
$repo/test_wavs/1089-134686-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0002.wav
|
||||
done
|
||||
|
||||
for method in modified_beam_search beam_search fast_beam_search; do
|
||||
log "$method"
|
||||
|
||||
./pruned_transducer_stateless8/pretrained.py \
|
||||
--method $method \
|
||||
--beam-size 4 \
|
||||
--checkpoint $repo/exp/pretrained.pt \
|
||||
--tokens $repo/data/lang_bpe_500/tokens.txt \
|
||||
$repo/test_wavs/1089-134686-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0002.wav
|
||||
done
|
||||
|
||||
echo "GITHUB_EVENT_NAME: ${GITHUB_EVENT_NAME}"
|
||||
echo "GITHUB_EVENT_LABEL_NAME: ${GITHUB_EVENT_LABEL_NAME}"
|
||||
if [[ x"${GITHUB_EVENT_NAME}" == x"schedule" || x"${GITHUB_EVENT_LABEL_NAME}" == x"run-decode" ]]; then
|
||||
mkdir -p pruned_transducer_stateless8/exp
|
||||
ln -s $PWD/$repo/exp/pretrained.pt pruned_transducer_stateless8/exp/epoch-999.pt
|
||||
ln -s $PWD/$repo/data/lang_bpe_500 data/
|
||||
|
||||
ls -lh data
|
||||
ls -lh pruned_transducer_stateless8/exp
|
||||
|
||||
log "Decoding test-clean and test-other"
|
||||
|
||||
# use a small value for decoding with CPU
|
||||
max_duration=100
|
||||
|
||||
for method in greedy_search fast_beam_search modified_beam_search; do
|
||||
log "Decoding with $method"
|
||||
|
||||
./pruned_transducer_stateless8/decode.py \
|
||||
--decoding-method $method \
|
||||
--epoch 999 \
|
||||
--avg 1 \
|
||||
--use-averaged-model 0 \
|
||||
--max-duration $max_duration \
|
||||
--exp-dir pruned_transducer_stateless8/exp
|
||||
done
|
||||
|
||||
rm pruned_transducer_stateless8/exp/*.pt
|
||||
fi
|
@ -1,101 +0,0 @@
|
||||
#!/usr/bin/env bash
|
||||
|
||||
set -e
|
||||
|
||||
log() {
|
||||
# This function is from espnet
|
||||
local fname=${BASH_SOURCE[1]##*/}
|
||||
echo -e "$(date '+%Y-%m-%d %H:%M:%S') (${fname}:${BASH_LINENO[0]}:${FUNCNAME[1]}) $*"
|
||||
}
|
||||
|
||||
cd egs/librispeech/ASR
|
||||
|
||||
repo_url=https://huggingface.co/pkufool/icefall_librispeech_streaming_pruned_transducer_stateless2_20220625
|
||||
|
||||
log "Downloading pre-trained model from $repo_url"
|
||||
git lfs install
|
||||
git clone $repo_url
|
||||
repo=$(basename $repo_url)
|
||||
|
||||
log "Display test files"
|
||||
tree $repo/
|
||||
ls -lh $repo/test_wavs/*.wav
|
||||
|
||||
pushd $repo/exp
|
||||
ln -s pretrained-epoch-24-avg-10.pt pretrained.pt
|
||||
popd
|
||||
|
||||
for sym in 1 2 3; do
|
||||
log "Greedy search with --max-sym-per-frame $sym"
|
||||
|
||||
./pruned_transducer_stateless2/pretrained.py \
|
||||
--method greedy_search \
|
||||
--max-sym-per-frame $sym \
|
||||
--checkpoint $repo/exp/pretrained.pt \
|
||||
--tokens $repo/data/lang_bpe_500/tokens.txt \
|
||||
--simulate-streaming 1 \
|
||||
--causal-convolution 1 \
|
||||
$repo/test_wavs/1089-134686-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0002.wav
|
||||
done
|
||||
|
||||
for method in modified_beam_search beam_search fast_beam_search; do
|
||||
log "$method"
|
||||
|
||||
./pruned_transducer_stateless2/pretrained.py \
|
||||
--method $method \
|
||||
--beam-size 4 \
|
||||
--checkpoint $repo/exp/pretrained.pt \
|
||||
--tokens $repo/data/lang_bpe_500/tokens.txt \
|
||||
--simulate-streaming 1 \
|
||||
--causal-convolution 1 \
|
||||
$repo/test_wavs/1089-134686-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0002.wav
|
||||
done
|
||||
|
||||
echo "GITHUB_EVENT_NAME: ${GITHUB_EVENT_NAME}"
|
||||
echo "GITHUB_EVENT_LABEL_NAME: ${GITHUB_EVENT_LABEL_NAME}"
|
||||
if [[ x"${GITHUB_EVENT_NAME}" == x"schedule" || x"${GITHUB_EVENT_LABEL_NAME}" == x"run-decode" ]]; then
|
||||
mkdir -p pruned_transducer_stateless2/exp
|
||||
ln -s $PWD/$repo/exp/pretrained-epoch-24-avg-10.pt pruned_transducer_stateless2/exp/epoch-999.pt
|
||||
ln -s $PWD/$repo/data/lang_bpe_500 data/
|
||||
|
||||
ls -lh data
|
||||
ls -lh pruned_transducer_stateless2/exp
|
||||
|
||||
log "Decoding test-clean and test-other"
|
||||
|
||||
# use a small value for decoding with CPU
|
||||
max_duration=100
|
||||
|
||||
for method in greedy_search fast_beam_search modified_beam_search; do
|
||||
log "Simulate streaming decoding with $method"
|
||||
|
||||
./pruned_transducer_stateless2/decode.py \
|
||||
--decoding-method $method \
|
||||
--epoch 999 \
|
||||
--avg 1 \
|
||||
--max-duration $max_duration \
|
||||
--exp-dir pruned_transducer_stateless2/exp \
|
||||
--simulate-streaming 1 \
|
||||
--causal-convolution 1
|
||||
done
|
||||
|
||||
for method in greedy_search fast_beam_search modified_beam_search; do
|
||||
log "Real streaming decoding with $method"
|
||||
|
||||
./pruned_transducer_stateless2/streaming_decode.py \
|
||||
--decoding-method $method \
|
||||
--epoch 999 \
|
||||
--avg 1 \
|
||||
--num-decode-streams 100 \
|
||||
--exp-dir pruned_transducer_stateless2/exp \
|
||||
--left-context 32 \
|
||||
--decode-chunk-size 8 \
|
||||
--right-context 0
|
||||
done
|
||||
|
||||
rm pruned_transducer_stateless2/exp/*.pt
|
||||
fi
|
@ -1,116 +0,0 @@
|
||||
#!/usr/bin/env bash
|
||||
|
||||
set -e
|
||||
|
||||
log() {
|
||||
# This function is from espnet
|
||||
local fname=${BASH_SOURCE[1]##*/}
|
||||
echo -e "$(date '+%Y-%m-%d %H:%M:%S') (${fname}:${BASH_LINENO[0]}:${FUNCNAME[1]}) $*"
|
||||
}
|
||||
|
||||
cd egs/librispeech/ASR
|
||||
|
||||
repo_url=https://huggingface.co/Zengwei/icefall-asr-librispeech-streaming-zipformer-2023-05-17
|
||||
|
||||
log "Downloading pre-trained model from $repo_url"
|
||||
git lfs install
|
||||
GIT_LFS_SKIP_SMUDGE=1 git clone $repo_url
|
||||
repo=$(basename $repo_url)
|
||||
|
||||
log "Display test files"
|
||||
tree $repo/
|
||||
ls -lh $repo/test_wavs/*.wav
|
||||
|
||||
pushd $repo/exp
|
||||
git lfs pull --include "data/lang_bpe_500/bpe.model"
|
||||
git lfs pull --include "data/lang_bpe_500/tokens.txt"
|
||||
git lfs pull --include "exp/jit_script_chunk_16_left_128.pt"
|
||||
git lfs pull --include "exp/pretrained.pt"
|
||||
ln -s pretrained.pt epoch-99.pt
|
||||
ls -lh *.pt
|
||||
popd
|
||||
|
||||
log "Export to torchscript model"
|
||||
./zipformer/export.py \
|
||||
--exp-dir $repo/exp \
|
||||
--use-averaged-model false \
|
||||
--tokens $repo/data/lang_bpe_500/tokens.txt \
|
||||
--causal 1 \
|
||||
--chunk-size 16 \
|
||||
--left-context-frames 128 \
|
||||
--epoch 99 \
|
||||
--avg 1 \
|
||||
--jit 1
|
||||
|
||||
ls -lh $repo/exp/*.pt
|
||||
|
||||
log "Decode with models exported by torch.jit.script()"
|
||||
|
||||
./zipformer/jit_pretrained_streaming.py \
|
||||
--tokens $repo/data/lang_bpe_500/tokens.txt \
|
||||
--nn-model-filename $repo/exp/jit_script_chunk_16_left_128.pt \
|
||||
$repo/test_wavs/1089-134686-0001.wav
|
||||
|
||||
for method in greedy_search modified_beam_search fast_beam_search; do
|
||||
log "$method"
|
||||
|
||||
./zipformer/pretrained.py \
|
||||
--causal 1 \
|
||||
--chunk-size 16 \
|
||||
--left-context-frames 128 \
|
||||
--method $method \
|
||||
--beam-size 4 \
|
||||
--checkpoint $repo/exp/pretrained.pt \
|
||||
--tokens $repo/data/lang_bpe_500/tokens.txt \
|
||||
$repo/test_wavs/1089-134686-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0002.wav
|
||||
done
|
||||
|
||||
echo "GITHUB_EVENT_NAME: ${GITHUB_EVENT_NAME}"
|
||||
echo "GITHUB_EVENT_LABEL_NAME: ${GITHUB_EVENT_LABEL_NAME}"
|
||||
if [[ x"${GITHUB_EVENT_NAME}" == x"schedule" || x"${GITHUB_EVENT_LABEL_NAME}" == x"run-decode" ]]; then
|
||||
mkdir -p zipformer/exp
|
||||
ln -s $PWD/$repo/exp/pretrained.pt zipformer/exp/epoch-999.pt
|
||||
ln -s $PWD/$repo/data/lang_bpe_500 data/
|
||||
|
||||
ls -lh data
|
||||
ls -lh zipformer/exp
|
||||
|
||||
log "Decoding test-clean and test-other"
|
||||
|
||||
# use a small value for decoding with CPU
|
||||
max_duration=100
|
||||
|
||||
for method in greedy_search fast_beam_search modified_beam_search; do
|
||||
log "Simulated streaming decoding with $method"
|
||||
|
||||
./zipformer/decode.py \
|
||||
--causal 1 \
|
||||
--chunk-size 16 \
|
||||
--left-context-frames 128 \
|
||||
--decoding-method $method \
|
||||
--epoch 999 \
|
||||
--avg 1 \
|
||||
--use-averaged-model 0 \
|
||||
--max-duration $max_duration \
|
||||
--exp-dir zipformer/exp
|
||||
done
|
||||
|
||||
for method in greedy_search fast_beam_search modified_beam_search; do
|
||||
log "Chunk-wise streaming decoding with $method"
|
||||
|
||||
./zipformer/streaming_decode.py \
|
||||
--causal 1 \
|
||||
--chunk-size 16 \
|
||||
--left-context-frames 128 \
|
||||
--decoding-method $method \
|
||||
--epoch 999 \
|
||||
--avg 1 \
|
||||
--use-averaged-model 0 \
|
||||
--max-duration $max_duration \
|
||||
--exp-dir zipformer/exp
|
||||
done
|
||||
|
||||
rm zipformer/exp/*.pt
|
||||
fi
|
@ -1,77 +0,0 @@
|
||||
#!/usr/bin/env bash
|
||||
|
||||
set -e
|
||||
|
||||
log() {
|
||||
# This function is from espnet
|
||||
local fname=${BASH_SOURCE[1]##*/}
|
||||
echo -e "$(date '+%Y-%m-%d %H:%M:%S') (${fname}:${BASH_LINENO[0]}:${FUNCNAME[1]}) $*"
|
||||
}
|
||||
|
||||
cd egs/librispeech/ASR
|
||||
|
||||
repo_url=https://huggingface.co/csukuangfj/icefall-asr-librispeech-transducer-stateless2-torchaudio-2022-04-19
|
||||
|
||||
log "Downloading pre-trained model from $repo_url"
|
||||
git lfs install
|
||||
git clone $repo_url
|
||||
repo=$(basename $repo_url)
|
||||
|
||||
log "Display test files"
|
||||
tree $repo/
|
||||
ls -lh $repo/test_wavs/*.wav
|
||||
|
||||
for sym in 1 2 3; do
|
||||
log "Greedy search with --max-sym-per-frame $sym"
|
||||
|
||||
./transducer_stateless2/pretrained.py \
|
||||
--method greedy_search \
|
||||
--max-sym-per-frame $sym \
|
||||
--checkpoint $repo/exp/pretrained.pt \
|
||||
--tokens $repo/data/lang_bpe_500/tokens.txt \
|
||||
$repo/test_wavs/1089-134686-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0002.wav
|
||||
done
|
||||
|
||||
for method in fast_beam_search modified_beam_search beam_search; do
|
||||
log "$method"
|
||||
|
||||
./transducer_stateless2/pretrained.py \
|
||||
--method $method \
|
||||
--beam-size 4 \
|
||||
--checkpoint $repo/exp/pretrained.pt \
|
||||
--tokens $repo/data/lang_bpe_500/tokens.txt \
|
||||
$repo/test_wavs/1089-134686-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0002.wav
|
||||
done
|
||||
|
||||
echo "GITHUB_EVENT_NAME: ${GITHUB_EVENT_NAME}"
|
||||
echo "GITHUB_EVENT_LABEL_NAME: ${GITHUB_EVENT_LABEL_NAME}"
|
||||
if [[ x"${GITHUB_EVENT_NAME}" == x"schedule" || x"${GITHUB_EVENT_LABEL_NAME}" == x"run-decode" ]]; then
|
||||
mkdir -p transducer_stateless2/exp
|
||||
ln -s $PWD/$repo/exp/pretrained.pt transducer_stateless2/exp/epoch-999.pt
|
||||
ln -s $PWD/$repo/data/lang_bpe_500 data/
|
||||
|
||||
ls -lh data
|
||||
ls -lh transducer_stateless2/exp
|
||||
|
||||
log "Decoding test-clean and test-other"
|
||||
|
||||
# use a small value for decoding with CPU
|
||||
max_duration=100
|
||||
|
||||
for method in greedy_search fast_beam_search modified_beam_search; do
|
||||
log "Decoding with $method"
|
||||
|
||||
./transducer_stateless2/decode.py \
|
||||
--decoding-method $method \
|
||||
--epoch 999 \
|
||||
--avg 1 \
|
||||
--max-duration $max_duration \
|
||||
--exp-dir transducer_stateless2/exp
|
||||
done
|
||||
|
||||
rm transducer_stateless2/exp/*.pt
|
||||
fi
|
@ -1,94 +0,0 @@
|
||||
#!/usr/bin/env bash
|
||||
|
||||
set -e
|
||||
|
||||
log() {
|
||||
# This function is from espnet
|
||||
local fname=${BASH_SOURCE[1]##*/}
|
||||
echo -e "$(date '+%Y-%m-%d %H:%M:%S') (${fname}:${BASH_LINENO[0]}:${FUNCNAME[1]}) $*"
|
||||
}
|
||||
|
||||
cd egs/librispeech/ASR
|
||||
|
||||
repo_url=https://huggingface.co/Zengwei/icefall-asr-librispeech-zipformer-2023-05-15
|
||||
|
||||
log "Downloading pre-trained model from $repo_url"
|
||||
git lfs install
|
||||
GIT_LFS_SKIP_SMUDGE=1 git clone $repo_url
|
||||
repo=$(basename $repo_url)
|
||||
|
||||
log "Display test files"
|
||||
tree $repo/
|
||||
ls -lh $repo/test_wavs/*.wav
|
||||
|
||||
pushd $repo/exp
|
||||
git lfs pull --include "data/lang_bpe_500/bpe.model"
|
||||
git lfs pull --include "data/lang_bpe_500/tokens.txt"
|
||||
git lfs pull --include "exp/jit_script.pt"
|
||||
git lfs pull --include "exp/pretrained.pt"
|
||||
ln -s pretrained.pt epoch-99.pt
|
||||
ls -lh *.pt
|
||||
popd
|
||||
|
||||
log "Export to torchscript model"
|
||||
./zipformer/export.py \
|
||||
--exp-dir $repo/exp \
|
||||
--use-averaged-model false \
|
||||
--tokens $repo/data/lang_bpe_500/tokens.txt \
|
||||
--epoch 99 \
|
||||
--avg 1 \
|
||||
--jit 1
|
||||
|
||||
ls -lh $repo/exp/*.pt
|
||||
|
||||
log "Decode with models exported by torch.jit.script()"
|
||||
|
||||
./zipformer/jit_pretrained.py \
|
||||
--tokens $repo/data/lang_bpe_500/tokens.txt \
|
||||
--nn-model-filename $repo/exp/jit_script.pt \
|
||||
$repo/test_wavs/1089-134686-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0002.wav
|
||||
|
||||
for method in greedy_search modified_beam_search fast_beam_search; do
|
||||
log "$method"
|
||||
|
||||
./zipformer/pretrained.py \
|
||||
--method $method \
|
||||
--beam-size 4 \
|
||||
--checkpoint $repo/exp/pretrained.pt \
|
||||
--tokens $repo/data/lang_bpe_500/tokens.txt \
|
||||
$repo/test_wavs/1089-134686-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0002.wav
|
||||
done
|
||||
|
||||
echo "GITHUB_EVENT_NAME: ${GITHUB_EVENT_NAME}"
|
||||
echo "GITHUB_EVENT_LABEL_NAME: ${GITHUB_EVENT_LABEL_NAME}"
|
||||
if [[ x"${GITHUB_EVENT_NAME}" == x"schedule" || x"${GITHUB_EVENT_LABEL_NAME}" == x"run-decode" ]]; then
|
||||
mkdir -p zipformer/exp
|
||||
ln -s $PWD/$repo/exp/pretrained.pt zipformer/exp/epoch-999.pt
|
||||
ln -s $PWD/$repo/data/lang_bpe_500 data/
|
||||
|
||||
ls -lh data
|
||||
ls -lh zipformer/exp
|
||||
|
||||
log "Decoding test-clean and test-other"
|
||||
|
||||
# use a small value for decoding with CPU
|
||||
max_duration=100
|
||||
|
||||
for method in greedy_search fast_beam_search modified_beam_search; do
|
||||
log "Decoding with $method"
|
||||
|
||||
./zipformer/decode.py \
|
||||
--decoding-method $method \
|
||||
--epoch 999 \
|
||||
--avg 1 \
|
||||
--use-averaged-model 0 \
|
||||
--max-duration $max_duration \
|
||||
--exp-dir zipformer/exp
|
||||
done
|
||||
|
||||
rm zipformer/exp/*.pt
|
||||
fi
|
@ -1,117 +0,0 @@
|
||||
#!/usr/bin/env bash
|
||||
|
||||
set -e
|
||||
|
||||
log() {
|
||||
# This function is from espnet
|
||||
local fname=${BASH_SOURCE[1]##*/}
|
||||
echo -e "$(date '+%Y-%m-%d %H:%M:%S') (${fname}:${BASH_LINENO[0]}:${FUNCNAME[1]}) $*"
|
||||
}
|
||||
|
||||
cd egs/librispeech/ASR
|
||||
|
||||
repo_url=https://huggingface.co/Zengwei/icefall-asr-librispeech-zipformer-transducer-ctc-2023-06-13
|
||||
|
||||
log "Downloading pre-trained model from $repo_url"
|
||||
git lfs install
|
||||
GIT_LFS_SKIP_SMUDGE=1 git clone $repo_url
|
||||
repo=$(basename $repo_url)
|
||||
|
||||
log "Display test files"
|
||||
tree $repo/
|
||||
ls -lh $repo/test_wavs/*.wav
|
||||
|
||||
pushd $repo/exp
|
||||
git lfs pull --include "data/lang_bpe_500/bpe.model"
|
||||
git lfs pull --include "data/lang_bpe_500/tokens.txt"
|
||||
git lfs pull --include "data/lang_bpe_500/HLG.pt"
|
||||
git lfs pull --include "data/lang_bpe_500/L.pt"
|
||||
git lfs pull --include "data/lang_bpe_500/LG.pt"
|
||||
git lfs pull --include "data/lang_bpe_500/Linv.pt"
|
||||
git lfs pull --include "data/lm/G_4_gram.pt"
|
||||
git lfs pull --include "exp/jit_script.pt"
|
||||
git lfs pull --include "exp/pretrained.pt"
|
||||
ln -s pretrained.pt epoch-99.pt
|
||||
ls -lh *.pt
|
||||
popd
|
||||
|
||||
log "Export to torchscript model"
|
||||
./zipformer/export.py \
|
||||
--exp-dir $repo/exp \
|
||||
--use-transducer 1 \
|
||||
--use-ctc 1 \
|
||||
--use-averaged-model false \
|
||||
--tokens $repo/data/lang_bpe_500/tokens.txt \
|
||||
--epoch 99 \
|
||||
--avg 1 \
|
||||
--jit 1
|
||||
|
||||
ls -lh $repo/exp/*.pt
|
||||
|
||||
log "Decode with models exported by torch.jit.script()"
|
||||
|
||||
for method in ctc-decoding 1best; do
|
||||
./zipformer/jit_pretrained_ctc.py \
|
||||
--tokens $repo/data/lang_bpe_500/tokens.txt \
|
||||
--model-filename $repo/exp/jit_script.pt \
|
||||
--HLG $repo/data/lang_bpe_500/HLG.pt \
|
||||
--words-file $repo/data/lang_bpe_500/words.txt \
|
||||
--G $repo/data/lm/G_4_gram.pt \
|
||||
--method $method \
|
||||
--sample-rate 16000 \
|
||||
$repo/test_wavs/1089-134686-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0002.wav
|
||||
done
|
||||
|
||||
for method in ctc-decoding 1best; do
|
||||
log "$method"
|
||||
|
||||
./zipformer/pretrained_ctc.py \
|
||||
--use-transducer 1 \
|
||||
--use-ctc 1 \
|
||||
--method $method \
|
||||
--checkpoint $repo/exp/pretrained.pt \
|
||||
--tokens $repo/data/lang_bpe_500/tokens.txt \
|
||||
--HLG $repo/data/lang_bpe_500/HLG.pt \
|
||||
--G $repo/data/lm/G_4_gram.pt \
|
||||
--words-file $repo/data/lang_bpe_500/words.txt \
|
||||
--sample-rate 16000 \
|
||||
$repo/test_wavs/1089-134686-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0002.wav
|
||||
done
|
||||
|
||||
echo "GITHUB_EVENT_NAME: ${GITHUB_EVENT_NAME}"
|
||||
echo "GITHUB_EVENT_LABEL_NAME: ${GITHUB_EVENT_LABEL_NAME}"
|
||||
if [[ x"${GITHUB_EVENT_NAME}" == x"schedule" || x"${GITHUB_EVENT_LABEL_NAME}" == x"run-decode" ]]; then
|
||||
mkdir -p zipformer/exp
|
||||
ln -s $PWD/$repo/exp/pretrained.pt zipformer/exp/epoch-999.pt
|
||||
ln -s $PWD/$repo/data/lang_bpe_500 data/
|
||||
|
||||
ls -lh data
|
||||
ls -lh zipformer/exp
|
||||
|
||||
log "Decoding test-clean and test-other"
|
||||
|
||||
# use a small value for decoding with CPU
|
||||
max_duration=100
|
||||
|
||||
for method in ctc-decoding 1best; do
|
||||
log "Decoding with $method"
|
||||
|
||||
./zipformer/ctc_decode.py \
|
||||
--use-transducer 1 \
|
||||
--use-ctc 1 \
|
||||
--decoding-method $method \
|
||||
--nbest-scale 1.0 \
|
||||
--hlg-scale 0.6 \
|
||||
--epoch 999 \
|
||||
--avg 1 \
|
||||
--use-averaged-model 0 \
|
||||
--max-duration $max_duration \
|
||||
--exp-dir zipformer/exp
|
||||
done
|
||||
|
||||
rm zipformer/exp/*.pt
|
||||
fi
|
@ -1,102 +0,0 @@
|
||||
#!/usr/bin/env bash
|
||||
|
||||
set -e
|
||||
|
||||
log() {
|
||||
# This function is from espnet
|
||||
local fname=${BASH_SOURCE[1]##*/}
|
||||
echo -e "$(date '+%Y-%m-%d %H:%M:%S') (${fname}:${BASH_LINENO[0]}:${FUNCNAME[1]}) $*"
|
||||
}
|
||||
|
||||
cd egs/librispeech/ASR
|
||||
|
||||
repo_url=https://huggingface.co/Zengwei/icefall-asr-librispeech-zipformer-mmi-2022-12-08
|
||||
|
||||
log "Downloading pre-trained model from $repo_url"
|
||||
GIT_LFS_SKIP_SMUDGE=1 git clone $repo_url
|
||||
repo=$(basename $repo_url)
|
||||
|
||||
log "Display test files"
|
||||
tree $repo/
|
||||
ls -lh $repo/test_wavs/*.wav
|
||||
|
||||
pushd $repo/exp
|
||||
git lfs pull --include "data/lang_bpe_500/3gram.pt"
|
||||
git lfs pull --include "data/lang_bpe_500/4gram.pt"
|
||||
git lfs pull --include "data/lang_bpe_500/L.pt"
|
||||
git lfs pull --include "data/lang_bpe_500/LG.pt"
|
||||
git lfs pull --include "data/lang_bpe_500/Linv.pt"
|
||||
git lfs pull --include "data/lang_bpe_500/bpe.model"
|
||||
git lfs pull --include "exp/cpu_jit.pt"
|
||||
git lfs pull --include "exp/pretrained.pt"
|
||||
ln -s pretrained.pt epoch-99.pt
|
||||
ls -lh *.pt
|
||||
popd
|
||||
|
||||
log "Export to torchscript model"
|
||||
./zipformer_mmi/export.py \
|
||||
--exp-dir $repo/exp \
|
||||
--use-averaged-model false \
|
||||
--tokens $repo/data/lang_bpe_500/tokens.txt \
|
||||
--epoch 99 \
|
||||
--avg 1 \
|
||||
--jit 1
|
||||
|
||||
ls -lh $repo/exp/*.pt
|
||||
|
||||
log "Decode with models exported by torch.jit.script()"
|
||||
|
||||
./zipformer_mmi/jit_pretrained.py \
|
||||
--bpe-model $repo/data/lang_bpe_500/bpe.model \
|
||||
--nn-model-filename $repo/exp/cpu_jit.pt \
|
||||
--lang-dir $repo/data/lang_bpe_500 \
|
||||
$repo/test_wavs/1089-134686-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0002.wav
|
||||
|
||||
for method in 1best nbest nbest-rescoring-LG nbest-rescoring-3-gram nbest-rescoring-4-gram; do
|
||||
log "$method"
|
||||
|
||||
./zipformer_mmi/pretrained.py \
|
||||
--method $method \
|
||||
--checkpoint $repo/exp/pretrained.pt \
|
||||
--lang-dir $repo/data/lang_bpe_500 \
|
||||
--tokens $repo/data/lang_bpe_500/tokens.txt \
|
||||
$repo/test_wavs/1089-134686-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0002.wav
|
||||
done
|
||||
|
||||
|
||||
echo "GITHUB_EVENT_NAME: ${GITHUB_EVENT_NAME}"
|
||||
echo "GITHUB_EVENT_LABEL_NAME: ${GITHUB_EVENT_LABEL_NAME}"
|
||||
if [[ x"${GITHUB_EVENT_NAME}" == x"schedule" || x"${GITHUB_EVENT_LABEL_NAME}" == x"run-decode" ]]; then
|
||||
mkdir -p zipformer_mmi/exp
|
||||
ln -s $PWD/$repo/exp/pretrained.pt zipformer_mmi/exp/epoch-999.pt
|
||||
ln -s $PWD/$repo/data/lang_bpe_500 data/
|
||||
|
||||
ls -lh data
|
||||
ls -lh zipformer_mmi/exp
|
||||
|
||||
log "Decoding test-clean and test-other"
|
||||
|
||||
# use a small value for decoding with CPU
|
||||
max_duration=100
|
||||
|
||||
for method in 1best nbest nbest-rescoring-LG nbest-rescoring-3-gram nbest-rescoring-4-gram; do
|
||||
log "Decoding with $method"
|
||||
|
||||
./zipformer_mmi/decode.py \
|
||||
--decoding-method $method \
|
||||
--epoch 999 \
|
||||
--avg 1 \
|
||||
--use-averaged-model 0 \
|
||||
--nbest-scale 1.2 \
|
||||
--hp-scale 1.0 \
|
||||
--max-duration $max_duration \
|
||||
--lang-dir $repo/data/lang_bpe_500 \
|
||||
--exp-dir zipformer_mmi/exp
|
||||
done
|
||||
|
||||
rm zipformer_mmi/exp/*.pt
|
||||
fi
|
240
.github/scripts/run-pre-trained-ctc.sh
vendored
240
.github/scripts/run-pre-trained-ctc.sh
vendored
@ -1,240 +0,0 @@
|
||||
#!/usr/bin/env bash
|
||||
|
||||
set -e
|
||||
|
||||
log() {
|
||||
# This function is from espnet
|
||||
local fname=${BASH_SOURCE[1]##*/}
|
||||
echo -e "$(date '+%Y-%m-%d %H:%M:%S') (${fname}:${BASH_LINENO[0]}:${FUNCNAME[1]}) $*"
|
||||
}
|
||||
|
||||
pushd egs/librispeech/ASR
|
||||
|
||||
repo_url=https://huggingface.co/csukuangfj/sherpa-onnx-zipformer-ctc-en-2023-10-02
|
||||
log "Downloading pre-trained model from $repo_url"
|
||||
git lfs install
|
||||
git clone $repo_url
|
||||
repo=$(basename $repo_url)
|
||||
|
||||
log "Display test files"
|
||||
tree $repo/
|
||||
ls -lh $repo/test_wavs/*.wav
|
||||
|
||||
log "CTC greedy search"
|
||||
|
||||
./zipformer/onnx_pretrained_ctc.py \
|
||||
--nn-model $repo/model.onnx \
|
||||
--tokens $repo/tokens.txt \
|
||||
$repo/test_wavs/0.wav \
|
||||
$repo/test_wavs/1.wav \
|
||||
$repo/test_wavs/2.wav
|
||||
|
||||
log "CTC H decoding"
|
||||
|
||||
./zipformer/onnx_pretrained_ctc_H.py \
|
||||
--nn-model $repo/model.onnx \
|
||||
--tokens $repo/tokens.txt \
|
||||
--H $repo/H.fst \
|
||||
$repo/test_wavs/0.wav \
|
||||
$repo/test_wavs/1.wav \
|
||||
$repo/test_wavs/2.wav
|
||||
|
||||
log "CTC HL decoding"
|
||||
|
||||
./zipformer/onnx_pretrained_ctc_HL.py \
|
||||
--nn-model $repo/model.onnx \
|
||||
--words $repo/words.txt \
|
||||
--HL $repo/HL.fst \
|
||||
$repo/test_wavs/0.wav \
|
||||
$repo/test_wavs/1.wav \
|
||||
$repo/test_wavs/2.wav
|
||||
|
||||
log "CTC HLG decoding"
|
||||
|
||||
./zipformer/onnx_pretrained_ctc_HLG.py \
|
||||
--nn-model $repo/model.onnx \
|
||||
--words $repo/words.txt \
|
||||
--HLG $repo/HLG.fst \
|
||||
$repo/test_wavs/0.wav \
|
||||
$repo/test_wavs/1.wav \
|
||||
$repo/test_wavs/2.wav
|
||||
|
||||
rm -rf $repo
|
||||
|
||||
repo_url=https://huggingface.co/csukuangfj/icefall-asr-librispeech-conformer-ctc-jit-bpe-500-2021-11-09
|
||||
log "Downloading pre-trained model from $repo_url"
|
||||
GIT_LFS_SKIP_SMUDGE=1 git clone $repo_url
|
||||
repo=$(basename $repo_url)
|
||||
pushd $repo
|
||||
|
||||
git lfs pull --include "exp/pretrained.pt"
|
||||
git lfs pull --include "data/lang_bpe_500/HLG.pt"
|
||||
git lfs pull --include "data/lang_bpe_500/L.pt"
|
||||
git lfs pull --include "data/lang_bpe_500/L_disambig.pt"
|
||||
git lfs pull --include "data/lang_bpe_500/Linv.pt"
|
||||
git lfs pull --include "data/lang_bpe_500/bpe.model"
|
||||
git lfs pull --include "data/lang_bpe_500/lexicon.txt"
|
||||
git lfs pull --include "data/lang_bpe_500/lexicon_disambig.txt"
|
||||
git lfs pull --include "data/lang_bpe_500/tokens.txt"
|
||||
git lfs pull --include "data/lang_bpe_500/words.txt"
|
||||
git lfs pull --include "data/lm/G_3_gram.fst.txt"
|
||||
|
||||
popd
|
||||
|
||||
log "Display test files"
|
||||
tree $repo/
|
||||
ls -lh $repo/test_wavs/*.wav
|
||||
|
||||
log "CTC decoding"
|
||||
|
||||
./conformer_ctc/pretrained.py \
|
||||
--method ctc-decoding \
|
||||
--num-classes 500 \
|
||||
--checkpoint $repo/exp/pretrained.pt \
|
||||
--tokens $repo/data/lang_bpe_500/tokens.txt \
|
||||
$repo/test_wavs/1089-134686-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0002.wav
|
||||
|
||||
log "HLG decoding"
|
||||
|
||||
./conformer_ctc/pretrained.py \
|
||||
--method 1best \
|
||||
--num-classes 500 \
|
||||
--checkpoint $repo/exp/pretrained.pt \
|
||||
--tokens $repo/data/lang_bpe_500/tokens.txt \
|
||||
--words-file $repo/data/lang_bpe_500/words.txt \
|
||||
--HLG $repo/data/lang_bpe_500/HLG.pt \
|
||||
$repo/test_wavs/1089-134686-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0002.wav
|
||||
|
||||
log "CTC decoding on CPU with kaldi decoders using OpenFst"
|
||||
|
||||
log "Exporting model with torchscript"
|
||||
|
||||
pushd $repo/exp
|
||||
ln -s pretrained.pt epoch-99.pt
|
||||
popd
|
||||
|
||||
./conformer_ctc/export.py \
|
||||
--epoch 99 \
|
||||
--avg 1 \
|
||||
--exp-dir $repo/exp \
|
||||
--tokens $repo/data/lang_bpe_500/tokens.txt \
|
||||
--jit 1
|
||||
|
||||
ls -lh $repo/exp
|
||||
|
||||
|
||||
log "Generating H.fst, HL.fst"
|
||||
|
||||
./local/prepare_lang_fst.py --lang-dir $repo/data/lang_bpe_500 --ngram-G $repo/data/lm/G_3_gram.fst.txt
|
||||
|
||||
ls -lh $repo/data/lang_bpe_500
|
||||
|
||||
log "Decoding with H on CPU with OpenFst"
|
||||
|
||||
./conformer_ctc/jit_pretrained_decode_with_H.py \
|
||||
--nn-model $repo/exp/cpu_jit.pt \
|
||||
--H $repo/data/lang_bpe_500/H.fst \
|
||||
--tokens $repo/data/lang_bpe_500/tokens.txt \
|
||||
$repo/test_wavs/1089-134686-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0002.wav
|
||||
|
||||
log "Decoding with HL on CPU with OpenFst"
|
||||
|
||||
./conformer_ctc/jit_pretrained_decode_with_HL.py \
|
||||
--nn-model $repo/exp/cpu_jit.pt \
|
||||
--HL $repo/data/lang_bpe_500/HL.fst \
|
||||
--words $repo/data/lang_bpe_500/words.txt \
|
||||
$repo/test_wavs/1089-134686-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0002.wav
|
||||
|
||||
log "Decoding with HLG on CPU with OpenFst"
|
||||
|
||||
./conformer_ctc/jit_pretrained_decode_with_HLG.py \
|
||||
--nn-model $repo/exp/cpu_jit.pt \
|
||||
--HLG $repo/data/lang_bpe_500/HLG.fst \
|
||||
--words $repo/data/lang_bpe_500/words.txt \
|
||||
$repo/test_wavs/1089-134686-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0002.wav
|
||||
|
||||
rm -rf $repo
|
||||
|
||||
popd
|
||||
|
||||
log "Test aishell"
|
||||
|
||||
pushd egs/aishell/ASR
|
||||
|
||||
repo_url=https://huggingface.co/csukuangfj/icefall_asr_aishell_conformer_ctc
|
||||
log "Downloading pre-trained model from $repo_url"
|
||||
GIT_LFS_SKIP_SMUDGE=1 git clone $repo_url
|
||||
repo=$(basename $repo_url)
|
||||
pushd $repo
|
||||
|
||||
git lfs pull --include "exp/pretrained.pt"
|
||||
git lfs pull --include "data/lang_char/H.fst"
|
||||
git lfs pull --include "data/lang_char/HL.fst"
|
||||
git lfs pull --include "data/lang_char/HLG.fst"
|
||||
|
||||
popd
|
||||
|
||||
log "Display test files"
|
||||
tree $repo/
|
||||
ls -lh $repo/test_wavs/*.wav
|
||||
|
||||
log "CTC decoding"
|
||||
|
||||
log "Exporting model with torchscript"
|
||||
|
||||
pushd $repo/exp
|
||||
ln -s pretrained.pt epoch-99.pt
|
||||
popd
|
||||
|
||||
./conformer_ctc/export.py \
|
||||
--epoch 99 \
|
||||
--avg 1 \
|
||||
--exp-dir $repo/exp \
|
||||
--tokens $repo/data/lang_char/tokens.txt \
|
||||
--jit 1
|
||||
|
||||
ls -lh $repo/exp
|
||||
|
||||
ls -lh $repo/data/lang_char
|
||||
|
||||
log "Decoding with H on CPU with OpenFst"
|
||||
|
||||
./conformer_ctc/jit_pretrained_decode_with_H.py \
|
||||
--nn-model $repo/exp/cpu_jit.pt \
|
||||
--H $repo/data/lang_char/H.fst \
|
||||
--tokens $repo/data/lang_char/tokens.txt \
|
||||
$repo/test_wavs/0.wav \
|
||||
$repo/test_wavs/1.wav \
|
||||
$repo/test_wavs/2.wav
|
||||
|
||||
log "Decoding with HL on CPU with OpenFst"
|
||||
|
||||
./conformer_ctc/jit_pretrained_decode_with_HL.py \
|
||||
--nn-model $repo/exp/cpu_jit.pt \
|
||||
--HL $repo/data/lang_char/HL.fst \
|
||||
--words $repo/data/lang_char/words.txt \
|
||||
$repo/test_wavs/0.wav \
|
||||
$repo/test_wavs/1.wav \
|
||||
$repo/test_wavs/2.wav
|
||||
|
||||
log "Decoding with HLG on CPU with OpenFst"
|
||||
|
||||
./conformer_ctc/jit_pretrained_decode_with_HLG.py \
|
||||
--nn-model $repo/exp/cpu_jit.pt \
|
||||
--HLG $repo/data/lang_char/HLG.fst \
|
||||
--words $repo/data/lang_char/words.txt \
|
||||
$repo/test_wavs/0.wav \
|
||||
$repo/test_wavs/1.wav \
|
||||
$repo/test_wavs/2.wav
|
||||
|
||||
rm -rf $repo
|
@ -1,77 +0,0 @@
|
||||
#!/usr/bin/env bash
|
||||
|
||||
set -e
|
||||
|
||||
log() {
|
||||
# This function is from espnet
|
||||
local fname=${BASH_SOURCE[1]##*/}
|
||||
echo -e "$(date '+%Y-%m-%d %H:%M:%S') (${fname}:${BASH_LINENO[0]}:${FUNCNAME[1]}) $*"
|
||||
}
|
||||
|
||||
cd egs/librispeech/ASR
|
||||
|
||||
repo_url=https://huggingface.co/csukuangfj/icefall-asr-librispeech-100h-transducer-stateless-multi-datasets-bpe-500-2022-02-21
|
||||
|
||||
log "Downloading pre-trained model from $repo_url"
|
||||
git lfs install
|
||||
git clone $repo_url
|
||||
repo=$(basename $repo_url)
|
||||
|
||||
log "Display test files"
|
||||
tree $repo/
|
||||
ls -lh $repo/test_wavs/*.wav
|
||||
|
||||
for sym in 1 2 3; do
|
||||
log "Greedy search with --max-sym-per-frame $sym"
|
||||
|
||||
./transducer_stateless_multi_datasets/pretrained.py \
|
||||
--method greedy_search \
|
||||
--max-sym-per-frame $sym \
|
||||
--checkpoint $repo/exp/pretrained.pt \
|
||||
--tokens $repo/data/lang_bpe_500/tokens.txt \
|
||||
$repo/test_wavs/1089-134686-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0002.wav
|
||||
done
|
||||
|
||||
for method in modified_beam_search beam_search fast_beam_search; do
|
||||
log "$method"
|
||||
|
||||
./transducer_stateless_multi_datasets/pretrained.py \
|
||||
--method $method \
|
||||
--beam-size 4 \
|
||||
--checkpoint $repo/exp/pretrained.pt \
|
||||
--tokens $repo/data/lang_bpe_500/tokens.txt \
|
||||
$repo/test_wavs/1089-134686-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0002.wav
|
||||
done
|
||||
|
||||
echo "GITHUB_EVENT_NAME: ${GITHUB_EVENT_NAME}"
|
||||
echo "GITHUB_EVENT_LABEL_NAME: ${GITHUB_EVENT_LABEL_NAME}"
|
||||
if [[ x"${GITHUB_EVENT_NAME}" == x"schedule" || x"${GITHUB_EVENT_LABEL_NAME}" == x"run-decode" ]]; then
|
||||
mkdir -p transducer_stateless_multi_datasets/exp
|
||||
ln -s $PWD/$repo/exp/pretrained.pt transducer_stateless_multi_datasets/exp/epoch-999.pt
|
||||
ln -s $PWD/$repo/data/lang_bpe_500 data/
|
||||
|
||||
ls -lh data
|
||||
ls -lh transducer_stateless_multi_datasets/exp
|
||||
|
||||
log "Decoding test-clean and test-other"
|
||||
|
||||
# use a small value for decoding with CPU
|
||||
max_duration=100
|
||||
|
||||
for method in greedy_search fast_beam_search modified_beam_search; do
|
||||
log "Decoding with $method"
|
||||
|
||||
./transducer_stateless_multi_datasets/decode.py \
|
||||
--decoding-method $method \
|
||||
--epoch 999 \
|
||||
--avg 1 \
|
||||
--max-duration $max_duration \
|
||||
--exp-dir transducer_stateless_multi_datasets/exp
|
||||
done
|
||||
|
||||
rm transducer_stateless_multi_datasets/exp/*.pt
|
||||
fi
|
@ -1,77 +0,0 @@
|
||||
#!/usr/bin/env bash
|
||||
|
||||
set -e
|
||||
|
||||
log() {
|
||||
# This function is from espnet
|
||||
local fname=${BASH_SOURCE[1]##*/}
|
||||
echo -e "$(date '+%Y-%m-%d %H:%M:%S') (${fname}:${BASH_LINENO[0]}:${FUNCNAME[1]}) $*"
|
||||
}
|
||||
|
||||
cd egs/librispeech/ASR
|
||||
|
||||
repo_url=https://huggingface.co/csukuangfj/icefall-asr-librispeech-transducer-stateless-multi-datasets-bpe-500-2022-03-01
|
||||
|
||||
log "Downloading pre-trained model from $repo_url"
|
||||
git lfs install
|
||||
git clone $repo_url
|
||||
repo=$(basename $repo_url)
|
||||
|
||||
log "Display test files"
|
||||
tree $repo/
|
||||
ls -lh $repo/test_wavs/*.wav
|
||||
|
||||
for sym in 1 2 3; do
|
||||
log "Greedy search with --max-sym-per-frame $sym"
|
||||
|
||||
./transducer_stateless_multi_datasets/pretrained.py \
|
||||
--method greedy_search \
|
||||
--max-sym-per-frame $sym \
|
||||
--checkpoint $repo/exp/pretrained.pt \
|
||||
--tokens $repo/data/lang_bpe_500/tokens.txt \
|
||||
$repo/test_wavs/1089-134686-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0002.wav
|
||||
done
|
||||
|
||||
for method in modified_beam_search beam_search fast_beam_search; do
|
||||
log "$method"
|
||||
|
||||
./transducer_stateless_multi_datasets/pretrained.py \
|
||||
--method $method \
|
||||
--beam-size 4 \
|
||||
--checkpoint $repo/exp/pretrained.pt \
|
||||
--tokens $repo/data/lang_bpe_500/tokens.txt \
|
||||
$repo/test_wavs/1089-134686-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0002.wav
|
||||
done
|
||||
|
||||
echo "GITHUB_EVENT_NAME: ${GITHUB_EVENT_NAME}"
|
||||
echo "GITHUB_EVENT_LABEL_NAME: ${GITHUB_EVENT_LABEL_NAME}"
|
||||
if [[ x"${GITHUB_EVENT_NAME}" == x"schedule" || x"${GITHUB_EVENT_LABEL_NAME}" == x"run-decode" ]]; then
|
||||
mkdir -p transducer_stateless_multi_datasets/exp
|
||||
ln -s $PWD/$repo/exp/pretrained.pt transducer_stateless_multi_datasets/exp/epoch-999.pt
|
||||
ln -s $PWD/$repo/data/lang_bpe_500 data/
|
||||
|
||||
ls -lh data
|
||||
ls -lh transducer_stateless_multi_datasets/exp
|
||||
|
||||
log "Decoding test-clean and test-other"
|
||||
|
||||
# use a small value for decoding with CPU
|
||||
max_duration=100
|
||||
|
||||
for method in greedy_search fast_beam_search modified_beam_search; do
|
||||
log "Decoding with $method"
|
||||
|
||||
./transducer_stateless_multi_datasets/decode.py \
|
||||
--decoding-method $method \
|
||||
--epoch 999 \
|
||||
--avg 1 \
|
||||
--max-duration $max_duration \
|
||||
--exp-dir transducer_stateless_multi_datasets/exp
|
||||
done
|
||||
|
||||
rm transducer_stateless_multi_datasets/exp/*.pt
|
||||
fi
|
@ -1,48 +0,0 @@
|
||||
#!/usr/bin/env bash
|
||||
|
||||
set -e
|
||||
|
||||
log() {
|
||||
# This function is from espnet
|
||||
local fname=${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 \
|
||||
$repo/test_wavs/BAC009S0764W0123.wav
|
||||
done
|
@ -1,48 +0,0 @@
|
||||
#!/usr/bin/env bash
|
||||
|
||||
set -e
|
||||
|
||||
log() {
|
||||
# This function is from espnet
|
||||
local fname=${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 \
|
||||
$repo/test_wavs/BAC009S0764W0123.wav
|
||||
done
|
@ -1,77 +0,0 @@
|
||||
#!/usr/bin/env bash
|
||||
|
||||
set -e
|
||||
|
||||
log() {
|
||||
# This function is from espnet
|
||||
local fname=${BASH_SOURCE[1]##*/}
|
||||
echo -e "$(date '+%Y-%m-%d %H:%M:%S') (${fname}:${BASH_LINENO[0]}:${FUNCNAME[1]}) $*"
|
||||
}
|
||||
|
||||
cd egs/librispeech/ASR
|
||||
|
||||
repo_url=https://huggingface.co/csukuangfj/icefall-asr-librispeech-transducer-stateless-bpe-500-2022-02-07
|
||||
|
||||
log "Downloading pre-trained model from $repo_url"
|
||||
git lfs install
|
||||
git clone $repo_url
|
||||
repo=$(basename $repo_url)
|
||||
|
||||
log "Display test files"
|
||||
tree $repo/
|
||||
ls -lh $repo/test_wavs/*.wav
|
||||
|
||||
for sym in 1 2 3; do
|
||||
log "Greedy search with --max-sym-per-frame $sym"
|
||||
|
||||
./transducer_stateless/pretrained.py \
|
||||
--method greedy_search \
|
||||
--max-sym-per-frame $sym \
|
||||
--checkpoint $repo/exp/pretrained.pt \
|
||||
--tokens $repo/data/lang_bpe_500/tokens.txt \
|
||||
$repo/test_wavs/1089-134686-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0002.wav
|
||||
done
|
||||
|
||||
for method in fast_beam_search modified_beam_search beam_search; do
|
||||
log "$method"
|
||||
|
||||
./transducer_stateless/pretrained.py \
|
||||
--method $method \
|
||||
--beam-size 4 \
|
||||
--checkpoint $repo/exp/pretrained.pt \
|
||||
--tokens $repo/data/lang_bpe_500/tokens.txt \
|
||||
$repo/test_wavs/1089-134686-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0002.wav
|
||||
done
|
||||
|
||||
echo "GITHUB_EVENT_NAME: ${GITHUB_EVENT_NAME}"
|
||||
echo "GITHUB_EVENT_LABEL_NAME: ${GITHUB_EVENT_LABEL_NAME}"
|
||||
if [[ x"${GITHUB_EVENT_NAME}" == x"schedule" || x"${GITHUB_EVENT_LABEL_NAME}" == x"run-decode" ]]; then
|
||||
mkdir -p transducer_stateless/exp
|
||||
ln -s $PWD/$repo/exp/pretrained.pt transducer_stateless/exp/epoch-999.pt
|
||||
ln -s $PWD/$repo/data/lang_bpe_500 data/
|
||||
|
||||
ls -lh data
|
||||
ls -lh transducer_stateless/exp
|
||||
|
||||
log "Decoding test-clean and test-other"
|
||||
|
||||
# use a small value for decoding with CPU
|
||||
max_duration=100
|
||||
|
||||
for method in greedy_search fast_beam_search modified_beam_search; do
|
||||
log "Decoding with $method"
|
||||
|
||||
./transducer_stateless/decode.py \
|
||||
--decoding-method $method \
|
||||
--epoch 999 \
|
||||
--avg 1 \
|
||||
--max-duration $max_duration \
|
||||
--exp-dir transducer_stateless/exp
|
||||
done
|
||||
|
||||
rm transducer_stateless/exp/*.pt
|
||||
fi
|
33
.github/scripts/run-pre-trained-transducer.sh
vendored
33
.github/scripts/run-pre-trained-transducer.sh
vendored
@ -1,33 +0,0 @@
|
||||
#!/usr/bin/env bash
|
||||
|
||||
set -e
|
||||
|
||||
log() {
|
||||
# This function is from espnet
|
||||
local fname=${BASH_SOURCE[1]##*/}
|
||||
echo -e "$(date '+%Y-%m-%d %H:%M:%S') (${fname}:${BASH_LINENO[0]}:${FUNCNAME[1]}) $*"
|
||||
}
|
||||
|
||||
cd egs/librispeech/ASR
|
||||
|
||||
repo_url=https://huggingface.co/csukuangfj/icefall-asr-librispeech-transducer-bpe-500-2021-12-23
|
||||
|
||||
log "Downloading pre-trained model from $repo_url"
|
||||
git lfs install
|
||||
git clone $repo_url
|
||||
repo=$(basename $repo_url)
|
||||
|
||||
log "Display test files"
|
||||
tree $repo/
|
||||
ls -lh $repo/test_wavs/*.wav
|
||||
|
||||
log "Beam search decoding"
|
||||
|
||||
./transducer/pretrained.py \
|
||||
--method beam_search \
|
||||
--beam-size 4 \
|
||||
--checkpoint $repo/exp/pretrained.pt \
|
||||
--tokens $repo/data/lang_bpe_500/tokens.txt \
|
||||
$repo/test_wavs/1089-134686-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0001.wav \
|
||||
$repo/test_wavs/1221-135766-0002.wav
|
86
.github/scripts/yesno/ASR/run.sh
vendored
Executable file
86
.github/scripts/yesno/ASR/run.sh
vendored
Executable file
@ -0,0 +1,86 @@
|
||||
#!/usr/bin/env bash
|
||||
|
||||
set -ex
|
||||
|
||||
log() {
|
||||
# This function is from espnet
|
||||
local fname=${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
72
.github/workflows/aishell.yml
vendored
Normal 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
|
81
.github/workflows/build-cpu-docker.yml
vendored
Normal file
81
.github/workflows/build-cpu-docker.yml
vendored
Normal 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 $tag
|
71
.github/workflows/librispeech.yml
vendored
Normal file
71
.github/workflows/librispeech.yml
vendored
Normal file
@ -0,0 +1,71 @@
|
||||
name: librispeech
|
||||
on:
|
||||
push:
|
||||
branches:
|
||||
- master
|
||||
|
||||
pull_request:
|
||||
branches:
|
||||
- master
|
||||
|
||||
workflow_dispatch:
|
||||
|
||||
concurrency:
|
||||
group: librispeech-${{ 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
|
123
.github/workflows/run-aishell-2022-06-20.yml
vendored
123
.github/workflows/run-aishell-2022-06-20.yml
vendored
@ -1,123 +0,0 @@
|
||||
# Copyright 2022 Fangjun Kuang (csukuangfj@gmail.com)
|
||||
|
||||
# See ../../LICENSE for clarification regarding multiple authors
|
||||
#
|
||||
# Licensed under the Apache License, Version 2.0 (the "License");
|
||||
# you may not use this file except in compliance with the License.
|
||||
# You may obtain a copy of the License at
|
||||
#
|
||||
# http://www.apache.org/licenses/LICENSE-2.0
|
||||
#
|
||||
# Unless required by applicable law or agreed to in writing, software
|
||||
# distributed under the License is distributed on an "AS IS" BASIS,
|
||||
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
||||
# See the License for the specific language governing permissions and
|
||||
# limitations under the License.
|
||||
|
||||
name: run-aishell-2022-06-20
|
||||
# pruned RNN-T + reworked model with random combiner
|
||||
# https://huggingface.co/csukuangfj/icefall-aishell-pruned-transducer-stateless3-2022-06-20
|
||||
|
||||
on:
|
||||
push:
|
||||
branches:
|
||||
- master
|
||||
pull_request:
|
||||
types: [labeled]
|
||||
|
||||
schedule:
|
||||
# minute (0-59)
|
||||
# hour (0-23)
|
||||
# day of the month (1-31)
|
||||
# month (1-12)
|
||||
# day of the week (0-6)
|
||||
# nightly build at 15:50 UTC time every day
|
||||
- cron: "50 15 * * *"
|
||||
|
||||
concurrency:
|
||||
group: run_aishell_2022_06_20-${{ github.ref }}
|
||||
cancel-in-progress: true
|
||||
|
||||
jobs:
|
||||
run_aishell_2022_06_20:
|
||||
if: github.event.label.name == 'ready' || github.event.label.name == 'run-decode' || github.event_name == 'push' || github.event_name == 'schedule'
|
||||
runs-on: ${{ matrix.os }}
|
||||
strategy:
|
||||
matrix:
|
||||
os: [ubuntu-latest]
|
||||
python-version: [3.8]
|
||||
|
||||
fail-fast: false
|
||||
|
||||
steps:
|
||||
- uses: actions/checkout@v2
|
||||
with:
|
||||
fetch-depth: 0
|
||||
|
||||
- name: Setup Python ${{ matrix.python-version }}
|
||||
uses: actions/setup-python@v2
|
||||
with:
|
||||
python-version: ${{ matrix.python-version }}
|
||||
cache: 'pip'
|
||||
cache-dependency-path: '**/requirements-ci.txt'
|
||||
|
||||
- name: Install Python dependencies
|
||||
run: |
|
||||
grep -v '^#' ./requirements-ci.txt | xargs -n 1 -L 1 pip install
|
||||
pip uninstall -y protobuf
|
||||
pip install --no-binary protobuf protobuf==3.20.*
|
||||
|
||||
- name: Cache kaldifeat
|
||||
id: my-cache
|
||||
uses: actions/cache@v2
|
||||
with:
|
||||
path: |
|
||||
~/tmp/kaldifeat
|
||||
key: cache-tmp-${{ matrix.python-version }}-2023-05-22
|
||||
|
||||
- name: Install kaldifeat
|
||||
if: steps.my-cache.outputs.cache-hit != 'true'
|
||||
shell: bash
|
||||
run: |
|
||||
.github/scripts/install-kaldifeat.sh
|
||||
|
||||
- name: Inference with pre-trained model
|
||||
shell: bash
|
||||
env:
|
||||
GITHUB_EVENT_NAME: ${{ github.event_name }}
|
||||
GITHUB_EVENT_LABEL_NAME: ${{ github.event.label.name }}
|
||||
run: |
|
||||
sudo apt-get -qq install git-lfs tree
|
||||
export PYTHONPATH=$PWD:$PYTHONPATH
|
||||
export PYTHONPATH=~/tmp/kaldifeat/kaldifeat/python:$PYTHONPATH
|
||||
export PYTHONPATH=~/tmp/kaldifeat/build/lib:$PYTHONPATH
|
||||
|
||||
.github/scripts/run-aishell-pruned-transducer-stateless3-2022-06-20.sh
|
||||
|
||||
- name: Display decoding results for aishell pruned_transducer_stateless3
|
||||
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
|
||||
shell: bash
|
||||
run: |
|
||||
cd egs/aishell/ASR/
|
||||
tree ./pruned_transducer_stateless3/exp
|
||||
|
||||
cd pruned_transducer_stateless3
|
||||
echo "results for pruned_transducer_stateless3"
|
||||
echo "===greedy search==="
|
||||
find exp/greedy_search -name "log-*" -exec grep -n --color "best for test" {} + | sort -n -k2
|
||||
find exp/greedy_search -name "log-*" -exec grep -n --color "best for dev" {} + | sort -n -k2
|
||||
|
||||
echo "===fast_beam_search==="
|
||||
find exp/fast_beam_search -name "log-*" -exec grep -n --color "best for test" {} + | sort -n -k2
|
||||
find exp/fast_beam_search -name "log-*" -exec grep -n --color "best for dev" {} + | sort -n -k2
|
||||
|
||||
echo "===modified beam search==="
|
||||
find exp/modified_beam_search -name "log-*" -exec grep -n --color "best for test" {} + | sort -n -k2
|
||||
find exp/modified_beam_search -name "log-*" -exec grep -n --color "best for dev" {} + | sort -n -k2
|
||||
|
||||
- name: Upload decoding results for aishell pruned_transducer_stateless3
|
||||
uses: actions/upload-artifact@v2
|
||||
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
|
||||
with:
|
||||
name: aishell-torch-${{ matrix.torch }}-python-${{ matrix.python-version }}-ubuntu-latest-cpu-pruned_transducer_stateless3-2022-06-20
|
||||
path: egs/aishell/ASR/pruned_transducer_stateless3/exp/
|
@ -1,95 +0,0 @@
|
||||
# Copyright 2023 Zengrui Jin (Xiaomi Corp.)
|
||||
|
||||
# See ../../LICENSE for clarification regarding multiple authors
|
||||
#
|
||||
# Licensed under the Apache License, Version 2.0 (the "License");
|
||||
# you may not use this file except in compliance with the License.
|
||||
# You may obtain a copy of the License at
|
||||
#
|
||||
# http://www.apache.org/licenses/LICENSE-2.0
|
||||
#
|
||||
# Unless required by applicable law or agreed to in writing, software
|
||||
# distributed under the License is distributed on an "AS IS" BASIS,
|
||||
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
||||
# See the License for the specific language governing permissions and
|
||||
# limitations under the License.
|
||||
|
||||
name: run-aishell-zipformer-2023-10-24
|
||||
|
||||
on:
|
||||
push:
|
||||
branches:
|
||||
- master
|
||||
pull_request:
|
||||
types: [labeled]
|
||||
|
||||
schedule:
|
||||
# minute (0-59)
|
||||
# hour (0-23)
|
||||
# day of the month (1-31)
|
||||
# month (1-12)
|
||||
# day of the week (0-6)
|
||||
# nightly build at 15:50 UTC time every day
|
||||
- cron: "50 15 * * *"
|
||||
|
||||
concurrency:
|
||||
group: run_aishell_zipformer_2023_10_24-${{ github.ref }}
|
||||
cancel-in-progress: true
|
||||
|
||||
jobs:
|
||||
run_aishell_zipformer_2023_10_24:
|
||||
if: github.event.label.name == 'ready' || github.event.label.name == 'zipformer' || github.event.label.name == 'run-decode' || github.event_name == 'push' || github.event_name == 'schedule'
|
||||
runs-on: ${{ matrix.os }}
|
||||
strategy:
|
||||
matrix:
|
||||
os: [ubuntu-latest]
|
||||
python-version: [3.8]
|
||||
|
||||
fail-fast: false
|
||||
|
||||
steps:
|
||||
- uses: actions/checkout@v2
|
||||
with:
|
||||
fetch-depth: 0
|
||||
|
||||
- name: Setup Python ${{ matrix.python-version }}
|
||||
uses: actions/setup-python@v2
|
||||
with:
|
||||
python-version: ${{ matrix.python-version }}
|
||||
cache: 'pip'
|
||||
cache-dependency-path: '**/requirements-ci.txt'
|
||||
|
||||
- name: Install Python dependencies
|
||||
run: |
|
||||
grep -v '^#' ./requirements-ci.txt | xargs -n 1 -L 1 pip install
|
||||
pip uninstall -y protobuf
|
||||
pip install --no-binary protobuf protobuf==3.20.*
|
||||
|
||||
- name: Cache kaldifeat
|
||||
id: my-cache
|
||||
uses: actions/cache@v2
|
||||
with:
|
||||
path: |
|
||||
~/tmp/kaldifeat
|
||||
key: cache-tmp-${{ matrix.python-version }}-2023-05-22
|
||||
|
||||
- name: Install kaldifeat
|
||||
if: steps.my-cache.outputs.cache-hit != 'true'
|
||||
shell: bash
|
||||
run: |
|
||||
.github/scripts/install-kaldifeat.sh
|
||||
|
||||
- name: Inference with pre-trained model
|
||||
shell: bash
|
||||
env:
|
||||
GITHUB_EVENT_NAME: ${{ github.event_name }}
|
||||
GITHUB_EVENT_LABEL_NAME: ${{ github.event.label.name }}
|
||||
run: |
|
||||
sudo apt-get -qq install git-lfs tree
|
||||
export PYTHONPATH=$PWD:$PYTHONPATH
|
||||
export PYTHONPATH=~/tmp/kaldifeat/kaldifeat/python:$PYTHONPATH
|
||||
export PYTHONPATH=~/tmp/kaldifeat/build/lib:$PYTHONPATH
|
||||
|
||||
.github/scripts/run-aishell-zipformer-2023-10-24.sh
|
||||
|
||||
|
159
.github/workflows/run-librispeech-2022-03-12.yml
vendored
159
.github/workflows/run-librispeech-2022-03-12.yml
vendored
@ -1,159 +0,0 @@
|
||||
# Copyright 2021 Fangjun Kuang (csukuangfj@gmail.com)
|
||||
|
||||
# See ../../LICENSE for clarification regarding multiple authors
|
||||
#
|
||||
# Licensed under the Apache License, Version 2.0 (the "License");
|
||||
# you may not use this file except in compliance with the License.
|
||||
# You may obtain a copy of the License at
|
||||
#
|
||||
# http://www.apache.org/licenses/LICENSE-2.0
|
||||
#
|
||||
# Unless required by applicable law or agreed to in writing, software
|
||||
# distributed under the License is distributed on an "AS IS" BASIS,
|
||||
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
||||
# See the License for the specific language governing permissions and
|
||||
# limitations under the License.
|
||||
|
||||
name: run-librispeech-2022-03-12
|
||||
# stateless transducer + k2 pruned rnnt-loss
|
||||
|
||||
on:
|
||||
push:
|
||||
branches:
|
||||
- master
|
||||
pull_request:
|
||||
types: [labeled]
|
||||
|
||||
schedule:
|
||||
# minute (0-59)
|
||||
# hour (0-23)
|
||||
# day of the month (1-31)
|
||||
# month (1-12)
|
||||
# day of the week (0-6)
|
||||
# nightly build at 15:50 UTC time every day
|
||||
- cron: "50 15 * * *"
|
||||
|
||||
concurrency:
|
||||
group: run_librispeech_2022_03_12-${{ github.ref }}
|
||||
cancel-in-progress: true
|
||||
|
||||
jobs:
|
||||
run_librispeech_2022_03_12:
|
||||
if: github.event.label.name == 'ready' || github.event.label.name == 'run-decode' || github.event_name == 'push' || github.event_name == 'schedule'
|
||||
runs-on: ${{ matrix.os }}
|
||||
strategy:
|
||||
matrix:
|
||||
os: [ubuntu-latest]
|
||||
python-version: [3.8]
|
||||
|
||||
fail-fast: false
|
||||
|
||||
steps:
|
||||
- uses: actions/checkout@v2
|
||||
with:
|
||||
fetch-depth: 0
|
||||
|
||||
- name: Setup Python ${{ matrix.python-version }}
|
||||
uses: actions/setup-python@v2
|
||||
with:
|
||||
python-version: ${{ matrix.python-version }}
|
||||
cache: 'pip'
|
||||
cache-dependency-path: '**/requirements-ci.txt'
|
||||
|
||||
- name: Install Python dependencies
|
||||
run: |
|
||||
grep -v '^#' ./requirements-ci.txt | xargs -n 1 -L 1 pip install
|
||||
pip uninstall -y protobuf
|
||||
pip install --no-binary protobuf protobuf==3.20.*
|
||||
|
||||
- name: Cache kaldifeat
|
||||
id: my-cache
|
||||
uses: actions/cache@v2
|
||||
with:
|
||||
path: |
|
||||
~/tmp/kaldifeat
|
||||
key: cache-tmp-${{ matrix.python-version }}-2023-05-22
|
||||
|
||||
- name: Install kaldifeat
|
||||
if: steps.my-cache.outputs.cache-hit != 'true'
|
||||
shell: bash
|
||||
run: |
|
||||
.github/scripts/install-kaldifeat.sh
|
||||
|
||||
- name: Cache LibriSpeech test-clean and test-other datasets
|
||||
id: libri-test-clean-and-test-other-data
|
||||
uses: actions/cache@v2
|
||||
with:
|
||||
path: |
|
||||
~/tmp/download
|
||||
key: cache-libri-test-clean-and-test-other
|
||||
|
||||
- name: Download LibriSpeech test-clean and test-other
|
||||
if: steps.libri-test-clean-and-test-other-data.outputs.cache-hit != 'true'
|
||||
shell: bash
|
||||
run: |
|
||||
.github/scripts/download-librispeech-test-clean-and-test-other-dataset.sh
|
||||
|
||||
- name: Prepare manifests for LibriSpeech test-clean and test-other
|
||||
shell: bash
|
||||
run: |
|
||||
.github/scripts/prepare-librispeech-test-clean-and-test-other-manifests.sh
|
||||
|
||||
- name: Cache LibriSpeech test-clean and test-other fbank features
|
||||
id: libri-test-clean-and-test-other-fbank
|
||||
uses: actions/cache@v2
|
||||
with:
|
||||
path: |
|
||||
~/tmp/fbank-libri
|
||||
key: cache-libri-fbank-test-clean-and-test-other-v2
|
||||
|
||||
- name: Compute fbank for LibriSpeech test-clean and test-other
|
||||
if: steps.libri-test-clean-and-test-other-fbank.outputs.cache-hit != 'true'
|
||||
shell: bash
|
||||
run: |
|
||||
.github/scripts/compute-fbank-librispeech-test-clean-and-test-other.sh
|
||||
|
||||
- name: Inference with pre-trained model
|
||||
shell: bash
|
||||
env:
|
||||
GITHUB_EVENT_NAME: ${{ github.event_name }}
|
||||
GITHUB_EVENT_LABEL_NAME: ${{ github.event.label.name }}
|
||||
run: |
|
||||
mkdir -p egs/librispeech/ASR/data
|
||||
ln -sfv ~/tmp/fbank-libri egs/librispeech/ASR/data/fbank
|
||||
ls -lh egs/librispeech/ASR/data/*
|
||||
|
||||
sudo apt-get -qq install git-lfs tree
|
||||
export PYTHONPATH=$PWD:$PYTHONPATH
|
||||
export PYTHONPATH=~/tmp/kaldifeat/kaldifeat/python:$PYTHONPATH
|
||||
export PYTHONPATH=~/tmp/kaldifeat/build/lib:$PYTHONPATH
|
||||
|
||||
.github/scripts/run-librispeech-pruned-transducer-stateless-2022-03-12.sh
|
||||
|
||||
- name: Display decoding results for pruned_transducer_stateless
|
||||
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
|
||||
shell: bash
|
||||
run: |
|
||||
cd egs/librispeech/ASR/
|
||||
tree ./pruned_transducer_stateless/exp
|
||||
|
||||
cd pruned_transducer_stateless
|
||||
echo "results for pruned_transducer_stateless"
|
||||
echo "===greedy search==="
|
||||
find exp/greedy_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
|
||||
find exp/greedy_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
|
||||
|
||||
echo "===fast_beam_search==="
|
||||
find exp/fast_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
|
||||
find exp/fast_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
|
||||
|
||||
echo "===modified beam search==="
|
||||
find exp/modified_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
|
||||
find exp/modified_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
|
||||
|
||||
- name: Upload decoding results for pruned_transducer_stateless
|
||||
uses: actions/upload-artifact@v2
|
||||
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
|
||||
with:
|
||||
name: torch-${{ matrix.torch }}-python-${{ matrix.python-version }}-ubuntu-latest-cpu-pruned_transducer_stateless-2022-03-12
|
||||
path: egs/librispeech/ASR/pruned_transducer_stateless/exp/
|
185
.github/workflows/run-librispeech-2022-04-29.yml
vendored
185
.github/workflows/run-librispeech-2022-04-29.yml
vendored
@ -1,185 +0,0 @@
|
||||
# Copyright 2021 Fangjun Kuang (csukuangfj@gmail.com)
|
||||
|
||||
# See ../../LICENSE for clarification regarding multiple authors
|
||||
#
|
||||
# Licensed under the Apache License, Version 2.0 (the "License");
|
||||
# you may not use this file except in compliance with the License.
|
||||
# You may obtain a copy of the License at
|
||||
#
|
||||
# http://www.apache.org/licenses/LICENSE-2.0
|
||||
#
|
||||
# Unless required by applicable law or agreed to in writing, software
|
||||
# distributed under the License is distributed on an "AS IS" BASIS,
|
||||
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
||||
# See the License for the specific language governing permissions and
|
||||
# limitations under the License.
|
||||
|
||||
name: run-librispeech-2022-04-29
|
||||
# stateless pruned transducer (reworked model) + giga speech
|
||||
|
||||
on:
|
||||
push:
|
||||
branches:
|
||||
- master
|
||||
pull_request:
|
||||
types: [labeled]
|
||||
|
||||
schedule:
|
||||
# minute (0-59)
|
||||
# hour (0-23)
|
||||
# day of the month (1-31)
|
||||
# month (1-12)
|
||||
# day of the week (0-6)
|
||||
# nightly build at 15:50 UTC time every day
|
||||
- cron: "50 15 * * *"
|
||||
|
||||
concurrency:
|
||||
group: run_librispeech_2022_04_29-${{ github.ref }}
|
||||
cancel-in-progress: true
|
||||
|
||||
jobs:
|
||||
run_librispeech_2022_04_29:
|
||||
if: github.event.label.name == 'ready' || github.event.label.name == 'run-decode' || github.event_name == 'push' || github.event_name == 'schedule'
|
||||
runs-on: ${{ matrix.os }}
|
||||
strategy:
|
||||
matrix:
|
||||
os: [ubuntu-latest]
|
||||
python-version: [3.8]
|
||||
|
||||
fail-fast: false
|
||||
|
||||
steps:
|
||||
- uses: actions/checkout@v2
|
||||
with:
|
||||
fetch-depth: 0
|
||||
|
||||
- name: Setup Python ${{ matrix.python-version }}
|
||||
uses: actions/setup-python@v2
|
||||
with:
|
||||
python-version: ${{ matrix.python-version }}
|
||||
cache: 'pip'
|
||||
cache-dependency-path: '**/requirements-ci.txt'
|
||||
|
||||
- name: Install Python dependencies
|
||||
run: |
|
||||
grep -v '^#' ./requirements-ci.txt | xargs -n 1 -L 1 pip install
|
||||
pip uninstall -y protobuf
|
||||
pip install --no-binary protobuf protobuf==3.20.*
|
||||
|
||||
- name: Cache kaldifeat
|
||||
id: my-cache
|
||||
uses: actions/cache@v2
|
||||
with:
|
||||
path: |
|
||||
~/tmp/kaldifeat
|
||||
key: cache-tmp-${{ matrix.python-version }}-2023-05-22
|
||||
|
||||
- name: Install kaldifeat
|
||||
if: steps.my-cache.outputs.cache-hit != 'true'
|
||||
shell: bash
|
||||
run: |
|
||||
.github/scripts/install-kaldifeat.sh
|
||||
|
||||
- name: Cache LibriSpeech test-clean and test-other datasets
|
||||
id: libri-test-clean-and-test-other-data
|
||||
uses: actions/cache@v2
|
||||
with:
|
||||
path: |
|
||||
~/tmp/download
|
||||
key: cache-libri-test-clean-and-test-other
|
||||
|
||||
- name: Download LibriSpeech test-clean and test-other
|
||||
if: steps.libri-test-clean-and-test-other-data.outputs.cache-hit != 'true'
|
||||
shell: bash
|
||||
run: |
|
||||
.github/scripts/download-librispeech-test-clean-and-test-other-dataset.sh
|
||||
|
||||
- name: Prepare manifests for LibriSpeech test-clean and test-other
|
||||
shell: bash
|
||||
run: |
|
||||
.github/scripts/prepare-librispeech-test-clean-and-test-other-manifests.sh
|
||||
|
||||
- name: Cache LibriSpeech test-clean and test-other fbank features
|
||||
id: libri-test-clean-and-test-other-fbank
|
||||
uses: actions/cache@v2
|
||||
with:
|
||||
path: |
|
||||
~/tmp/fbank-libri
|
||||
key: cache-libri-fbank-test-clean-and-test-other-v2
|
||||
|
||||
- name: Compute fbank for LibriSpeech test-clean and test-other
|
||||
if: steps.libri-test-clean-and-test-other-fbank.outputs.cache-hit != 'true'
|
||||
shell: bash
|
||||
run: |
|
||||
.github/scripts/compute-fbank-librispeech-test-clean-and-test-other.sh
|
||||
|
||||
- name: Inference with pre-trained model
|
||||
shell: bash
|
||||
env:
|
||||
GITHUB_EVENT_NAME: ${{ github.event_name }}
|
||||
GITHUB_EVENT_LABEL_NAME: ${{ github.event.label.name }}
|
||||
run: |
|
||||
mkdir -p egs/librispeech/ASR/data
|
||||
ln -sfv ~/tmp/fbank-libri egs/librispeech/ASR/data/fbank
|
||||
ls -lh egs/librispeech/ASR/data/*
|
||||
|
||||
sudo apt-get -qq install git-lfs tree
|
||||
export PYTHONPATH=$PWD:$PYTHONPATH
|
||||
export PYTHONPATH=~/tmp/kaldifeat/kaldifeat/python:$PYTHONPATH
|
||||
export PYTHONPATH=~/tmp/kaldifeat/build/lib:$PYTHONPATH
|
||||
|
||||
.github/scripts/run-librispeech-pruned-transducer-stateless2-2022-04-29.sh
|
||||
|
||||
.github/scripts/run-librispeech-pruned-transducer-stateless3-2022-04-29.sh
|
||||
|
||||
- name: Display decoding results for pruned_transducer_stateless2
|
||||
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
|
||||
shell: bash
|
||||
run: |
|
||||
cd egs/librispeech/ASR
|
||||
tree pruned_transducer_stateless2/exp
|
||||
cd pruned_transducer_stateless2/exp
|
||||
echo "===greedy search==="
|
||||
find greedy_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
|
||||
find greedy_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
|
||||
|
||||
echo "===fast_beam_search==="
|
||||
find fast_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
|
||||
find fast_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
|
||||
|
||||
echo "===modified beam search==="
|
||||
find modified_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
|
||||
find modified_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
|
||||
|
||||
- name: Display decoding results for pruned_transducer_stateless3
|
||||
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
|
||||
shell: bash
|
||||
run: |
|
||||
cd egs/librispeech/ASR
|
||||
tree pruned_transducer_stateless3/exp
|
||||
cd pruned_transducer_stateless3/exp
|
||||
echo "===greedy search==="
|
||||
find greedy_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
|
||||
find greedy_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
|
||||
|
||||
echo "===fast_beam_search==="
|
||||
find fast_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
|
||||
find fast_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
|
||||
|
||||
echo "===modified beam search==="
|
||||
find modified_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
|
||||
find modified_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
|
||||
|
||||
- name: Upload decoding results for pruned_transducer_stateless2
|
||||
uses: actions/upload-artifact@v2
|
||||
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
|
||||
with:
|
||||
name: torch-${{ matrix.torch }}-python-${{ matrix.python-version }}-ubuntu-latest-cpu-pruned_transducer_stateless2-2022-04-29
|
||||
path: egs/librispeech/ASR/pruned_transducer_stateless2/exp/
|
||||
|
||||
- name: Upload decoding results for pruned_transducer_stateless3
|
||||
uses: actions/upload-artifact@v2
|
||||
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
|
||||
with:
|
||||
name: torch-${{ matrix.torch }}-python-${{ matrix.python-version }}-ubuntu-latest-cpu-pruned_transducer_stateless3-2022-04-29
|
||||
path: egs/librispeech/ASR/pruned_transducer_stateless3/exp/
|
159
.github/workflows/run-librispeech-2022-05-13.yml
vendored
159
.github/workflows/run-librispeech-2022-05-13.yml
vendored
@ -1,159 +0,0 @@
|
||||
# Copyright 2022 Fangjun Kuang (csukuangfj@gmail.com)
|
||||
|
||||
# See ../../LICENSE for clarification regarding multiple authors
|
||||
#
|
||||
# Licensed under the Apache License, Version 2.0 (the "License");
|
||||
# you may not use this file except in compliance with the License.
|
||||
# You may obtain a copy of the License at
|
||||
#
|
||||
# http://www.apache.org/licenses/LICENSE-2.0
|
||||
#
|
||||
# Unless required by applicable law or agreed to in writing, software
|
||||
# distributed under the License is distributed on an "AS IS" BASIS,
|
||||
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
||||
# See the License for the specific language governing permissions and
|
||||
# limitations under the License.
|
||||
|
||||
name: run-librispeech-2022-05-13
|
||||
# stateless transducer + k2 pruned rnnt-loss + deeper model
|
||||
|
||||
on:
|
||||
push:
|
||||
branches:
|
||||
- master
|
||||
pull_request:
|
||||
types: [labeled]
|
||||
|
||||
schedule:
|
||||
# minute (0-59)
|
||||
# hour (0-23)
|
||||
# day of the month (1-31)
|
||||
# month (1-12)
|
||||
# day of the week (0-6)
|
||||
# nightly build at 15:50 UTC time every day
|
||||
- cron: "50 15 * * *"
|
||||
|
||||
concurrency:
|
||||
group: run_librispeech_2022_05_13-${{ github.ref }}
|
||||
cancel-in-progress: true
|
||||
|
||||
jobs:
|
||||
run_librispeech_2022_05_13:
|
||||
if: github.event.label.name == 'ready' || github.event.label.name == 'run-decode' || github.event_name == 'push' || github.event_name == 'schedule'
|
||||
runs-on: ${{ matrix.os }}
|
||||
strategy:
|
||||
matrix:
|
||||
os: [ubuntu-latest]
|
||||
python-version: [3.8]
|
||||
|
||||
fail-fast: false
|
||||
|
||||
steps:
|
||||
- uses: actions/checkout@v2
|
||||
with:
|
||||
fetch-depth: 0
|
||||
|
||||
- name: Setup Python ${{ matrix.python-version }}
|
||||
uses: actions/setup-python@v2
|
||||
with:
|
||||
python-version: ${{ matrix.python-version }}
|
||||
cache: 'pip'
|
||||
cache-dependency-path: '**/requirements-ci.txt'
|
||||
|
||||
- name: Install Python dependencies
|
||||
run: |
|
||||
grep -v '^#' ./requirements-ci.txt | xargs -n 1 -L 1 pip install
|
||||
pip uninstall -y protobuf
|
||||
pip install --no-binary protobuf protobuf==3.20.*
|
||||
|
||||
- name: Cache kaldifeat
|
||||
id: my-cache
|
||||
uses: actions/cache@v2
|
||||
with:
|
||||
path: |
|
||||
~/tmp/kaldifeat
|
||||
key: cache-tmp-${{ matrix.python-version }}-2023-05-22
|
||||
|
||||
- name: Install kaldifeat
|
||||
if: steps.my-cache.outputs.cache-hit != 'true'
|
||||
shell: bash
|
||||
run: |
|
||||
.github/scripts/install-kaldifeat.sh
|
||||
|
||||
- name: Cache LibriSpeech test-clean and test-other datasets
|
||||
id: libri-test-clean-and-test-other-data
|
||||
uses: actions/cache@v2
|
||||
with:
|
||||
path: |
|
||||
~/tmp/download
|
||||
key: cache-libri-test-clean-and-test-other
|
||||
|
||||
- name: Download LibriSpeech test-clean and test-other
|
||||
if: steps.libri-test-clean-and-test-other-data.outputs.cache-hit != 'true'
|
||||
shell: bash
|
||||
run: |
|
||||
.github/scripts/download-librispeech-test-clean-and-test-other-dataset.sh
|
||||
|
||||
- name: Prepare manifests for LibriSpeech test-clean and test-other
|
||||
shell: bash
|
||||
run: |
|
||||
.github/scripts/prepare-librispeech-test-clean-and-test-other-manifests.sh
|
||||
|
||||
- name: Cache LibriSpeech test-clean and test-other fbank features
|
||||
id: libri-test-clean-and-test-other-fbank
|
||||
uses: actions/cache@v2
|
||||
with:
|
||||
path: |
|
||||
~/tmp/fbank-libri
|
||||
key: cache-libri-fbank-test-clean-and-test-other-v2
|
||||
|
||||
- name: Compute fbank for LibriSpeech test-clean and test-other
|
||||
if: steps.libri-test-clean-and-test-other-fbank.outputs.cache-hit != 'true'
|
||||
shell: bash
|
||||
run: |
|
||||
.github/scripts/compute-fbank-librispeech-test-clean-and-test-other.sh
|
||||
|
||||
- name: Inference with pre-trained model
|
||||
shell: bash
|
||||
env:
|
||||
GITHUB_EVENT_NAME: ${{ github.event_name }}
|
||||
GITHUB_EVENT_LABEL_NAME: ${{ github.event.label.name }}
|
||||
run: |
|
||||
mkdir -p egs/librispeech/ASR/data
|
||||
ln -sfv ~/tmp/fbank-libri egs/librispeech/ASR/data/fbank
|
||||
ls -lh egs/librispeech/ASR/data/*
|
||||
|
||||
sudo apt-get -qq install git-lfs tree
|
||||
export PYTHONPATH=$PWD:$PYTHONPATH
|
||||
export PYTHONPATH=~/tmp/kaldifeat/kaldifeat/python:$PYTHONPATH
|
||||
export PYTHONPATH=~/tmp/kaldifeat/build/lib:$PYTHONPATH
|
||||
|
||||
.github/scripts/run-librispeech-pruned-transducer-stateless5-2022-05-13.sh
|
||||
|
||||
- name: Display decoding results for librispeech pruned_transducer_stateless5
|
||||
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
|
||||
shell: bash
|
||||
run: |
|
||||
cd egs/librispeech/ASR/
|
||||
tree ./pruned_transducer_stateless5/exp
|
||||
|
||||
cd pruned_transducer_stateless5
|
||||
echo "results for pruned_transducer_stateless5"
|
||||
echo "===greedy search==="
|
||||
find exp/greedy_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
|
||||
find exp/greedy_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
|
||||
|
||||
echo "===fast_beam_search==="
|
||||
find exp/fast_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
|
||||
find exp/fast_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
|
||||
|
||||
echo "===modified beam search==="
|
||||
find exp/modified_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
|
||||
find exp/modified_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
|
||||
|
||||
- name: Upload decoding results for librispeech pruned_transducer_stateless5
|
||||
uses: actions/upload-artifact@v2
|
||||
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
|
||||
with:
|
||||
name: torch-${{ matrix.torch }}-python-${{ matrix.python-version }}-ubuntu-latest-cpu-pruned_transducer_stateless5-2022-05-13
|
||||
path: egs/librispeech/ASR/pruned_transducer_stateless5/exp/
|
@ -1,159 +0,0 @@
|
||||
# Copyright 2022 Fangjun Kuang (csukuangfj@gmail.com)
|
||||
|
||||
# See ../../LICENSE for clarification regarding multiple authors
|
||||
#
|
||||
# Licensed under the Apache License, Version 2.0 (the "License");
|
||||
# you may not use this file except in compliance with the License.
|
||||
# You may obtain a copy of the License at
|
||||
#
|
||||
# http://www.apache.org/licenses/LICENSE-2.0
|
||||
#
|
||||
# Unless required by applicable law or agreed to in writing, software
|
||||
# distributed under the License is distributed on an "AS IS" BASIS,
|
||||
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
||||
# See the License for the specific language governing permissions and
|
||||
# limitations under the License.
|
||||
|
||||
name: run-librispeech-2022-11-11-stateless7
|
||||
# zipformer
|
||||
|
||||
on:
|
||||
push:
|
||||
branches:
|
||||
- master
|
||||
pull_request:
|
||||
types: [labeled]
|
||||
|
||||
schedule:
|
||||
# minute (0-59)
|
||||
# hour (0-23)
|
||||
# day of the month (1-31)
|
||||
# month (1-12)
|
||||
# day of the week (0-6)
|
||||
# nightly build at 15:50 UTC time every day
|
||||
- cron: "50 15 * * *"
|
||||
|
||||
concurrency:
|
||||
group: run_librispeech_2022_11_11_zipformer-${{ github.ref }}
|
||||
cancel-in-progress: true
|
||||
|
||||
jobs:
|
||||
run_librispeech_2022_11_11_zipformer:
|
||||
if: github.event.label.name == 'ready' || github.event.label.name == 'run-decode' || github.event_name == 'push' || github.event_name == 'schedule'
|
||||
runs-on: ${{ matrix.os }}
|
||||
strategy:
|
||||
matrix:
|
||||
os: [ubuntu-latest]
|
||||
python-version: [3.8]
|
||||
|
||||
fail-fast: false
|
||||
|
||||
steps:
|
||||
- uses: actions/checkout@v2
|
||||
with:
|
||||
fetch-depth: 0
|
||||
|
||||
- name: Setup Python ${{ matrix.python-version }}
|
||||
uses: actions/setup-python@v2
|
||||
with:
|
||||
python-version: ${{ matrix.python-version }}
|
||||
cache: 'pip'
|
||||
cache-dependency-path: '**/requirements-ci.txt'
|
||||
|
||||
- name: Install Python dependencies
|
||||
run: |
|
||||
grep -v '^#' ./requirements-ci.txt | xargs -n 1 -L 1 pip install
|
||||
pip uninstall -y protobuf
|
||||
pip install --no-binary protobuf protobuf==3.20.*
|
||||
|
||||
- name: Cache kaldifeat
|
||||
id: my-cache
|
||||
uses: actions/cache@v2
|
||||
with:
|
||||
path: |
|
||||
~/tmp/kaldifeat
|
||||
key: cache-tmp-${{ matrix.python-version }}-2023-05-22
|
||||
|
||||
- name: Install kaldifeat
|
||||
if: steps.my-cache.outputs.cache-hit != 'true'
|
||||
shell: bash
|
||||
run: |
|
||||
.github/scripts/install-kaldifeat.sh
|
||||
|
||||
- name: Cache LibriSpeech test-clean and test-other datasets
|
||||
id: libri-test-clean-and-test-other-data
|
||||
uses: actions/cache@v2
|
||||
with:
|
||||
path: |
|
||||
~/tmp/download
|
||||
key: cache-libri-test-clean-and-test-other
|
||||
|
||||
- name: Download LibriSpeech test-clean and test-other
|
||||
if: steps.libri-test-clean-and-test-other-data.outputs.cache-hit != 'true'
|
||||
shell: bash
|
||||
run: |
|
||||
.github/scripts/download-librispeech-test-clean-and-test-other-dataset.sh
|
||||
|
||||
- name: Prepare manifests for LibriSpeech test-clean and test-other
|
||||
shell: bash
|
||||
run: |
|
||||
.github/scripts/prepare-librispeech-test-clean-and-test-other-manifests.sh
|
||||
|
||||
- name: Cache LibriSpeech test-clean and test-other fbank features
|
||||
id: libri-test-clean-and-test-other-fbank
|
||||
uses: actions/cache@v2
|
||||
with:
|
||||
path: |
|
||||
~/tmp/fbank-libri
|
||||
key: cache-libri-fbank-test-clean-and-test-other-v2
|
||||
|
||||
- name: Compute fbank for LibriSpeech test-clean and test-other
|
||||
if: steps.libri-test-clean-and-test-other-fbank.outputs.cache-hit != 'true'
|
||||
shell: bash
|
||||
run: |
|
||||
.github/scripts/compute-fbank-librispeech-test-clean-and-test-other.sh
|
||||
|
||||
- name: Inference with pre-trained model
|
||||
shell: bash
|
||||
env:
|
||||
GITHUB_EVENT_NAME: ${{ github.event_name }}
|
||||
GITHUB_EVENT_LABEL_NAME: ${{ github.event.label.name }}
|
||||
run: |
|
||||
mkdir -p egs/librispeech/ASR/data
|
||||
ln -sfv ~/tmp/fbank-libri egs/librispeech/ASR/data/fbank
|
||||
ls -lh egs/librispeech/ASR/data/*
|
||||
|
||||
sudo apt-get -qq install git-lfs tree
|
||||
export PYTHONPATH=$PWD:$PYTHONPATH
|
||||
export PYTHONPATH=~/tmp/kaldifeat/kaldifeat/python:$PYTHONPATH
|
||||
export PYTHONPATH=~/tmp/kaldifeat/build/lib:$PYTHONPATH
|
||||
|
||||
.github/scripts/run-librispeech-pruned-transducer-stateless7-2022-11-11.sh
|
||||
|
||||
- name: Display decoding results for librispeech pruned_transducer_stateless7
|
||||
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
|
||||
shell: bash
|
||||
run: |
|
||||
cd egs/librispeech/ASR/
|
||||
tree ./pruned_transducer_stateless7/exp
|
||||
|
||||
cd pruned_transducer_stateless7
|
||||
echo "results for pruned_transducer_stateless7"
|
||||
echo "===greedy search==="
|
||||
find exp/greedy_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
|
||||
find exp/greedy_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
|
||||
|
||||
echo "===fast_beam_search==="
|
||||
find exp/fast_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
|
||||
find exp/fast_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
|
||||
|
||||
echo "===modified beam search==="
|
||||
find exp/modified_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
|
||||
find exp/modified_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
|
||||
|
||||
- name: Upload decoding results for librispeech pruned_transducer_stateless7
|
||||
uses: actions/upload-artifact@v2
|
||||
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
|
||||
with:
|
||||
name: torch-${{ matrix.torch }}-python-${{ matrix.python-version }}-ubuntu-latest-cpu-pruned_transducer_stateless7-2022-11-11
|
||||
path: egs/librispeech/ASR/pruned_transducer_stateless7/exp/
|
@ -1,159 +0,0 @@
|
||||
# Copyright 2022 Fangjun Kuang (csukuangfj@gmail.com)
|
||||
|
||||
# See ../../LICENSE for clarification regarding multiple authors
|
||||
#
|
||||
# Licensed under the Apache License, Version 2.0 (the "License");
|
||||
# you may not use this file except in compliance with the License.
|
||||
# You may obtain a copy of the License at
|
||||
#
|
||||
# http://www.apache.org/licenses/LICENSE-2.0
|
||||
#
|
||||
# Unless required by applicable law or agreed to in writing, software
|
||||
# distributed under the License is distributed on an "AS IS" BASIS,
|
||||
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
||||
# See the License for the specific language governing permissions and
|
||||
# limitations under the License.
|
||||
|
||||
name: run-librispeech-2022-11-14-stateless8
|
||||
# zipformer
|
||||
|
||||
on:
|
||||
push:
|
||||
branches:
|
||||
- master
|
||||
pull_request:
|
||||
types: [labeled]
|
||||
|
||||
schedule:
|
||||
# minute (0-59)
|
||||
# hour (0-23)
|
||||
# day of the month (1-31)
|
||||
# month (1-12)
|
||||
# day of the week (0-6)
|
||||
# nightly build at 15:50 UTC time every day
|
||||
- cron: "50 15 * * *"
|
||||
|
||||
concurrency:
|
||||
group: run_librispeech_2022_11_14_zipformer_stateless8-${{ github.ref }}
|
||||
cancel-in-progress: true
|
||||
|
||||
jobs:
|
||||
run_librispeech_2022_11_14_zipformer_stateless8:
|
||||
if: github.event.label.name == 'ready' || github.event.label.name == 'run-decode' || github.event_name == 'push' || github.event_name == 'schedule'
|
||||
runs-on: ${{ matrix.os }}
|
||||
strategy:
|
||||
matrix:
|
||||
os: [ubuntu-latest]
|
||||
python-version: [3.8]
|
||||
|
||||
fail-fast: false
|
||||
|
||||
steps:
|
||||
- uses: actions/checkout@v2
|
||||
with:
|
||||
fetch-depth: 0
|
||||
|
||||
- name: Setup Python ${{ matrix.python-version }}
|
||||
uses: actions/setup-python@v2
|
||||
with:
|
||||
python-version: ${{ matrix.python-version }}
|
||||
cache: 'pip'
|
||||
cache-dependency-path: '**/requirements-ci.txt'
|
||||
|
||||
- name: Install Python dependencies
|
||||
run: |
|
||||
grep -v '^#' ./requirements-ci.txt | xargs -n 1 -L 1 pip install
|
||||
pip uninstall -y protobuf
|
||||
pip install --no-binary protobuf protobuf==3.20.*
|
||||
|
||||
- name: Cache kaldifeat
|
||||
id: my-cache
|
||||
uses: actions/cache@v2
|
||||
with:
|
||||
path: |
|
||||
~/tmp/kaldifeat
|
||||
key: cache-tmp-${{ matrix.python-version }}-2023-05-22
|
||||
|
||||
- name: Install kaldifeat
|
||||
if: steps.my-cache.outputs.cache-hit != 'true'
|
||||
shell: bash
|
||||
run: |
|
||||
.github/scripts/install-kaldifeat.sh
|
||||
|
||||
- name: Cache LibriSpeech test-clean and test-other datasets
|
||||
id: libri-test-clean-and-test-other-data
|
||||
uses: actions/cache@v2
|
||||
with:
|
||||
path: |
|
||||
~/tmp/download
|
||||
key: cache-libri-test-clean-and-test-other
|
||||
|
||||
- name: Download LibriSpeech test-clean and test-other
|
||||
if: steps.libri-test-clean-and-test-other-data.outputs.cache-hit != 'true'
|
||||
shell: bash
|
||||
run: |
|
||||
.github/scripts/download-librispeech-test-clean-and-test-other-dataset.sh
|
||||
|
||||
- name: Prepare manifests for LibriSpeech test-clean and test-other
|
||||
shell: bash
|
||||
run: |
|
||||
.github/scripts/prepare-librispeech-test-clean-and-test-other-manifests.sh
|
||||
|
||||
- name: Cache LibriSpeech test-clean and test-other fbank features
|
||||
id: libri-test-clean-and-test-other-fbank
|
||||
uses: actions/cache@v2
|
||||
with:
|
||||
path: |
|
||||
~/tmp/fbank-libri
|
||||
key: cache-libri-fbank-test-clean-and-test-other-v2
|
||||
|
||||
- name: Compute fbank for LibriSpeech test-clean and test-other
|
||||
if: steps.libri-test-clean-and-test-other-fbank.outputs.cache-hit != 'true'
|
||||
shell: bash
|
||||
run: |
|
||||
.github/scripts/compute-fbank-librispeech-test-clean-and-test-other.sh
|
||||
|
||||
- name: Inference with pre-trained model
|
||||
shell: bash
|
||||
env:
|
||||
GITHUB_EVENT_NAME: ${{ github.event_name }}
|
||||
GITHUB_EVENT_LABEL_NAME: ${{ github.event.label.name }}
|
||||
run: |
|
||||
mkdir -p egs/librispeech/ASR/data
|
||||
ln -sfv ~/tmp/fbank-libri egs/librispeech/ASR/data/fbank
|
||||
ls -lh egs/librispeech/ASR/data/*
|
||||
|
||||
sudo apt-get -qq install git-lfs tree
|
||||
export PYTHONPATH=$PWD:$PYTHONPATH
|
||||
export PYTHONPATH=~/tmp/kaldifeat/kaldifeat/python:$PYTHONPATH
|
||||
export PYTHONPATH=~/tmp/kaldifeat/build/lib:$PYTHONPATH
|
||||
|
||||
.github/scripts/run-librispeech-pruned-transducer-stateless8-2022-11-14.sh
|
||||
|
||||
- name: Display decoding results for librispeech pruned_transducer_stateless8
|
||||
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
|
||||
shell: bash
|
||||
run: |
|
||||
cd egs/librispeech/ASR/
|
||||
tree ./pruned_transducer_stateless8/exp
|
||||
|
||||
cd pruned_transducer_stateless8
|
||||
echo "results for pruned_transducer_stateless8"
|
||||
echo "===greedy search==="
|
||||
find exp/greedy_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
|
||||
find exp/greedy_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
|
||||
|
||||
echo "===fast_beam_search==="
|
||||
find exp/fast_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
|
||||
find exp/fast_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
|
||||
|
||||
echo "===modified beam search==="
|
||||
find exp/modified_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
|
||||
find exp/modified_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
|
||||
|
||||
- name: Upload decoding results for librispeech pruned_transducer_stateless8
|
||||
uses: actions/upload-artifact@v2
|
||||
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
|
||||
with:
|
||||
name: torch-${{ matrix.torch }}-python-${{ matrix.python-version }}-ubuntu-latest-cpu-pruned_transducer_stateless8-2022-11-14
|
||||
path: egs/librispeech/ASR/pruned_transducer_stateless8/exp/
|
@ -1,163 +0,0 @@
|
||||
# Copyright 2022 Fangjun Kuang (csukuangfj@gmail.com)
|
||||
|
||||
# See ../../LICENSE for clarification regarding multiple authors
|
||||
#
|
||||
# Licensed under the Apache License, Version 2.0 (the "License");
|
||||
# you may not use this file except in compliance with the License.
|
||||
# You may obtain a copy of the License at
|
||||
#
|
||||
# http://www.apache.org/licenses/LICENSE-2.0
|
||||
#
|
||||
# Unless required by applicable law or agreed to in writing, software
|
||||
# distributed under the License is distributed on an "AS IS" BASIS,
|
||||
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
||||
# See the License for the specific language governing permissions and
|
||||
# limitations under the License.
|
||||
|
||||
name: run-librispeech-2022-12-01-stateless7-ctc
|
||||
# zipformer
|
||||
|
||||
on:
|
||||
push:
|
||||
branches:
|
||||
- master
|
||||
pull_request:
|
||||
types: [labeled]
|
||||
|
||||
schedule:
|
||||
# minute (0-59)
|
||||
# hour (0-23)
|
||||
# day of the month (1-31)
|
||||
# month (1-12)
|
||||
# day of the week (0-6)
|
||||
# nightly build at 15:50 UTC time every day
|
||||
- cron: "50 15 * * *"
|
||||
|
||||
jobs:
|
||||
run_librispeech_2022_11_11_zipformer:
|
||||
if: github.event.label.name == 'ready' || github.event.label.name == 'run-decode' || github.event_name == 'push' || github.event_name == 'schedule'
|
||||
runs-on: ${{ matrix.os }}
|
||||
strategy:
|
||||
matrix:
|
||||
os: [ubuntu-latest]
|
||||
python-version: [3.8]
|
||||
|
||||
fail-fast: false
|
||||
|
||||
steps:
|
||||
- uses: actions/checkout@v2
|
||||
with:
|
||||
fetch-depth: 0
|
||||
|
||||
- name: Setup Python ${{ matrix.python-version }}
|
||||
uses: actions/setup-python@v2
|
||||
with:
|
||||
python-version: ${{ matrix.python-version }}
|
||||
cache: 'pip'
|
||||
cache-dependency-path: '**/requirements-ci.txt'
|
||||
|
||||
- name: Install Python dependencies
|
||||
run: |
|
||||
grep -v '^#' ./requirements-ci.txt | xargs -n 1 -L 1 pip install
|
||||
pip uninstall -y protobuf
|
||||
pip install --no-binary protobuf protobuf==3.20.*
|
||||
|
||||
- name: Cache kaldifeat
|
||||
id: my-cache
|
||||
uses: actions/cache@v2
|
||||
with:
|
||||
path: |
|
||||
~/tmp/kaldifeat
|
||||
key: cache-tmp-${{ matrix.python-version }}-2023-05-22
|
||||
|
||||
- name: Install kaldifeat
|
||||
if: steps.my-cache.outputs.cache-hit != 'true'
|
||||
shell: bash
|
||||
run: |
|
||||
.github/scripts/install-kaldifeat.sh
|
||||
|
||||
- name: Cache LibriSpeech test-clean and test-other datasets
|
||||
id: libri-test-clean-and-test-other-data
|
||||
uses: actions/cache@v2
|
||||
with:
|
||||
path: |
|
||||
~/tmp/download
|
||||
key: cache-libri-test-clean-and-test-other
|
||||
|
||||
- name: Download LibriSpeech test-clean and test-other
|
||||
if: steps.libri-test-clean-and-test-other-data.outputs.cache-hit != 'true'
|
||||
shell: bash
|
||||
run: |
|
||||
.github/scripts/download-librispeech-test-clean-and-test-other-dataset.sh
|
||||
|
||||
- name: Prepare manifests for LibriSpeech test-clean and test-other
|
||||
shell: bash
|
||||
run: |
|
||||
.github/scripts/prepare-librispeech-test-clean-and-test-other-manifests.sh
|
||||
|
||||
- name: Cache LibriSpeech test-clean and test-other fbank features
|
||||
id: libri-test-clean-and-test-other-fbank
|
||||
uses: actions/cache@v2
|
||||
with:
|
||||
path: |
|
||||
~/tmp/fbank-libri
|
||||
key: cache-libri-fbank-test-clean-and-test-other-v2
|
||||
|
||||
- name: Compute fbank for LibriSpeech test-clean and test-other
|
||||
if: steps.libri-test-clean-and-test-other-fbank.outputs.cache-hit != 'true'
|
||||
shell: bash
|
||||
run: |
|
||||
.github/scripts/compute-fbank-librispeech-test-clean-and-test-other.sh
|
||||
|
||||
- name: Inference with pre-trained model
|
||||
shell: bash
|
||||
env:
|
||||
GITHUB_EVENT_NAME: ${{ github.event_name }}
|
||||
GITHUB_EVENT_LABEL_NAME: ${{ github.event.label.name }}
|
||||
run: |
|
||||
mkdir -p egs/librispeech/ASR/data
|
||||
ln -sfv ~/tmp/fbank-libri egs/librispeech/ASR/data/fbank
|
||||
ls -lh egs/librispeech/ASR/data/*
|
||||
|
||||
sudo apt-get -qq install git-lfs tree
|
||||
export PYTHONPATH=$PWD:$PYTHONPATH
|
||||
export PYTHONPATH=~/tmp/kaldifeat/kaldifeat/python:$PYTHONPATH
|
||||
export PYTHONPATH=~/tmp/kaldifeat/build/lib:$PYTHONPATH
|
||||
|
||||
.github/scripts/run-librispeech-pruned-transducer-stateless7-ctc-2022-12-01.sh
|
||||
|
||||
- name: Display decoding results for librispeech pruned_transducer_stateless7_ctc
|
||||
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
|
||||
shell: bash
|
||||
run: |
|
||||
cd egs/librispeech/ASR/
|
||||
tree ./pruned_transducer_stateless7_ctc/exp
|
||||
|
||||
cd pruned_transducer_stateless7_ctc
|
||||
echo "results for pruned_transducer_stateless7_ctc"
|
||||
echo "===greedy search==="
|
||||
find exp/greedy_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
|
||||
find exp/greedy_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
|
||||
|
||||
echo "===fast_beam_search==="
|
||||
find exp/fast_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
|
||||
find exp/fast_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
|
||||
|
||||
echo "===modified beam search==="
|
||||
find exp/modified_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
|
||||
find exp/modified_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
|
||||
|
||||
echo "===ctc decoding==="
|
||||
find exp/ctc-decoding -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
|
||||
find exp/ctc-decoding -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
|
||||
|
||||
echo "===1best==="
|
||||
find exp/1best -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
|
||||
find exp/1best -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
|
||||
|
||||
- name: Upload decoding results for librispeech pruned_transducer_stateless7_ctc
|
||||
uses: actions/upload-artifact@v2
|
||||
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
|
||||
with:
|
||||
name: torch-${{ matrix.torch }}-python-${{ matrix.python-version }}-ubuntu-latest-cpu-pruned_transducer_stateless7-ctc-2022-12-01
|
||||
path: egs/librispeech/ASR/pruned_transducer_stateless7_ctc/exp/
|
@ -1,167 +0,0 @@
|
||||
# Copyright 2022 Zengwei Yao
|
||||
|
||||
# See ../../LICENSE for clarification regarding multiple authors
|
||||
#
|
||||
# Licensed under the Apache License, Version 2.0 (the "License");
|
||||
# you may not use this file except in compliance with the License.
|
||||
# You may obtain a copy of the License at
|
||||
#
|
||||
# http://www.apache.org/licenses/LICENSE-2.0
|
||||
#
|
||||
# Unless required by applicable law or agreed to in writing, software
|
||||
# distributed under the License is distributed on an "AS IS" BASIS,
|
||||
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
||||
# See the License for the specific language governing permissions and
|
||||
# limitations under the License.
|
||||
|
||||
name: run-librispeech-2022-12-08-zipformer-mmi
|
||||
# zipformer
|
||||
|
||||
on:
|
||||
push:
|
||||
branches:
|
||||
- master
|
||||
pull_request:
|
||||
types: [labeled]
|
||||
|
||||
schedule:
|
||||
# minute (0-59)
|
||||
# hour (0-23)
|
||||
# day of the month (1-31)
|
||||
# month (1-12)
|
||||
# day of the week (0-6)
|
||||
# nightly build at 15:50 UTC time every day
|
||||
- cron: "50 15 * * *"
|
||||
|
||||
concurrency:
|
||||
group: run_librispeech_2022_12_08_zipformer-${{ github.ref }}
|
||||
cancel-in-progress: true
|
||||
|
||||
jobs:
|
||||
run_librispeech_2022_12_08_zipformer:
|
||||
if: github.event.label.name == 'ready' || github.event.label.name == 'run-decode' || github.event_name == 'push' || github.event_name == 'schedule'
|
||||
runs-on: ${{ matrix.os }}
|
||||
strategy:
|
||||
matrix:
|
||||
os: [ubuntu-latest]
|
||||
python-version: [3.8]
|
||||
|
||||
fail-fast: false
|
||||
|
||||
steps:
|
||||
- uses: actions/checkout@v2
|
||||
with:
|
||||
fetch-depth: 0
|
||||
|
||||
- name: Setup Python ${{ matrix.python-version }}
|
||||
uses: actions/setup-python@v2
|
||||
with:
|
||||
python-version: ${{ matrix.python-version }}
|
||||
cache: 'pip'
|
||||
cache-dependency-path: '**/requirements-ci.txt'
|
||||
|
||||
- name: Install Python dependencies
|
||||
run: |
|
||||
grep -v '^#' ./requirements-ci.txt | xargs -n 1 -L 1 pip install
|
||||
pip uninstall -y protobuf
|
||||
pip install --no-binary protobuf protobuf==3.20.*
|
||||
|
||||
- name: Cache kaldifeat
|
||||
id: my-cache
|
||||
uses: actions/cache@v2
|
||||
with:
|
||||
path: |
|
||||
~/tmp/kaldifeat
|
||||
key: cache-tmp-${{ matrix.python-version }}-2023-05-22
|
||||
|
||||
- name: Install kaldifeat
|
||||
if: steps.my-cache.outputs.cache-hit != 'true'
|
||||
shell: bash
|
||||
run: |
|
||||
.github/scripts/install-kaldifeat.sh
|
||||
|
||||
- name: Cache LibriSpeech test-clean and test-other datasets
|
||||
id: libri-test-clean-and-test-other-data
|
||||
uses: actions/cache@v2
|
||||
with:
|
||||
path: |
|
||||
~/tmp/download
|
||||
key: cache-libri-test-clean-and-test-other
|
||||
|
||||
- name: Download LibriSpeech test-clean and test-other
|
||||
if: steps.libri-test-clean-and-test-other-data.outputs.cache-hit != 'true'
|
||||
shell: bash
|
||||
run: |
|
||||
.github/scripts/download-librispeech-test-clean-and-test-other-dataset.sh
|
||||
|
||||
- name: Prepare manifests for LibriSpeech test-clean and test-other
|
||||
shell: bash
|
||||
run: |
|
||||
.github/scripts/prepare-librispeech-test-clean-and-test-other-manifests.sh
|
||||
|
||||
- name: Cache LibriSpeech test-clean and test-other fbank features
|
||||
id: libri-test-clean-and-test-other-fbank
|
||||
uses: actions/cache@v2
|
||||
with:
|
||||
path: |
|
||||
~/tmp/fbank-libri
|
||||
key: cache-libri-fbank-test-clean-and-test-other-v2
|
||||
|
||||
- name: Compute fbank for LibriSpeech test-clean and test-other
|
||||
if: steps.libri-test-clean-and-test-other-fbank.outputs.cache-hit != 'true'
|
||||
shell: bash
|
||||
run: |
|
||||
.github/scripts/compute-fbank-librispeech-test-clean-and-test-other.sh
|
||||
|
||||
- name: Inference with pre-trained model
|
||||
shell: bash
|
||||
env:
|
||||
GITHUB_EVENT_NAME: ${{ github.event_name }}
|
||||
GITHUB_EVENT_LABEL_NAME: ${{ github.event.label.name }}
|
||||
run: |
|
||||
mkdir -p egs/librispeech/ASR/data
|
||||
ln -sfv ~/tmp/fbank-libri egs/librispeech/ASR/data/fbank
|
||||
ls -lh egs/librispeech/ASR/data/*
|
||||
|
||||
sudo apt-get -qq install git-lfs tree
|
||||
export PYTHONPATH=$PWD:$PYTHONPATH
|
||||
export PYTHONPATH=~/tmp/kaldifeat/kaldifeat/python:$PYTHONPATH
|
||||
export PYTHONPATH=~/tmp/kaldifeat/build/lib:$PYTHONPATH
|
||||
|
||||
.github/scripts/run-librispeech-zipformer-mmi-2022-12-08.sh
|
||||
|
||||
- name: Display decoding results for librispeech zipformer-mmi
|
||||
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
|
||||
shell: bash
|
||||
run: |
|
||||
cd egs/librispeech/ASR/
|
||||
tree ./zipformer-mmi/exp
|
||||
|
||||
cd zipformer-mmi
|
||||
echo "results for zipformer-mmi"
|
||||
echo "===1best==="
|
||||
find exp/1best -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
|
||||
find exp/1best -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
|
||||
|
||||
echo "===nbest==="
|
||||
find exp/nbest -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
|
||||
find exp/nbest -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
|
||||
|
||||
echo "===nbest-rescoring-LG==="
|
||||
find exp/nbest-rescoring-LG -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
|
||||
find exp/nbest-rescoring-LG -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
|
||||
|
||||
echo "===nbest-rescoring-3-gram==="
|
||||
find exp/nbest-rescoring-3-gram -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
|
||||
find exp/nbest-rescoring-3-gram -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
|
||||
|
||||
echo "===nbest-rescoring-4-gram==="
|
||||
find exp/nbest-rescoring-4-gram -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
|
||||
find exp/nbest-rescoring-4-gram -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
|
||||
|
||||
- name: Upload decoding results for librispeech zipformer-mmi
|
||||
uses: actions/upload-artifact@v2
|
||||
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
|
||||
with:
|
||||
name: torch-${{ matrix.torch }}-python-${{ matrix.python-version }}-ubuntu-latest-cpu-zipformer_mmi-2022-12-08
|
||||
path: egs/librispeech/ASR/zipformer_mmi/exp/
|
@ -1,172 +0,0 @@
|
||||
# Copyright 2022 Fangjun Kuang (csukuangfj@gmail.com)
|
||||
|
||||
# See ../../LICENSE for clarification regarding multiple authors
|
||||
#
|
||||
# Licensed under the Apache License, Version 2.0 (the "License");
|
||||
# you may not use this file except in compliance with the License.
|
||||
# You may obtain a copy of the License at
|
||||
#
|
||||
# http://www.apache.org/licenses/LICENSE-2.0
|
||||
#
|
||||
# Unless required by applicable law or agreed to in writing, software
|
||||
# distributed under the License is distributed on an "AS IS" BASIS,
|
||||
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
||||
# See the License for the specific language governing permissions and
|
||||
# limitations under the License.
|
||||
|
||||
name: run-librispeech-2022-12-29-stateless7-streaming
|
||||
# zipformer
|
||||
|
||||
on:
|
||||
push:
|
||||
branches:
|
||||
- master
|
||||
pull_request:
|
||||
types: [labeled]
|
||||
|
||||
schedule:
|
||||
# minute (0-59)
|
||||
# hour (0-23)
|
||||
# day of the month (1-31)
|
||||
# month (1-12)
|
||||
# day of the week (0-6)
|
||||
# nightly build at 15:50 UTC time every day
|
||||
- cron: "50 15 * * *"
|
||||
|
||||
concurrency:
|
||||
group: run_librispeech_2022_12_29_zipformer_streaming-${{ github.ref }}
|
||||
cancel-in-progress: true
|
||||
|
||||
jobs:
|
||||
run_librispeech_2022_12_29_zipformer_streaming:
|
||||
if: github.event.label.name == 'ready' || github.event.label.name == 'run-decode' || github.event.label.name == 'streaming-zipformer' || github.event_name == 'push' || github.event_name == 'schedule'
|
||||
runs-on: ${{ matrix.os }}
|
||||
strategy:
|
||||
matrix:
|
||||
os: [ubuntu-latest]
|
||||
python-version: [3.8]
|
||||
|
||||
fail-fast: false
|
||||
|
||||
steps:
|
||||
- uses: actions/checkout@v2
|
||||
with:
|
||||
fetch-depth: 0
|
||||
|
||||
- name: Setup Python ${{ matrix.python-version }}
|
||||
uses: actions/setup-python@v2
|
||||
with:
|
||||
python-version: ${{ matrix.python-version }}
|
||||
cache: 'pip'
|
||||
cache-dependency-path: '**/requirements-ci.txt'
|
||||
|
||||
- name: Install Python dependencies
|
||||
run: |
|
||||
grep -v '^#' ./requirements-ci.txt | xargs -n 1 -L 1 pip install
|
||||
pip uninstall -y protobuf
|
||||
pip install --no-binary protobuf protobuf==3.20.*
|
||||
|
||||
- name: Cache kaldifeat
|
||||
id: my-cache
|
||||
uses: actions/cache@v2
|
||||
with:
|
||||
path: |
|
||||
~/tmp/kaldifeat
|
||||
key: cache-tmp-${{ matrix.python-version }}-2023-05-22
|
||||
|
||||
- name: Install kaldifeat
|
||||
if: steps.my-cache.outputs.cache-hit != 'true'
|
||||
shell: bash
|
||||
run: |
|
||||
.github/scripts/install-kaldifeat.sh
|
||||
|
||||
- name: Cache LibriSpeech test-clean and test-other datasets
|
||||
id: libri-test-clean-and-test-other-data
|
||||
uses: actions/cache@v2
|
||||
with:
|
||||
path: |
|
||||
~/tmp/download
|
||||
key: cache-libri-test-clean-and-test-other
|
||||
|
||||
- name: Download LibriSpeech test-clean and test-other
|
||||
if: steps.libri-test-clean-and-test-other-data.outputs.cache-hit != 'true'
|
||||
shell: bash
|
||||
run: |
|
||||
.github/scripts/download-librispeech-test-clean-and-test-other-dataset.sh
|
||||
|
||||
- name: Prepare manifests for LibriSpeech test-clean and test-other
|
||||
shell: bash
|
||||
run: |
|
||||
.github/scripts/prepare-librispeech-test-clean-and-test-other-manifests.sh
|
||||
|
||||
- name: Cache LibriSpeech test-clean and test-other fbank features
|
||||
id: libri-test-clean-and-test-other-fbank
|
||||
uses: actions/cache@v2
|
||||
with:
|
||||
path: |
|
||||
~/tmp/fbank-libri
|
||||
key: cache-libri-fbank-test-clean-and-test-other-v2
|
||||
|
||||
- name: Compute fbank for LibriSpeech test-clean and test-other
|
||||
if: steps.libri-test-clean-and-test-other-fbank.outputs.cache-hit != 'true'
|
||||
shell: bash
|
||||
run: |
|
||||
.github/scripts/compute-fbank-librispeech-test-clean-and-test-other.sh
|
||||
|
||||
- name: Inference with pre-trained model
|
||||
shell: bash
|
||||
env:
|
||||
GITHUB_EVENT_NAME: ${{ github.event_name }}
|
||||
GITHUB_EVENT_LABEL_NAME: ${{ github.event.label.name }}
|
||||
run: |
|
||||
mkdir -p egs/librispeech/ASR/data
|
||||
ln -sfv ~/tmp/fbank-libri egs/librispeech/ASR/data/fbank
|
||||
ls -lh egs/librispeech/ASR/data/*
|
||||
|
||||
sudo apt-get -qq install git-lfs tree
|
||||
export PYTHONPATH=$PWD:$PYTHONPATH
|
||||
export PYTHONPATH=~/tmp/kaldifeat/kaldifeat/python:$PYTHONPATH
|
||||
export PYTHONPATH=~/tmp/kaldifeat/build/lib:$PYTHONPATH
|
||||
|
||||
.github/scripts/run-librispeech-pruned-transducer-stateless7-streaming-2022-12-29.sh
|
||||
|
||||
- name: Display decoding results for librispeech pruned_transducer_stateless7_streaming
|
||||
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
|
||||
shell: bash
|
||||
run: |
|
||||
cd egs/librispeech/ASR/
|
||||
tree ./pruned_transducer_stateless7_streaming/exp
|
||||
|
||||
cd pruned_transducer_stateless7_streaming
|
||||
echo "results for pruned_transducer_stateless7_streaming"
|
||||
echo "===greedy search==="
|
||||
find exp/greedy_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
|
||||
find exp/greedy_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
|
||||
|
||||
echo "===fast_beam_search==="
|
||||
find exp/fast_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
|
||||
find exp/fast_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
|
||||
|
||||
echo "===modified beam search==="
|
||||
find exp/modified_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
|
||||
find exp/modified_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
|
||||
|
||||
echo "===streaming greedy search==="
|
||||
find exp/streaming/greedy_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
|
||||
find exp/streaming/greedy_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
|
||||
|
||||
echo "===streaming fast_beam_search==="
|
||||
find exp/streaming/fast_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
|
||||
find exp/streaming/fast_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
|
||||
|
||||
echo "===streaming modified beam search==="
|
||||
find exp/streaming/modified_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
|
||||
find exp/streaming/modified_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
|
||||
|
||||
|
||||
- name: Upload decoding results for librispeech pruned_transducer_stateless7_streaming
|
||||
uses: actions/upload-artifact@v2
|
||||
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
|
||||
with:
|
||||
name: torch-${{ matrix.torch }}-python-${{ matrix.python-version }}-ubuntu-latest-cpu-pruned_transducer_stateless7-streaming-2022-12-29
|
||||
path: egs/librispeech/ASR/pruned_transducer_stateless7_streaming/exp/
|
@ -1,163 +0,0 @@
|
||||
# Copyright 2022 Fangjun Kuang (csukuangfj@gmail.com)
|
||||
|
||||
# See ../../LICENSE for clarification regarding multiple authors
|
||||
#
|
||||
# Licensed under the Apache License, Version 2.0 (the "License");
|
||||
# you may not use this file except in compliance with the License.
|
||||
# You may obtain a copy of the License at
|
||||
#
|
||||
# http://www.apache.org/licenses/LICENSE-2.0
|
||||
#
|
||||
# Unless required by applicable law or agreed to in writing, software
|
||||
# distributed under the License is distributed on an "AS IS" BASIS,
|
||||
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
||||
# See the License for the specific language governing permissions and
|
||||
# limitations under the License.
|
||||
|
||||
name: run-librispeech-2023-01-29-stateless7-ctc-bs
|
||||
# zipformer
|
||||
|
||||
on:
|
||||
push:
|
||||
branches:
|
||||
- master
|
||||
pull_request:
|
||||
types: [labeled]
|
||||
|
||||
schedule:
|
||||
# minute (0-59)
|
||||
# hour (0-23)
|
||||
# day of the month (1-31)
|
||||
# month (1-12)
|
||||
# day of the week (0-6)
|
||||
# nightly build at 15:50 UTC time every day
|
||||
- cron: "50 15 * * *"
|
||||
|
||||
jobs:
|
||||
run_librispeech_2023_01_29_zipformer_ctc_bs:
|
||||
if: github.event.label.name == 'run-decode' || github.event.label.name == 'blank-skip' || github.event_name == 'push' || github.event_name == 'schedule'
|
||||
runs-on: ${{ matrix.os }}
|
||||
strategy:
|
||||
matrix:
|
||||
os: [ubuntu-latest]
|
||||
python-version: [3.8]
|
||||
|
||||
fail-fast: false
|
||||
|
||||
steps:
|
||||
- uses: actions/checkout@v2
|
||||
with:
|
||||
fetch-depth: 0
|
||||
|
||||
- name: Setup Python ${{ matrix.python-version }}
|
||||
uses: actions/setup-python@v2
|
||||
with:
|
||||
python-version: ${{ matrix.python-version }}
|
||||
cache: 'pip'
|
||||
cache-dependency-path: '**/requirements-ci.txt'
|
||||
|
||||
- name: Install Python dependencies
|
||||
run: |
|
||||
grep -v '^#' ./requirements-ci.txt | xargs -n 1 -L 1 pip install
|
||||
pip uninstall -y protobuf
|
||||
pip install --no-binary protobuf protobuf==3.20.*
|
||||
|
||||
- name: Cache kaldifeat
|
||||
id: my-cache
|
||||
uses: actions/cache@v2
|
||||
with:
|
||||
path: |
|
||||
~/tmp/kaldifeat
|
||||
key: cache-tmp-${{ matrix.python-version }}-2023-05-22
|
||||
|
||||
- name: Install kaldifeat
|
||||
if: steps.my-cache.outputs.cache-hit != 'true'
|
||||
shell: bash
|
||||
run: |
|
||||
.github/scripts/install-kaldifeat.sh
|
||||
|
||||
- name: Cache LibriSpeech test-clean and test-other datasets
|
||||
id: libri-test-clean-and-test-other-data
|
||||
uses: actions/cache@v2
|
||||
with:
|
||||
path: |
|
||||
~/tmp/download
|
||||
key: cache-libri-test-clean-and-test-other
|
||||
|
||||
- name: Download LibriSpeech test-clean and test-other
|
||||
if: steps.libri-test-clean-and-test-other-data.outputs.cache-hit != 'true'
|
||||
shell: bash
|
||||
run: |
|
||||
.github/scripts/download-librispeech-test-clean-and-test-other-dataset.sh
|
||||
|
||||
- name: Prepare manifests for LibriSpeech test-clean and test-other
|
||||
shell: bash
|
||||
run: |
|
||||
.github/scripts/prepare-librispeech-test-clean-and-test-other-manifests.sh
|
||||
|
||||
- name: Cache LibriSpeech test-clean and test-other fbank features
|
||||
id: libri-test-clean-and-test-other-fbank
|
||||
uses: actions/cache@v2
|
||||
with:
|
||||
path: |
|
||||
~/tmp/fbank-libri
|
||||
key: cache-libri-fbank-test-clean-and-test-other-v2
|
||||
|
||||
- name: Compute fbank for LibriSpeech test-clean and test-other
|
||||
if: steps.libri-test-clean-and-test-other-fbank.outputs.cache-hit != 'true'
|
||||
shell: bash
|
||||
run: |
|
||||
.github/scripts/compute-fbank-librispeech-test-clean-and-test-other.sh
|
||||
|
||||
- name: Inference with pre-trained model
|
||||
shell: bash
|
||||
env:
|
||||
GITHUB_EVENT_NAME: ${{ github.event_name }}
|
||||
GITHUB_EVENT_LABEL_NAME: ${{ github.event.label.name }}
|
||||
run: |
|
||||
mkdir -p egs/librispeech/ASR/data
|
||||
ln -sfv ~/tmp/fbank-libri egs/librispeech/ASR/data/fbank
|
||||
ls -lh egs/librispeech/ASR/data/*
|
||||
|
||||
sudo apt-get -qq install git-lfs tree
|
||||
export PYTHONPATH=$PWD:$PYTHONPATH
|
||||
export PYTHONPATH=~/tmp/kaldifeat/kaldifeat/python:$PYTHONPATH
|
||||
export PYTHONPATH=~/tmp/kaldifeat/build/lib:$PYTHONPATH
|
||||
|
||||
.github/scripts/run-librispeech-pruned-transducer-stateless7-ctc-bs-2023-01-29.sh
|
||||
|
||||
- name: Display decoding results for librispeech pruned_transducer_stateless7_ctc_bs
|
||||
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
|
||||
shell: bash
|
||||
run: |
|
||||
cd egs/librispeech/ASR/
|
||||
tree ./pruned_transducer_stateless7_ctc_bs/exp
|
||||
|
||||
cd pruned_transducer_stateless7_ctc_bs
|
||||
echo "results for pruned_transducer_stateless7_ctc_bs"
|
||||
echo "===greedy search==="
|
||||
find exp/greedy_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
|
||||
find exp/greedy_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
|
||||
|
||||
echo "===fast_beam_search==="
|
||||
find exp/fast_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
|
||||
find exp/fast_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
|
||||
|
||||
echo "===modified beam search==="
|
||||
find exp/modified_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
|
||||
find exp/modified_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
|
||||
|
||||
echo "===ctc decoding==="
|
||||
find exp/ctc-decoding -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
|
||||
find exp/ctc-decoding -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
|
||||
|
||||
echo "===1best==="
|
||||
find exp/1best -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
|
||||
find exp/1best -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
|
||||
|
||||
- name: Upload decoding results for librispeech pruned_transducer_stateless7_ctc_bs
|
||||
uses: actions/upload-artifact@v2
|
||||
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
|
||||
with:
|
||||
name: torch-${{ matrix.torch }}-python-${{ matrix.python-version }}-ubuntu-latest-cpu-pruned_transducer_stateless7-ctc-bs-2023-01-29
|
||||
path: egs/librispeech/ASR/pruned_transducer_stateless7_ctc_bs/exp/
|
@ -1,155 +0,0 @@
|
||||
# Copyright 2022 Fangjun Kuang (csukuangfj@gmail.com)
|
||||
|
||||
# See ../../LICENSE for clarification regarding multiple authors
|
||||
#
|
||||
# Licensed under the Apache License, Version 2.0 (the "License");
|
||||
# you may not use this file except in compliance with the License.
|
||||
# You may obtain a copy of the License at
|
||||
#
|
||||
# http://www.apache.org/licenses/LICENSE-2.0
|
||||
#
|
||||
# Unless required by applicable law or agreed to in writing, software
|
||||
# distributed under the License is distributed on an "AS IS" BASIS,
|
||||
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
||||
# See the License for the specific language governing permissions and
|
||||
# limitations under the License.
|
||||
|
||||
name: run-librispeech-conformer-ctc3-2022-11-28
|
||||
# zipformer
|
||||
|
||||
on:
|
||||
push:
|
||||
branches:
|
||||
- master
|
||||
pull_request:
|
||||
types: [labeled]
|
||||
|
||||
schedule:
|
||||
# minute (0-59)
|
||||
# hour (0-23)
|
||||
# day of the month (1-31)
|
||||
# month (1-12)
|
||||
# day of the week (0-6)
|
||||
# nightly build at 15:50 UTC time every day
|
||||
- cron: "50 15 * * *"
|
||||
|
||||
concurrency:
|
||||
group: run_librispeech_2022_11_28_conformer_ctc3-${{ github.ref }}
|
||||
cancel-in-progress: true
|
||||
|
||||
jobs:
|
||||
run_librispeech_2022_11_28_conformer_ctc3:
|
||||
if: github.event.label.name == 'ready' || github.event.label.name == 'run-decode' || github.event_name == 'push' || github.event_name == 'schedule'
|
||||
runs-on: ${{ matrix.os }}
|
||||
strategy:
|
||||
matrix:
|
||||
os: [ubuntu-latest]
|
||||
python-version: [3.8]
|
||||
|
||||
fail-fast: false
|
||||
|
||||
steps:
|
||||
- uses: actions/checkout@v2
|
||||
with:
|
||||
fetch-depth: 0
|
||||
|
||||
- name: Setup Python ${{ matrix.python-version }}
|
||||
uses: actions/setup-python@v2
|
||||
with:
|
||||
python-version: ${{ matrix.python-version }}
|
||||
cache: 'pip'
|
||||
cache-dependency-path: '**/requirements-ci.txt'
|
||||
|
||||
- name: Install Python dependencies
|
||||
run: |
|
||||
grep -v '^#' ./requirements-ci.txt | xargs -n 1 -L 1 pip install
|
||||
pip uninstall -y protobuf
|
||||
pip install --no-binary protobuf protobuf==3.20.*
|
||||
|
||||
- name: Cache kaldifeat
|
||||
id: my-cache
|
||||
uses: actions/cache@v2
|
||||
with:
|
||||
path: |
|
||||
~/tmp/kaldifeat
|
||||
key: cache-tmp-${{ matrix.python-version }}-2023-05-22
|
||||
|
||||
- name: Install kaldifeat
|
||||
if: steps.my-cache.outputs.cache-hit != 'true'
|
||||
shell: bash
|
||||
run: |
|
||||
.github/scripts/install-kaldifeat.sh
|
||||
|
||||
- name: Cache LibriSpeech test-clean and test-other datasets
|
||||
id: libri-test-clean-and-test-other-data
|
||||
uses: actions/cache@v2
|
||||
with:
|
||||
path: |
|
||||
~/tmp/download
|
||||
key: cache-libri-test-clean-and-test-other
|
||||
|
||||
- name: Download LibriSpeech test-clean and test-other
|
||||
if: steps.libri-test-clean-and-test-other-data.outputs.cache-hit != 'true'
|
||||
shell: bash
|
||||
run: |
|
||||
.github/scripts/download-librispeech-test-clean-and-test-other-dataset.sh
|
||||
|
||||
- name: Prepare manifests for LibriSpeech test-clean and test-other
|
||||
shell: bash
|
||||
run: |
|
||||
.github/scripts/prepare-librispeech-test-clean-and-test-other-manifests.sh
|
||||
|
||||
- name: Cache LibriSpeech test-clean and test-other fbank features
|
||||
id: libri-test-clean-and-test-other-fbank
|
||||
uses: actions/cache@v2
|
||||
with:
|
||||
path: |
|
||||
~/tmp/fbank-libri
|
||||
key: cache-libri-fbank-test-clean-and-test-other-v2
|
||||
|
||||
- name: Compute fbank for LibriSpeech test-clean and test-other
|
||||
if: steps.libri-test-clean-and-test-other-fbank.outputs.cache-hit != 'true'
|
||||
shell: bash
|
||||
run: |
|
||||
.github/scripts/compute-fbank-librispeech-test-clean-and-test-other.sh
|
||||
|
||||
- name: Inference with pre-trained model
|
||||
shell: bash
|
||||
env:
|
||||
GITHUB_EVENT_NAME: ${{ github.event_name }}
|
||||
GITHUB_EVENT_LABEL_NAME: ${{ github.event.label.name }}
|
||||
run: |
|
||||
mkdir -p egs/librispeech/ASR/data
|
||||
ln -sfv ~/tmp/fbank-libri egs/librispeech/ASR/data/fbank
|
||||
ls -lh egs/librispeech/ASR/data/*
|
||||
|
||||
sudo apt-get -qq install git-lfs tree
|
||||
export PYTHONPATH=$PWD:$PYTHONPATH
|
||||
export PYTHONPATH=~/tmp/kaldifeat/kaldifeat/python:$PYTHONPATH
|
||||
export PYTHONPATH=~/tmp/kaldifeat/build/lib:$PYTHONPATH
|
||||
|
||||
.github/scripts/run-librispeech-conformer-ctc3-2022-11-28.sh
|
||||
|
||||
- name: Display decoding results for librispeech conformer_ctc3
|
||||
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
|
||||
shell: bash
|
||||
run: |
|
||||
cd egs/librispeech/ASR/
|
||||
tree ./conformer_ctc3/exp
|
||||
|
||||
cd conformer_ctc3
|
||||
echo "results for conformer_ctc3"
|
||||
echo "===ctc-decoding==="
|
||||
find exp/ctc-decoding -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
|
||||
find exp/ctc-decoding -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
|
||||
|
||||
echo "===1best==="
|
||||
find exp/1best -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
|
||||
find exp/1best -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
|
||||
|
||||
- name: Upload decoding results for librispeech conformer_ctc3
|
||||
uses: actions/upload-artifact@v2
|
||||
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
|
||||
with:
|
||||
name: torch-${{ matrix.torch }}-python-${{ matrix.python-version }}-ubuntu-latest-cpu-conformer_ctc3-2022-11-28
|
||||
path: egs/librispeech/ASR/conformer_ctc3/exp/
|
@ -1,157 +0,0 @@
|
||||
# Copyright 2021 Fangjun Kuang (csukuangfj@gmail.com)
|
||||
|
||||
# See ../../LICENSE for clarification regarding multiple authors
|
||||
#
|
||||
# Licensed under the Apache License, Version 2.0 (the "License");
|
||||
# you may not use this file except in compliance with the License.
|
||||
# You may obtain a copy of the License at
|
||||
#
|
||||
# http://www.apache.org/licenses/LICENSE-2.0
|
||||
#
|
||||
# Unless required by applicable law or agreed to in writing, software
|
||||
# distributed under the License is distributed on an "AS IS" BASIS,
|
||||
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
||||
# See the License for the specific language governing permissions and
|
||||
# limitations under the License.
|
||||
|
||||
name: run-librispeech-pruned-transducer-stateless3-2022-05-13
|
||||
# stateless pruned transducer (reworked model) + giga speech
|
||||
|
||||
on:
|
||||
push:
|
||||
branches:
|
||||
- master
|
||||
pull_request:
|
||||
types: [labeled]
|
||||
|
||||
schedule:
|
||||
# minute (0-59)
|
||||
# hour (0-23)
|
||||
# day of the month (1-31)
|
||||
# month (1-12)
|
||||
# day of the week (0-6)
|
||||
# nightly build at 15:50 UTC time every day
|
||||
- cron: "50 15 * * *"
|
||||
|
||||
concurrency:
|
||||
group: run_librispeech_pruned_transducer_stateless3_2022_05_13-${{ github.ref }}
|
||||
cancel-in-progress: true
|
||||
|
||||
jobs:
|
||||
run_librispeech_pruned_transducer_stateless3_2022_05_13:
|
||||
if: github.event.label.name == 'ready' || github.event.label.name == 'run-decode' || github.event_name == 'push' || github.event_name == 'schedule'
|
||||
runs-on: ${{ matrix.os }}
|
||||
strategy:
|
||||
matrix:
|
||||
os: [ubuntu-latest]
|
||||
python-version: [3.8]
|
||||
|
||||
fail-fast: false
|
||||
|
||||
steps:
|
||||
- uses: actions/checkout@v2
|
||||
with:
|
||||
fetch-depth: 0
|
||||
|
||||
- name: Setup Python ${{ matrix.python-version }}
|
||||
uses: actions/setup-python@v2
|
||||
with:
|
||||
python-version: ${{ matrix.python-version }}
|
||||
cache: 'pip'
|
||||
cache-dependency-path: '**/requirements-ci.txt'
|
||||
|
||||
- name: Install Python dependencies
|
||||
run: |
|
||||
grep -v '^#' ./requirements-ci.txt | xargs -n 1 -L 1 pip install
|
||||
pip uninstall -y protobuf
|
||||
pip install --no-binary protobuf protobuf==3.20.*
|
||||
|
||||
- name: Cache kaldifeat
|
||||
id: my-cache
|
||||
uses: actions/cache@v2
|
||||
with:
|
||||
path: |
|
||||
~/tmp/kaldifeat
|
||||
key: cache-tmp-${{ matrix.python-version }}-2023-05-22
|
||||
|
||||
- name: Install kaldifeat
|
||||
if: steps.my-cache.outputs.cache-hit != 'true'
|
||||
shell: bash
|
||||
run: |
|
||||
.github/scripts/install-kaldifeat.sh
|
||||
|
||||
- name: Cache LibriSpeech test-clean and test-other datasets
|
||||
id: libri-test-clean-and-test-other-data
|
||||
uses: actions/cache@v2
|
||||
with:
|
||||
path: |
|
||||
~/tmp/download
|
||||
key: cache-libri-test-clean-and-test-other
|
||||
|
||||
- name: Download LibriSpeech test-clean and test-other
|
||||
if: steps.libri-test-clean-and-test-other-data.outputs.cache-hit != 'true'
|
||||
shell: bash
|
||||
run: |
|
||||
.github/scripts/download-librispeech-test-clean-and-test-other-dataset.sh
|
||||
|
||||
- name: Prepare manifests for LibriSpeech test-clean and test-other
|
||||
shell: bash
|
||||
run: |
|
||||
.github/scripts/prepare-librispeech-test-clean-and-test-other-manifests.sh
|
||||
|
||||
- name: Cache LibriSpeech test-clean and test-other fbank features
|
||||
id: libri-test-clean-and-test-other-fbank
|
||||
uses: actions/cache@v2
|
||||
with:
|
||||
path: |
|
||||
~/tmp/fbank-libri
|
||||
key: cache-libri-fbank-test-clean-and-test-other-v2
|
||||
|
||||
- name: Compute fbank for LibriSpeech test-clean and test-other
|
||||
if: steps.libri-test-clean-and-test-other-fbank.outputs.cache-hit != 'true'
|
||||
shell: bash
|
||||
run: |
|
||||
.github/scripts/compute-fbank-librispeech-test-clean-and-test-other.sh
|
||||
|
||||
- name: Inference with pre-trained model
|
||||
shell: bash
|
||||
env:
|
||||
GITHUB_EVENT_NAME: ${{ github.event_name }}
|
||||
GITHUB_EVENT_LABEL_NAME: ${{ github.event.label.name }}
|
||||
run: |
|
||||
mkdir -p egs/librispeech/ASR/data
|
||||
ln -sfv ~/tmp/fbank-libri egs/librispeech/ASR/data/fbank
|
||||
ls -lh egs/librispeech/ASR/data/*
|
||||
|
||||
sudo apt-get -qq install git-lfs tree
|
||||
export PYTHONPATH=$PWD:$PYTHONPATH
|
||||
export PYTHONPATH=~/tmp/kaldifeat/kaldifeat/python:$PYTHONPATH
|
||||
export PYTHONPATH=~/tmp/kaldifeat/build/lib:$PYTHONPATH
|
||||
|
||||
.github/scripts/run-librispeech-pruned-transducer-stateless3-2022-05-13.sh
|
||||
|
||||
- name: Display decoding results for pruned_transducer_stateless3
|
||||
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
|
||||
shell: bash
|
||||
run: |
|
||||
cd egs/librispeech/ASR
|
||||
tree pruned_transducer_stateless3/exp
|
||||
cd pruned_transducer_stateless3/exp
|
||||
echo "===greedy search==="
|
||||
find greedy_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
|
||||
find greedy_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
|
||||
|
||||
echo "===fast_beam_search==="
|
||||
find fast_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
|
||||
find fast_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
|
||||
|
||||
echo "===modified beam search==="
|
||||
find modified_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
|
||||
find modified_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
|
||||
|
||||
- name: Upload decoding results for pruned_transducer_stateless3
|
||||
uses: actions/upload-artifact@v2
|
||||
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
|
||||
with:
|
||||
name: torch-${{ matrix.torch }}-python-${{ matrix.python-version }}-ubuntu-latest-cpu-pruned_transducer_stateless3-2022-04-29
|
||||
path: egs/librispeech/ASR/pruned_transducer_stateless3/exp/
|
@ -1,159 +0,0 @@
|
||||
# Copyright 2021 Fangjun Kuang (csukuangfj@gmail.com)
|
||||
|
||||
# See ../../LICENSE for clarification regarding multiple authors
|
||||
#
|
||||
# Licensed under the Apache License, Version 2.0 (the "License");
|
||||
# you may not use this file except in compliance with the License.
|
||||
# You may obtain a copy of the License at
|
||||
#
|
||||
# http://www.apache.org/licenses/LICENSE-2.0
|
||||
#
|
||||
# Unless required by applicable law or agreed to in writing, software
|
||||
# distributed under the License is distributed on an "AS IS" BASIS,
|
||||
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
||||
# See the License for the specific language governing permissions and
|
||||
# limitations under the License.
|
||||
|
||||
name: run-librispeech-streaming-2022-06-26
|
||||
# streaming conformer stateless transducer2
|
||||
|
||||
on:
|
||||
push:
|
||||
branches:
|
||||
- master
|
||||
pull_request:
|
||||
types: [labeled]
|
||||
|
||||
schedule:
|
||||
# minute (0-59)
|
||||
# hour (0-23)
|
||||
# day of the month (1-31)
|
||||
# month (1-12)
|
||||
# day of the week (0-6)
|
||||
# nightly build at 15:50 UTC time every day
|
||||
- cron: "50 15 * * *"
|
||||
|
||||
concurrency:
|
||||
group: run_librispeech_streaming_2022_06_26-${{ github.ref }}
|
||||
cancel-in-progress: true
|
||||
|
||||
jobs:
|
||||
run_librispeech_streaming_2022_06_26:
|
||||
if: github.event.label.name == 'ready' || github.event.label.name == 'run-decode' || github.event_name == 'push' || github.event_name == 'schedule'
|
||||
runs-on: ${{ matrix.os }}
|
||||
strategy:
|
||||
matrix:
|
||||
os: [ubuntu-latest]
|
||||
python-version: [3.8]
|
||||
|
||||
fail-fast: false
|
||||
|
||||
steps:
|
||||
- uses: actions/checkout@v2
|
||||
with:
|
||||
fetch-depth: 0
|
||||
|
||||
- name: Setup Python ${{ matrix.python-version }}
|
||||
uses: actions/setup-python@v2
|
||||
with:
|
||||
python-version: ${{ matrix.python-version }}
|
||||
cache: 'pip'
|
||||
cache-dependency-path: '**/requirements-ci.txt'
|
||||
|
||||
- name: Install Python dependencies
|
||||
run: |
|
||||
grep -v '^#' ./requirements-ci.txt | xargs -n 1 -L 1 pip install
|
||||
pip uninstall -y protobuf
|
||||
pip install --no-binary protobuf protobuf==3.20.*
|
||||
|
||||
- name: Cache kaldifeat
|
||||
id: my-cache
|
||||
uses: actions/cache@v2
|
||||
with:
|
||||
path: |
|
||||
~/tmp/kaldifeat
|
||||
key: cache-tmp-${{ matrix.python-version }}-2023-05-22
|
||||
|
||||
- name: Install kaldifeat
|
||||
if: steps.my-cache.outputs.cache-hit != 'true'
|
||||
shell: bash
|
||||
run: |
|
||||
.github/scripts/install-kaldifeat.sh
|
||||
|
||||
- name: Cache LibriSpeech test-clean and test-other datasets
|
||||
id: libri-test-clean-and-test-other-data
|
||||
uses: actions/cache@v2
|
||||
with:
|
||||
path: |
|
||||
~/tmp/download
|
||||
key: cache-libri-test-clean-and-test-other
|
||||
|
||||
- name: Download LibriSpeech test-clean and test-other
|
||||
if: steps.libri-test-clean-and-test-other-data.outputs.cache-hit != 'true'
|
||||
shell: bash
|
||||
run: |
|
||||
.github/scripts/download-librispeech-test-clean-and-test-other-dataset.sh
|
||||
|
||||
- name: Prepare manifests for LibriSpeech test-clean and test-other
|
||||
shell: bash
|
||||
run: |
|
||||
.github/scripts/prepare-librispeech-test-clean-and-test-other-manifests.sh
|
||||
|
||||
- name: Cache LibriSpeech test-clean and test-other fbank features
|
||||
id: libri-test-clean-and-test-other-fbank
|
||||
uses: actions/cache@v2
|
||||
with:
|
||||
path: |
|
||||
~/tmp/fbank-libri
|
||||
key: cache-libri-fbank-test-clean-and-test-other-v2
|
||||
|
||||
- name: Compute fbank for LibriSpeech test-clean and test-other
|
||||
if: steps.libri-test-clean-and-test-other-fbank.outputs.cache-hit != 'true'
|
||||
shell: bash
|
||||
run: |
|
||||
.github/scripts/compute-fbank-librispeech-test-clean-and-test-other.sh
|
||||
|
||||
- name: Inference with pre-trained model
|
||||
shell: bash
|
||||
env:
|
||||
GITHUB_EVENT_NAME: ${{ github.event_name }}
|
||||
GITHUB_EVENT_LABEL_NAME: ${{ github.event.label.name }}
|
||||
run: |
|
||||
mkdir -p egs/librispeech/ASR/data
|
||||
ln -sfv ~/tmp/fbank-libri egs/librispeech/ASR/data/fbank
|
||||
ls -lh egs/librispeech/ASR/data/*
|
||||
|
||||
sudo apt-get -qq install git-lfs tree
|
||||
export PYTHONPATH=$PWD:$PYTHONPATH
|
||||
export PYTHONPATH=~/tmp/kaldifeat/kaldifeat/python:$PYTHONPATH
|
||||
export PYTHONPATH=~/tmp/kaldifeat/build/lib:$PYTHONPATH
|
||||
|
||||
.github/scripts/run-librispeech-streaming-pruned-transducer-stateless2-2022-06-26.sh
|
||||
|
||||
- name: Display decoding results
|
||||
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
|
||||
shell: bash
|
||||
run: |
|
||||
cd egs/librispeech/ASR/
|
||||
tree ./pruned_transducer_stateless2/exp
|
||||
|
||||
cd pruned_transducer_stateless2
|
||||
echo "results for pruned_transducer_stateless2"
|
||||
echo "===greedy search==="
|
||||
find exp/greedy_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
|
||||
find exp/greedy_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
|
||||
|
||||
echo "===fast_beam_search==="
|
||||
find exp/fast_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
|
||||
find exp/fast_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
|
||||
|
||||
echo "===modified_beam_search==="
|
||||
find exp/modified_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
|
||||
find exp/modified_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
|
||||
|
||||
- name: Upload decoding results for pruned_transducer_stateless2
|
||||
uses: actions/upload-artifact@v2
|
||||
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
|
||||
with:
|
||||
name: torch-${{ matrix.torch }}-python-${{ matrix.python-version }}-ubuntu-latest-cpu-pruned_transducer_stateless2-2022-06-26
|
||||
path: egs/librispeech/ASR/pruned_transducer_stateless2/exp/
|
@ -1,174 +0,0 @@
|
||||
# Copyright 2022 Fangjun Kuang (csukuangfj@gmail.com)
|
||||
|
||||
# See ../../LICENSE for clarification regarding multiple authors
|
||||
#
|
||||
# Licensed under the Apache License, Version 2.0 (the "License");
|
||||
# you may not use this file except in compliance with the License.
|
||||
# You may obtain a copy of the License at
|
||||
#
|
||||
# http://www.apache.org/licenses/LICENSE-2.0
|
||||
#
|
||||
# Unless required by applicable law or agreed to in writing, software
|
||||
# distributed under the License is distributed on an "AS IS" BASIS,
|
||||
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
||||
# See the License for the specific language governing permissions and
|
||||
# limitations under the License.
|
||||
|
||||
name: run-librispeech-streaming-zipformer-2023-05-18
|
||||
# zipformer
|
||||
|
||||
on:
|
||||
push:
|
||||
branches:
|
||||
- master
|
||||
pull_request:
|
||||
types: [labeled]
|
||||
|
||||
schedule:
|
||||
# minute (0-59)
|
||||
# hour (0-23)
|
||||
# day of the month (1-31)
|
||||
# month (1-12)
|
||||
# day of the week (0-6)
|
||||
# nightly build at 15:50 UTC time every day
|
||||
- cron: "50 15 * * *"
|
||||
|
||||
concurrency:
|
||||
group: run_librispeech_2023_05_18_streaming_zipformer-${{ github.ref }}
|
||||
cancel-in-progress: true
|
||||
|
||||
jobs:
|
||||
run_librispeech_2023_05_18_streaming_zipformer:
|
||||
if: github.event.label.name == 'zipformer' ||github.event.label.name == 'ready' || github.event.label.name == 'run-decode' || github.event_name == 'push' || github.event_name == 'schedule'
|
||||
runs-on: ${{ matrix.os }}
|
||||
strategy:
|
||||
matrix:
|
||||
os: [ubuntu-latest]
|
||||
python-version: [3.8]
|
||||
|
||||
fail-fast: false
|
||||
|
||||
steps:
|
||||
- uses: actions/checkout@v2
|
||||
with:
|
||||
fetch-depth: 0
|
||||
|
||||
- name: Setup Python ${{ matrix.python-version }}
|
||||
uses: actions/setup-python@v2
|
||||
with:
|
||||
python-version: ${{ matrix.python-version }}
|
||||
cache: 'pip'
|
||||
cache-dependency-path: '**/requirements-ci.txt'
|
||||
|
||||
- name: Install Python dependencies
|
||||
run: |
|
||||
grep -v '^#' ./requirements-ci.txt | xargs -n 1 -L 1 pip install
|
||||
pip uninstall -y protobuf
|
||||
pip install --no-binary protobuf protobuf==3.20.*
|
||||
|
||||
- name: Cache kaldifeat
|
||||
id: my-cache
|
||||
uses: actions/cache@v2
|
||||
with:
|
||||
path: |
|
||||
~/tmp/kaldifeat
|
||||
key: cache-tmp-${{ matrix.python-version }}-2023-05-22
|
||||
|
||||
- name: Install kaldifeat
|
||||
if: steps.my-cache.outputs.cache-hit != 'true'
|
||||
shell: bash
|
||||
run: |
|
||||
.github/scripts/install-kaldifeat.sh
|
||||
|
||||
- name: Cache LibriSpeech test-clean and test-other datasets
|
||||
id: libri-test-clean-and-test-other-data
|
||||
uses: actions/cache@v2
|
||||
with:
|
||||
path: |
|
||||
~/tmp/download
|
||||
key: cache-libri-test-clean-and-test-other
|
||||
|
||||
- name: Download LibriSpeech test-clean and test-other
|
||||
if: steps.libri-test-clean-and-test-other-data.outputs.cache-hit != 'true'
|
||||
shell: bash
|
||||
run: |
|
||||
.github/scripts/download-librispeech-test-clean-and-test-other-dataset.sh
|
||||
|
||||
- name: Prepare manifests for LibriSpeech test-clean and test-other
|
||||
shell: bash
|
||||
run: |
|
||||
.github/scripts/prepare-librispeech-test-clean-and-test-other-manifests.sh
|
||||
|
||||
- name: Cache LibriSpeech test-clean and test-other fbank features
|
||||
id: libri-test-clean-and-test-other-fbank
|
||||
uses: actions/cache@v2
|
||||
with:
|
||||
path: |
|
||||
~/tmp/fbank-libri
|
||||
key: cache-libri-fbank-test-clean-and-test-other-v2
|
||||
|
||||
- name: Compute fbank for LibriSpeech test-clean and test-other
|
||||
if: steps.libri-test-clean-and-test-other-fbank.outputs.cache-hit != 'true'
|
||||
shell: bash
|
||||
run: |
|
||||
.github/scripts/compute-fbank-librispeech-test-clean-and-test-other.sh
|
||||
|
||||
- name: Inference with pre-trained model
|
||||
shell: bash
|
||||
env:
|
||||
GITHUB_EVENT_NAME: ${{ github.event_name }}
|
||||
GITHUB_EVENT_LABEL_NAME: ${{ github.event.label.name }}
|
||||
run: |
|
||||
mkdir -p egs/librispeech/ASR/data
|
||||
ln -sfv ~/tmp/fbank-libri egs/librispeech/ASR/data/fbank
|
||||
ls -lh egs/librispeech/ASR/data/*
|
||||
|
||||
sudo apt-get -qq install git-lfs tree
|
||||
export PYTHONPATH=$PWD:$PYTHONPATH
|
||||
export PYTHONPATH=~/tmp/kaldifeat/kaldifeat/python:$PYTHONPATH
|
||||
export PYTHONPATH=~/tmp/kaldifeat/build/lib:$PYTHONPATH
|
||||
|
||||
.github/scripts/run-librispeech-streaming-zipformer-2023-05-18.sh
|
||||
|
||||
- name: Display decoding results for librispeech zipformer
|
||||
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
|
||||
shell: bash
|
||||
run: |
|
||||
cd egs/librispeech/ASR/
|
||||
tree ./zipformer/exp
|
||||
|
||||
cd zipformer
|
||||
|
||||
echo "results for zipformer, simulated streaming decoding"
|
||||
echo "===greedy search==="
|
||||
find exp/greedy_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
|
||||
find exp/greedy_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
|
||||
|
||||
echo "===fast_beam_search==="
|
||||
find exp/fast_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
|
||||
find exp/fast_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
|
||||
|
||||
echo "===modified beam search==="
|
||||
find exp/modified_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
|
||||
find exp/modified_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
|
||||
|
||||
echo "results for zipformer, chunk-wise streaming decoding"
|
||||
echo "===greedy search==="
|
||||
find exp/streaming/greedy_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
|
||||
find exp/streaming/greedy_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
|
||||
|
||||
echo "===fast_beam_search==="
|
||||
find exp/streaming/fast_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
|
||||
find exp/streaming/fast_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
|
||||
|
||||
echo "===modified beam search==="
|
||||
find exp/streaming/modified_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
|
||||
find exp/streaming/modified_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
|
||||
|
||||
|
||||
- name: Upload decoding results for librispeech zipformer
|
||||
uses: actions/upload-artifact@v2
|
||||
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
|
||||
with:
|
||||
name: torch-${{ matrix.torch }}-python-${{ matrix.python-version }}-ubuntu-latest-cpu-zipformer-2022-11-11
|
||||
path: egs/librispeech/ASR/zipformer/exp/
|
@ -1,159 +0,0 @@
|
||||
# Copyright 2021 Fangjun Kuang (csukuangfj@gmail.com)
|
||||
|
||||
# See ../../LICENSE for clarification regarding multiple authors
|
||||
#
|
||||
# Licensed under the Apache License, Version 2.0 (the "License");
|
||||
# you may not use this file except in compliance with the License.
|
||||
# You may obtain a copy of the License at
|
||||
#
|
||||
# http://www.apache.org/licenses/LICENSE-2.0
|
||||
#
|
||||
# Unless required by applicable law or agreed to in writing, software
|
||||
# distributed under the License is distributed on an "AS IS" BASIS,
|
||||
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
||||
# See the License for the specific language governing permissions and
|
||||
# limitations under the License.
|
||||
|
||||
name: run-librispeech-2022-04-19
|
||||
# stateless transducer + torchaudio rnn-t loss
|
||||
|
||||
on:
|
||||
push:
|
||||
branches:
|
||||
- master
|
||||
pull_request:
|
||||
types: [labeled]
|
||||
|
||||
schedule:
|
||||
# minute (0-59)
|
||||
# hour (0-23)
|
||||
# day of the month (1-31)
|
||||
# month (1-12)
|
||||
# day of the week (0-6)
|
||||
# nightly build at 15:50 UTC time every day
|
||||
- cron: "50 15 * * *"
|
||||
|
||||
concurrency:
|
||||
group: run_librispeech_2022_04_19-${{ github.ref }}
|
||||
cancel-in-progress: true
|
||||
|
||||
jobs:
|
||||
run_librispeech_2022_04_19:
|
||||
if: github.event.label.name == 'ready' || github.event.label.name == 'run-decode' || github.event_name == 'push' || github.event_name == 'schedule'
|
||||
runs-on: ${{ matrix.os }}
|
||||
strategy:
|
||||
matrix:
|
||||
os: [ubuntu-latest]
|
||||
python-version: [3.8]
|
||||
|
||||
fail-fast: false
|
||||
|
||||
steps:
|
||||
- uses: actions/checkout@v2
|
||||
with:
|
||||
fetch-depth: 0
|
||||
|
||||
- name: Setup Python ${{ matrix.python-version }}
|
||||
uses: actions/setup-python@v2
|
||||
with:
|
||||
python-version: ${{ matrix.python-version }}
|
||||
cache: 'pip'
|
||||
cache-dependency-path: '**/requirements-ci.txt'
|
||||
|
||||
- name: Install Python dependencies
|
||||
run: |
|
||||
grep -v '^#' ./requirements-ci.txt | xargs -n 1 -L 1 pip install
|
||||
pip uninstall -y protobuf
|
||||
pip install --no-binary protobuf protobuf==3.20.*
|
||||
|
||||
- name: Cache kaldifeat
|
||||
id: my-cache
|
||||
uses: actions/cache@v2
|
||||
with:
|
||||
path: |
|
||||
~/tmp/kaldifeat
|
||||
key: cache-tmp-${{ matrix.python-version }}-2023-05-22
|
||||
|
||||
- name: Install kaldifeat
|
||||
if: steps.my-cache.outputs.cache-hit != 'true'
|
||||
shell: bash
|
||||
run: |
|
||||
.github/scripts/install-kaldifeat.sh
|
||||
|
||||
- name: Cache LibriSpeech test-clean and test-other datasets
|
||||
id: libri-test-clean-and-test-other-data
|
||||
uses: actions/cache@v2
|
||||
with:
|
||||
path: |
|
||||
~/tmp/download
|
||||
key: cache-libri-test-clean-and-test-other
|
||||
|
||||
- name: Download LibriSpeech test-clean and test-other
|
||||
if: steps.libri-test-clean-and-test-other-data.outputs.cache-hit != 'true'
|
||||
shell: bash
|
||||
run: |
|
||||
.github/scripts/download-librispeech-test-clean-and-test-other-dataset.sh
|
||||
|
||||
- name: Prepare manifests for LibriSpeech test-clean and test-other
|
||||
shell: bash
|
||||
run: |
|
||||
.github/scripts/prepare-librispeech-test-clean-and-test-other-manifests.sh
|
||||
|
||||
- name: Cache LibriSpeech test-clean and test-other fbank features
|
||||
id: libri-test-clean-and-test-other-fbank
|
||||
uses: actions/cache@v2
|
||||
with:
|
||||
path: |
|
||||
~/tmp/fbank-libri
|
||||
key: cache-libri-fbank-test-clean-and-test-other-v2
|
||||
|
||||
- name: Compute fbank for LibriSpeech test-clean and test-other
|
||||
if: steps.libri-test-clean-and-test-other-fbank.outputs.cache-hit != 'true'
|
||||
shell: bash
|
||||
run: |
|
||||
.github/scripts/compute-fbank-librispeech-test-clean-and-test-other.sh
|
||||
|
||||
- name: Inference with pre-trained model
|
||||
shell: bash
|
||||
env:
|
||||
GITHUB_EVENT_NAME: ${{ github.event_name }}
|
||||
GITHUB_EVENT_LABEL_NAME: ${{ github.event.label.name }}
|
||||
run: |
|
||||
mkdir -p egs/librispeech/ASR/data
|
||||
ln -sfv ~/tmp/fbank-libri egs/librispeech/ASR/data/fbank
|
||||
ls -lh egs/librispeech/ASR/data/*
|
||||
|
||||
sudo apt-get -qq install git-lfs tree
|
||||
export PYTHONPATH=$PWD:$PYTHONPATH
|
||||
export PYTHONPATH=~/tmp/kaldifeat/kaldifeat/python:$PYTHONPATH
|
||||
export PYTHONPATH=~/tmp/kaldifeat/build/lib:$PYTHONPATH
|
||||
|
||||
.github/scripts/run-librispeech-transducer-stateless2-2022-04-19.sh
|
||||
|
||||
- name: Display decoding results
|
||||
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
|
||||
shell: bash
|
||||
run: |
|
||||
cd egs/librispeech/ASR/
|
||||
tree ./transducer_stateless2/exp
|
||||
|
||||
cd transducer_stateless2
|
||||
echo "results for transducer_stateless2"
|
||||
echo "===greedy search==="
|
||||
find exp/greedy_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
|
||||
find exp/greedy_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
|
||||
|
||||
echo "===fast_beam_search==="
|
||||
find exp/fast_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
|
||||
find exp/fast_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
|
||||
|
||||
echo "===modified_beam_search==="
|
||||
find exp/modified_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
|
||||
find exp/modified_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
|
||||
|
||||
- name: Upload decoding results for transducer_stateless2
|
||||
uses: actions/upload-artifact@v2
|
||||
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
|
||||
with:
|
||||
name: torch-${{ matrix.torch }}-python-${{ matrix.python-version }}-ubuntu-latest-cpu-transducer_stateless2-2022-04-19
|
||||
path: egs/librispeech/ASR/transducer_stateless2/exp/
|
@ -1,159 +0,0 @@
|
||||
# Copyright 2022 Fangjun Kuang (csukuangfj@gmail.com)
|
||||
|
||||
# See ../../LICENSE for clarification regarding multiple authors
|
||||
#
|
||||
# Licensed under the Apache License, Version 2.0 (the "License");
|
||||
# you may not use this file except in compliance with the License.
|
||||
# You may obtain a copy of the License at
|
||||
#
|
||||
# http://www.apache.org/licenses/LICENSE-2.0
|
||||
#
|
||||
# Unless required by applicable law or agreed to in writing, software
|
||||
# distributed under the License is distributed on an "AS IS" BASIS,
|
||||
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
||||
# See the License for the specific language governing permissions and
|
||||
# limitations under the License.
|
||||
|
||||
name: run-librispeech-zipformer-2023-05-18
|
||||
# zipformer
|
||||
|
||||
on:
|
||||
push:
|
||||
branches:
|
||||
- master
|
||||
pull_request:
|
||||
types: [labeled]
|
||||
|
||||
schedule:
|
||||
# minute (0-59)
|
||||
# hour (0-23)
|
||||
# day of the month (1-31)
|
||||
# month (1-12)
|
||||
# day of the week (0-6)
|
||||
# nightly build at 15:50 UTC time every day
|
||||
- cron: "50 15 * * *"
|
||||
|
||||
concurrency:
|
||||
group: run_librispeech_2023_05_18_zipformer-${{ github.ref }}
|
||||
cancel-in-progress: true
|
||||
|
||||
jobs:
|
||||
run_librispeech_2023_05_18_zipformer:
|
||||
if: github.event.label.name == 'zipformer' ||github.event.label.name == 'ready' || github.event.label.name == 'run-decode' || github.event_name == 'push' || github.event_name == 'schedule'
|
||||
runs-on: ${{ matrix.os }}
|
||||
strategy:
|
||||
matrix:
|
||||
os: [ubuntu-latest]
|
||||
python-version: [3.8]
|
||||
|
||||
fail-fast: false
|
||||
|
||||
steps:
|
||||
- uses: actions/checkout@v2
|
||||
with:
|
||||
fetch-depth: 0
|
||||
|
||||
- name: Setup Python ${{ matrix.python-version }}
|
||||
uses: actions/setup-python@v2
|
||||
with:
|
||||
python-version: ${{ matrix.python-version }}
|
||||
cache: 'pip'
|
||||
cache-dependency-path: '**/requirements-ci.txt'
|
||||
|
||||
- name: Install Python dependencies
|
||||
run: |
|
||||
grep -v '^#' ./requirements-ci.txt | xargs -n 1 -L 1 pip install
|
||||
pip uninstall -y protobuf
|
||||
pip install --no-binary protobuf protobuf==3.20.*
|
||||
|
||||
- name: Cache kaldifeat
|
||||
id: my-cache
|
||||
uses: actions/cache@v2
|
||||
with:
|
||||
path: |
|
||||
~/tmp/kaldifeat
|
||||
key: cache-tmp-${{ matrix.python-version }}-2023-05-22
|
||||
|
||||
- name: Install kaldifeat
|
||||
if: steps.my-cache.outputs.cache-hit != 'true'
|
||||
shell: bash
|
||||
run: |
|
||||
.github/scripts/install-kaldifeat.sh
|
||||
|
||||
- name: Cache LibriSpeech test-clean and test-other datasets
|
||||
id: libri-test-clean-and-test-other-data
|
||||
uses: actions/cache@v2
|
||||
with:
|
||||
path: |
|
||||
~/tmp/download
|
||||
key: cache-libri-test-clean-and-test-other
|
||||
|
||||
- name: Download LibriSpeech test-clean and test-other
|
||||
if: steps.libri-test-clean-and-test-other-data.outputs.cache-hit != 'true'
|
||||
shell: bash
|
||||
run: |
|
||||
.github/scripts/download-librispeech-test-clean-and-test-other-dataset.sh
|
||||
|
||||
- name: Prepare manifests for LibriSpeech test-clean and test-other
|
||||
shell: bash
|
||||
run: |
|
||||
.github/scripts/prepare-librispeech-test-clean-and-test-other-manifests.sh
|
||||
|
||||
- name: Cache LibriSpeech test-clean and test-other fbank features
|
||||
id: libri-test-clean-and-test-other-fbank
|
||||
uses: actions/cache@v2
|
||||
with:
|
||||
path: |
|
||||
~/tmp/fbank-libri
|
||||
key: cache-libri-fbank-test-clean-and-test-other-v2
|
||||
|
||||
- name: Compute fbank for LibriSpeech test-clean and test-other
|
||||
if: steps.libri-test-clean-and-test-other-fbank.outputs.cache-hit != 'true'
|
||||
shell: bash
|
||||
run: |
|
||||
.github/scripts/compute-fbank-librispeech-test-clean-and-test-other.sh
|
||||
|
||||
- name: Inference with pre-trained model
|
||||
shell: bash
|
||||
env:
|
||||
GITHUB_EVENT_NAME: ${{ github.event_name }}
|
||||
GITHUB_EVENT_LABEL_NAME: ${{ github.event.label.name }}
|
||||
run: |
|
||||
mkdir -p egs/librispeech/ASR/data
|
||||
ln -sfv ~/tmp/fbank-libri egs/librispeech/ASR/data/fbank
|
||||
ls -lh egs/librispeech/ASR/data/*
|
||||
|
||||
sudo apt-get -qq install git-lfs tree
|
||||
export PYTHONPATH=$PWD:$PYTHONPATH
|
||||
export PYTHONPATH=~/tmp/kaldifeat/kaldifeat/python:$PYTHONPATH
|
||||
export PYTHONPATH=~/tmp/kaldifeat/build/lib:$PYTHONPATH
|
||||
|
||||
.github/scripts/run-librispeech-zipformer-2023-05-18.sh
|
||||
|
||||
- name: Display decoding results for librispeech zipformer
|
||||
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
|
||||
shell: bash
|
||||
run: |
|
||||
cd egs/librispeech/ASR/
|
||||
tree ./zipformer/exp
|
||||
|
||||
cd zipformer
|
||||
echo "results for zipformer"
|
||||
echo "===greedy search==="
|
||||
find exp/greedy_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
|
||||
find exp/greedy_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
|
||||
|
||||
echo "===fast_beam_search==="
|
||||
find exp/fast_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
|
||||
find exp/fast_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
|
||||
|
||||
echo "===modified beam search==="
|
||||
find exp/modified_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
|
||||
find exp/modified_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
|
||||
|
||||
- name: Upload decoding results for librispeech zipformer
|
||||
uses: actions/upload-artifact@v2
|
||||
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
|
||||
with:
|
||||
name: torch-${{ matrix.torch }}-python-${{ matrix.python-version }}-ubuntu-latest-cpu-zipformer-2022-11-11
|
||||
path: egs/librispeech/ASR/zipformer/exp/
|
@ -1,155 +0,0 @@
|
||||
# Copyright 2022 Fangjun Kuang (csukuangfj@gmail.com)
|
||||
|
||||
# See ../../LICENSE for clarification regarding multiple authors
|
||||
#
|
||||
# Licensed under the Apache License, Version 2.0 (the "License");
|
||||
# you may not use this file except in compliance with the License.
|
||||
# You may obtain a copy of the License at
|
||||
#
|
||||
# http://www.apache.org/licenses/LICENSE-2.0
|
||||
#
|
||||
# Unless required by applicable law or agreed to in writing, software
|
||||
# distributed under the License is distributed on an "AS IS" BASIS,
|
||||
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
||||
# See the License for the specific language governing permissions and
|
||||
# limitations under the License.
|
||||
|
||||
name: run-librispeech-zipformer-ctc-2023-06-14
|
||||
# zipformer
|
||||
|
||||
on:
|
||||
push:
|
||||
branches:
|
||||
- master
|
||||
pull_request:
|
||||
types: [labeled]
|
||||
|
||||
schedule:
|
||||
# minute (0-59)
|
||||
# hour (0-23)
|
||||
# day of the month (1-31)
|
||||
# month (1-12)
|
||||
# day of the week (0-6)
|
||||
# nightly build at 15:50 UTC time every day
|
||||
- cron: "50 15 * * *"
|
||||
|
||||
concurrency:
|
||||
group: run_librispeech_2023_06_14_zipformer-ctc-${{ github.ref }}
|
||||
cancel-in-progress: true
|
||||
|
||||
jobs:
|
||||
run_librispeech_2023_06_14_zipformer_ctc:
|
||||
if: github.event.label.name == 'zipformer' ||github.event.label.name == 'ready' || github.event.label.name == 'run-decode' || github.event_name == 'push' || github.event_name == 'schedule'
|
||||
runs-on: ${{ matrix.os }}
|
||||
strategy:
|
||||
matrix:
|
||||
os: [ubuntu-latest]
|
||||
python-version: [3.8]
|
||||
|
||||
fail-fast: false
|
||||
|
||||
steps:
|
||||
- uses: actions/checkout@v2
|
||||
with:
|
||||
fetch-depth: 0
|
||||
|
||||
- name: Setup Python ${{ matrix.python-version }}
|
||||
uses: actions/setup-python@v2
|
||||
with:
|
||||
python-version: ${{ matrix.python-version }}
|
||||
cache: 'pip'
|
||||
cache-dependency-path: '**/requirements-ci.txt'
|
||||
|
||||
- name: Install Python dependencies
|
||||
run: |
|
||||
grep -v '^#' ./requirements-ci.txt | xargs -n 1 -L 1 pip install
|
||||
pip uninstall -y protobuf
|
||||
pip install --no-binary protobuf protobuf==3.20.*
|
||||
|
||||
- name: Cache kaldifeat
|
||||
id: my-cache
|
||||
uses: actions/cache@v2
|
||||
with:
|
||||
path: |
|
||||
~/tmp/kaldifeat
|
||||
key: cache-tmp-${{ matrix.python-version }}-2023-05-22
|
||||
|
||||
- name: Install kaldifeat
|
||||
if: steps.my-cache.outputs.cache-hit != 'true'
|
||||
shell: bash
|
||||
run: |
|
||||
.github/scripts/install-kaldifeat.sh
|
||||
|
||||
- name: Cache LibriSpeech test-clean and test-other datasets
|
||||
id: libri-test-clean-and-test-other-data
|
||||
uses: actions/cache@v2
|
||||
with:
|
||||
path: |
|
||||
~/tmp/download
|
||||
key: cache-libri-test-clean-and-test-other
|
||||
|
||||
- name: Download LibriSpeech test-clean and test-other
|
||||
if: steps.libri-test-clean-and-test-other-data.outputs.cache-hit != 'true'
|
||||
shell: bash
|
||||
run: |
|
||||
.github/scripts/download-librispeech-test-clean-and-test-other-dataset.sh
|
||||
|
||||
- name: Prepare manifests for LibriSpeech test-clean and test-other
|
||||
shell: bash
|
||||
run: |
|
||||
.github/scripts/prepare-librispeech-test-clean-and-test-other-manifests.sh
|
||||
|
||||
- name: Cache LibriSpeech test-clean and test-other fbank features
|
||||
id: libri-test-clean-and-test-other-fbank
|
||||
uses: actions/cache@v2
|
||||
with:
|
||||
path: |
|
||||
~/tmp/fbank-libri
|
||||
key: cache-libri-fbank-test-clean-and-test-other-v2
|
||||
|
||||
- name: Compute fbank for LibriSpeech test-clean and test-other
|
||||
if: steps.libri-test-clean-and-test-other-fbank.outputs.cache-hit != 'true'
|
||||
shell: bash
|
||||
run: |
|
||||
.github/scripts/compute-fbank-librispeech-test-clean-and-test-other.sh
|
||||
|
||||
- name: Inference with pre-trained model
|
||||
shell: bash
|
||||
env:
|
||||
GITHUB_EVENT_NAME: ${{ github.event_name }}
|
||||
GITHUB_EVENT_LABEL_NAME: ${{ github.event.label.name }}
|
||||
run: |
|
||||
mkdir -p egs/librispeech/ASR/data
|
||||
ln -sfv ~/tmp/fbank-libri egs/librispeech/ASR/data/fbank
|
||||
ls -lh egs/librispeech/ASR/data/*
|
||||
|
||||
sudo apt-get -qq install git-lfs tree
|
||||
export PYTHONPATH=$PWD:$PYTHONPATH
|
||||
export PYTHONPATH=~/tmp/kaldifeat/kaldifeat/python:$PYTHONPATH
|
||||
export PYTHONPATH=~/tmp/kaldifeat/build/lib:$PYTHONPATH
|
||||
|
||||
.github/scripts/run-librispeech-zipformer-ctc-2023-06-14.sh
|
||||
|
||||
- name: Display decoding results for librispeech zipformer
|
||||
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
|
||||
shell: bash
|
||||
run: |
|
||||
cd egs/librispeech/ASR/
|
||||
tree ./zipformer/exp
|
||||
|
||||
cd zipformer
|
||||
echo "results for zipformer"
|
||||
echo "===ctc-decoding==="
|
||||
find exp/ctc-decoding -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
|
||||
find exp/ctc-decoding -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
|
||||
|
||||
echo "===1best==="
|
||||
find exp/1best -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
|
||||
find exp/1best -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
|
||||
|
||||
- name: Upload decoding results for librispeech zipformer
|
||||
uses: actions/upload-artifact@v2
|
||||
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
|
||||
with:
|
||||
name: torch-${{ matrix.torch }}-python-${{ matrix.python-version }}-ubuntu-latest-cpu-zipformer-2022-11-11
|
||||
path: egs/librispeech/ASR/zipformer/exp/
|
87
.github/workflows/run-pretrained-ctc.yml
vendored
87
.github/workflows/run-pretrained-ctc.yml
vendored
@ -1,87 +0,0 @@
|
||||
# Copyright 2021 Fangjun Kuang (csukuangfj@gmail.com)
|
||||
|
||||
# See ../../LICENSE for clarification regarding multiple authors
|
||||
#
|
||||
# Licensed under the Apache License, Version 2.0 (the "License");
|
||||
# you may not use this file except in compliance with the License.
|
||||
# You may obtain a copy of the License at
|
||||
#
|
||||
# http://www.apache.org/licenses/LICENSE-2.0
|
||||
#
|
||||
# Unless required by applicable law or agreed to in writing, software
|
||||
# distributed under the License is distributed on an "AS IS" BASIS,
|
||||
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
||||
# See the License for the specific language governing permissions and
|
||||
# limitations under the License.
|
||||
|
||||
name: run-pre-trained-ctc
|
||||
|
||||
on:
|
||||
push:
|
||||
branches:
|
||||
- master
|
||||
pull_request:
|
||||
types: [labeled]
|
||||
|
||||
workflow_dispatch:
|
||||
inputs:
|
||||
test-run:
|
||||
description: 'Test (y/n)?'
|
||||
required: true
|
||||
default: 'y'
|
||||
|
||||
concurrency:
|
||||
group: run_pre_trained_ctc-${{ github.ref }}
|
||||
cancel-in-progress: true
|
||||
|
||||
jobs:
|
||||
run_pre_trained_ctc:
|
||||
if: github.event.label.name == 'ready' || github.event_name == 'push' || github.event.inputs.test-run == 'y' || github.event.label.name == 'ctc'
|
||||
runs-on: ${{ matrix.os }}
|
||||
strategy:
|
||||
matrix:
|
||||
os: [ubuntu-latest]
|
||||
python-version: [3.8]
|
||||
|
||||
fail-fast: false
|
||||
|
||||
steps:
|
||||
- uses: actions/checkout@v2
|
||||
with:
|
||||
fetch-depth: 0
|
||||
|
||||
- name: Setup Python ${{ matrix.python-version }}
|
||||
uses: actions/setup-python@v2
|
||||
with:
|
||||
python-version: ${{ matrix.python-version }}
|
||||
cache: 'pip'
|
||||
cache-dependency-path: '**/requirements-ci.txt'
|
||||
|
||||
- name: Install Python dependencies
|
||||
run: |
|
||||
grep -v '^#' ./requirements-ci.txt | xargs -n 1 -L 1 pip install
|
||||
pip uninstall -y protobuf
|
||||
pip install --no-binary protobuf protobuf==3.20.*
|
||||
|
||||
- name: Cache kaldifeat
|
||||
id: my-cache
|
||||
uses: actions/cache@v2
|
||||
with:
|
||||
path: |
|
||||
~/tmp/kaldifeat
|
||||
key: cache-tmp-${{ matrix.python-version }}-2023-05-22
|
||||
|
||||
- name: Install kaldifeat
|
||||
if: steps.my-cache.outputs.cache-hit != 'true'
|
||||
shell: bash
|
||||
run: |
|
||||
.github/scripts/install-kaldifeat.sh
|
||||
|
||||
- name: Inference with pre-trained model
|
||||
shell: bash
|
||||
run: |
|
||||
sudo apt-get -qq install git-lfs tree
|
||||
export PYTHONPATH=$PWD:$PYTHONPATH
|
||||
export PYTHONPATH=~/tmp/kaldifeat/kaldifeat/python:$PYTHONPATH
|
||||
export PYTHONPATH=~/tmp/kaldifeat/build/lib:$PYTHONPATH
|
||||
.github/scripts/run-pre-trained-ctc.sh
|
@ -1,158 +0,0 @@
|
||||
# Copyright 2021 Fangjun Kuang (csukuangfj@gmail.com)
|
||||
|
||||
# See ../../LICENSE for clarification regarding multiple authors
|
||||
#
|
||||
# Licensed under the Apache License, Version 2.0 (the "License");
|
||||
# you may not use this file except in compliance with the License.
|
||||
# You may obtain a copy of the License at
|
||||
#
|
||||
# http://www.apache.org/licenses/LICENSE-2.0
|
||||
#
|
||||
# Unless required by applicable law or agreed to in writing, software
|
||||
# distributed under the License is distributed on an "AS IS" BASIS,
|
||||
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
||||
# See the License for the specific language governing permissions and
|
||||
# limitations under the License.
|
||||
|
||||
name: run-pre-trained-trandsucer-stateless-multi-datasets-librispeech-100h
|
||||
|
||||
on:
|
||||
push:
|
||||
branches:
|
||||
- master
|
||||
pull_request:
|
||||
types: [labeled]
|
||||
|
||||
schedule:
|
||||
# minute (0-59)
|
||||
# hour (0-23)
|
||||
# day of the month (1-31)
|
||||
# month (1-12)
|
||||
# day of the week (0-6)
|
||||
# nightly build at 15:50 UTC time every day
|
||||
- cron: "50 15 * * *"
|
||||
|
||||
concurrency:
|
||||
group: run_pre_trained_transducer_stateless_multi_datasets_librispeech_100h-${{ github.ref }}
|
||||
cancel-in-progress: true
|
||||
|
||||
jobs:
|
||||
run_pre_trained_transducer_stateless_multi_datasets_librispeech_100h:
|
||||
if: github.event.label.name == 'ready' || github.event.label.name == 'run-decode' || github.event_name == 'push' || github.event_name == 'schedule'
|
||||
runs-on: ${{ matrix.os }}
|
||||
strategy:
|
||||
matrix:
|
||||
os: [ubuntu-latest]
|
||||
python-version: [3.8]
|
||||
|
||||
fail-fast: false
|
||||
|
||||
steps:
|
||||
- uses: actions/checkout@v2
|
||||
with:
|
||||
fetch-depth: 0
|
||||
|
||||
- name: Setup Python ${{ matrix.python-version }}
|
||||
uses: actions/setup-python@v2
|
||||
with:
|
||||
python-version: ${{ matrix.python-version }}
|
||||
cache: 'pip'
|
||||
cache-dependency-path: '**/requirements-ci.txt'
|
||||
|
||||
- name: Install Python dependencies
|
||||
run: |
|
||||
grep -v '^#' ./requirements-ci.txt | xargs -n 1 -L 1 pip install
|
||||
pip uninstall -y protobuf
|
||||
pip install --no-binary protobuf protobuf==3.20.*
|
||||
|
||||
- name: Cache kaldifeat
|
||||
id: my-cache
|
||||
uses: actions/cache@v2
|
||||
with:
|
||||
path: |
|
||||
~/tmp/kaldifeat
|
||||
key: cache-tmp-${{ matrix.python-version }}-2023-05-22
|
||||
|
||||
- name: Install kaldifeat
|
||||
if: steps.my-cache.outputs.cache-hit != 'true'
|
||||
shell: bash
|
||||
run: |
|
||||
.github/scripts/install-kaldifeat.sh
|
||||
|
||||
- name: Cache LibriSpeech test-clean and test-other datasets
|
||||
id: libri-test-clean-and-test-other-data
|
||||
uses: actions/cache@v2
|
||||
with:
|
||||
path: |
|
||||
~/tmp/download
|
||||
key: cache-libri-test-clean-and-test-other
|
||||
|
||||
- name: Download LibriSpeech test-clean and test-other
|
||||
if: steps.libri-test-clean-and-test-other-data.outputs.cache-hit != 'true'
|
||||
shell: bash
|
||||
run: |
|
||||
.github/scripts/download-librispeech-test-clean-and-test-other-dataset.sh
|
||||
|
||||
- name: Prepare manifests for LibriSpeech test-clean and test-other
|
||||
shell: bash
|
||||
run: |
|
||||
.github/scripts/prepare-librispeech-test-clean-and-test-other-manifests.sh
|
||||
|
||||
- name: Cache LibriSpeech test-clean and test-other fbank features
|
||||
id: libri-test-clean-and-test-other-fbank
|
||||
uses: actions/cache@v2
|
||||
with:
|
||||
path: |
|
||||
~/tmp/fbank-libri
|
||||
key: cache-libri-fbank-test-clean-and-test-other-v2
|
||||
|
||||
- name: Compute fbank for LibriSpeech test-clean and test-other
|
||||
if: steps.libri-test-clean-and-test-other-fbank.outputs.cache-hit != 'true'
|
||||
shell: bash
|
||||
run: |
|
||||
.github/scripts/compute-fbank-librispeech-test-clean-and-test-other.sh
|
||||
|
||||
- name: Inference with pre-trained model
|
||||
shell: bash
|
||||
env:
|
||||
GITHUB_EVENT_NAME: ${{ github.event_name }}
|
||||
GITHUB_EVENT_LABEL_NAME: ${{ github.event.label.name }}
|
||||
run: |
|
||||
mkdir -p egs/librispeech/ASR/data
|
||||
ln -sfv ~/tmp/fbank-libri egs/librispeech/ASR/data/fbank
|
||||
ls -lh egs/librispeech/ASR/data/*
|
||||
|
||||
sudo apt-get -qq install git-lfs tree
|
||||
export PYTHONPATH=$PWD:$PYTHONPATH
|
||||
export PYTHONPATH=~/tmp/kaldifeat/kaldifeat/python:$PYTHONPATH
|
||||
export PYTHONPATH=~/tmp/kaldifeat/build/lib:$PYTHONPATH
|
||||
|
||||
.github/scripts/run-pre-trained-transducer-stateless-librispeech-100h.sh
|
||||
|
||||
- name: Display decoding results for transducer_stateless_multi_datasets
|
||||
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
|
||||
shell: bash
|
||||
run: |
|
||||
cd egs/librispeech/ASR/
|
||||
tree ./transducer_stateless_multi_datasets/exp
|
||||
|
||||
cd transducer_stateless_multi_datasets
|
||||
echo "results for transducer_stateless_multi_datasets"
|
||||
echo "===greedy search==="
|
||||
find exp/greedy_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
|
||||
find exp/greedy_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
|
||||
|
||||
echo "===fast_beam_search==="
|
||||
find exp/fast_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
|
||||
find exp/fast_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
|
||||
|
||||
echo "===modified beam search==="
|
||||
find exp/modified_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
|
||||
find exp/modified_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
|
||||
|
||||
- name: Upload decoding results for transducer_stateless_multi_datasets
|
||||
uses: actions/upload-artifact@v2
|
||||
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
|
||||
with:
|
||||
name: torch-${{ matrix.torch }}-python-${{ matrix.python-version }}-ubuntu-latest-cpu-transducer_stateless_multi_datasets-100h-2022-02-21
|
||||
path: egs/librispeech/ASR/transducer_stateless_multi_datasets/exp/
|
@ -1,158 +0,0 @@
|
||||
# Copyright 2021 Fangjun Kuang (csukuangfj@gmail.com)
|
||||
|
||||
# See ../../LICENSE for clarification regarding multiple authors
|
||||
#
|
||||
# Licensed under the Apache License, Version 2.0 (the "License");
|
||||
# you may not use this file except in compliance with the License.
|
||||
# You may obtain a copy of the License at
|
||||
#
|
||||
# http://www.apache.org/licenses/LICENSE-2.0
|
||||
#
|
||||
# Unless required by applicable law or agreed to in writing, software
|
||||
# distributed under the License is distributed on an "AS IS" BASIS,
|
||||
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
||||
# See the License for the specific language governing permissions and
|
||||
# limitations under the License.
|
||||
|
||||
name: run-pre-trained-trandsucer-stateless-multi-datasets-librispeech-960h
|
||||
|
||||
on:
|
||||
push:
|
||||
branches:
|
||||
- master
|
||||
pull_request:
|
||||
types: [labeled]
|
||||
|
||||
schedule:
|
||||
# minute (0-59)
|
||||
# hour (0-23)
|
||||
# day of the month (1-31)
|
||||
# month (1-12)
|
||||
# day of the week (0-6)
|
||||
# nightly build at 15:50 UTC time every day
|
||||
- cron: "50 15 * * *"
|
||||
|
||||
concurrency:
|
||||
group: run_pre_trained_transducer_stateless_multi_datasets_librispeech_960h-${{ github.ref }}
|
||||
cancel-in-progress: true
|
||||
|
||||
jobs:
|
||||
run_pre_trained_transducer_stateless_multi_datasets_librispeech_960h:
|
||||
if: github.event.label.name == 'ready' || github.event.label.name == 'run-decode' || github.event_name == 'push' || github.event_name == 'schedule'
|
||||
runs-on: ${{ matrix.os }}
|
||||
strategy:
|
||||
matrix:
|
||||
os: [ubuntu-latest]
|
||||
python-version: [3.8]
|
||||
|
||||
fail-fast: false
|
||||
|
||||
steps:
|
||||
- uses: actions/checkout@v2
|
||||
with:
|
||||
fetch-depth: 0
|
||||
|
||||
- name: Setup Python ${{ matrix.python-version }}
|
||||
uses: actions/setup-python@v2
|
||||
with:
|
||||
python-version: ${{ matrix.python-version }}
|
||||
cache: 'pip'
|
||||
cache-dependency-path: '**/requirements-ci.txt'
|
||||
|
||||
- name: Install Python dependencies
|
||||
run: |
|
||||
grep -v '^#' ./requirements-ci.txt | xargs -n 1 -L 1 pip install
|
||||
pip uninstall -y protobuf
|
||||
pip install --no-binary protobuf protobuf==3.20.*
|
||||
|
||||
- name: Cache kaldifeat
|
||||
id: my-cache
|
||||
uses: actions/cache@v2
|
||||
with:
|
||||
path: |
|
||||
~/tmp/kaldifeat
|
||||
key: cache-tmp-${{ matrix.python-version }}-2023-05-22
|
||||
|
||||
- name: Install kaldifeat
|
||||
if: steps.my-cache.outputs.cache-hit != 'true'
|
||||
shell: bash
|
||||
run: |
|
||||
.github/scripts/install-kaldifeat.sh
|
||||
|
||||
- name: Cache LibriSpeech test-clean and test-other datasets
|
||||
id: libri-test-clean-and-test-other-data
|
||||
uses: actions/cache@v2
|
||||
with:
|
||||
path: |
|
||||
~/tmp/download
|
||||
key: cache-libri-test-clean-and-test-other
|
||||
|
||||
- name: Download LibriSpeech test-clean and test-other
|
||||
if: steps.libri-test-clean-and-test-other-data.outputs.cache-hit != 'true'
|
||||
shell: bash
|
||||
run: |
|
||||
.github/scripts/download-librispeech-test-clean-and-test-other-dataset.sh
|
||||
|
||||
- name: Prepare manifests for LibriSpeech test-clean and test-other
|
||||
shell: bash
|
||||
run: |
|
||||
.github/scripts/prepare-librispeech-test-clean-and-test-other-manifests.sh
|
||||
|
||||
- name: Cache LibriSpeech test-clean and test-other fbank features
|
||||
id: libri-test-clean-and-test-other-fbank
|
||||
uses: actions/cache@v2
|
||||
with:
|
||||
path: |
|
||||
~/tmp/fbank-libri
|
||||
key: cache-libri-fbank-test-clean-and-test-other-v2
|
||||
|
||||
- name: Compute fbank for LibriSpeech test-clean and test-other
|
||||
if: steps.libri-test-clean-and-test-other-fbank.outputs.cache-hit != 'true'
|
||||
shell: bash
|
||||
run: |
|
||||
.github/scripts/compute-fbank-librispeech-test-clean-and-test-other.sh
|
||||
|
||||
- name: Inference with pre-trained model
|
||||
shell: bash
|
||||
env:
|
||||
GITHUB_EVENT_NAME: ${{ github.event_name }}
|
||||
GITHUB_EVENT_LABEL_NAME: ${{ github.event.label.name }}
|
||||
run: |
|
||||
mkdir -p egs/librispeech/ASR/data
|
||||
ln -sfv ~/tmp/fbank-libri egs/librispeech/ASR/data/fbank
|
||||
ls -lh egs/librispeech/ASR/data/*
|
||||
|
||||
sudo apt-get -qq install git-lfs tree
|
||||
export PYTHONPATH=$PWD:$PYTHONPATH
|
||||
export PYTHONPATH=~/tmp/kaldifeat/kaldifeat/python:$PYTHONPATH
|
||||
export PYTHONPATH=~/tmp/kaldifeat/build/lib:$PYTHONPATH
|
||||
|
||||
.github/scripts/run-pre-trained-transducer-stateless-librispeech-960h.sh
|
||||
|
||||
- name: Display decoding results for transducer_stateless_multi_datasets
|
||||
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
|
||||
shell: bash
|
||||
run: |
|
||||
cd egs/librispeech/ASR/
|
||||
tree ./transducer_stateless_multi_datasets/exp
|
||||
|
||||
cd transducer_stateless_multi_datasets
|
||||
echo "results for transducer_stateless_multi_datasets"
|
||||
echo "===greedy search==="
|
||||
find exp/greedy_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
|
||||
find exp/greedy_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
|
||||
|
||||
echo "===fast_beam_search==="
|
||||
find exp/fast_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
|
||||
find exp/fast_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
|
||||
|
||||
echo "===modified beam search==="
|
||||
find exp/modified_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
|
||||
find exp/modified_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
|
||||
|
||||
- name: Upload decoding results for transducer_stateless_multi_datasets
|
||||
uses: actions/upload-artifact@v2
|
||||
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
|
||||
with:
|
||||
name: torch-${{ matrix.torch }}-python-${{ matrix.python-version }}-ubuntu-latest-cpu-transducer_stateless_multi_datasets-100h-2022-03-01
|
||||
path: egs/librispeech/ASR/transducer_stateless_multi_datasets/exp/
|
@ -1,80 +0,0 @@
|
||||
# Copyright 2021 Fangjun Kuang (csukuangfj@gmail.com)
|
||||
|
||||
# See ../../LICENSE for clarification regarding multiple authors
|
||||
#
|
||||
# Licensed under the Apache License, Version 2.0 (the "License");
|
||||
# you may not use this file except in compliance with the License.
|
||||
# You may obtain a copy of the License at
|
||||
#
|
||||
# http://www.apache.org/licenses/LICENSE-2.0
|
||||
#
|
||||
# Unless required by applicable law or agreed to in writing, software
|
||||
# distributed under the License is distributed on an "AS IS" BASIS,
|
||||
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
||||
# See the License for the specific language governing permissions and
|
||||
# limitations under the License.
|
||||
|
||||
name: run-pre-trained-trandsucer-stateless-modified-2-aishell
|
||||
|
||||
on:
|
||||
push:
|
||||
branches:
|
||||
- master
|
||||
pull_request:
|
||||
types: [labeled]
|
||||
|
||||
concurrency:
|
||||
group: run_pre_trained_transducer_stateless_modified_2_aishell-${{ github.ref }}
|
||||
cancel-in-progress: true
|
||||
|
||||
jobs:
|
||||
run_pre_trained_transducer_stateless_modified_2_aishell:
|
||||
if: github.event.label.name == 'ready' || github.event_name == 'push'
|
||||
runs-on: ${{ matrix.os }}
|
||||
strategy:
|
||||
matrix:
|
||||
os: [ubuntu-latest]
|
||||
python-version: [3.8]
|
||||
|
||||
fail-fast: false
|
||||
|
||||
steps:
|
||||
- uses: actions/checkout@v2
|
||||
with:
|
||||
fetch-depth: 0
|
||||
|
||||
- name: Setup Python ${{ matrix.python-version }}
|
||||
uses: actions/setup-python@v2
|
||||
with:
|
||||
python-version: ${{ matrix.python-version }}
|
||||
cache: 'pip'
|
||||
cache-dependency-path: '**/requirements-ci.txt'
|
||||
|
||||
- name: Install Python dependencies
|
||||
run: |
|
||||
grep -v '^#' ./requirements-ci.txt | xargs -n 1 -L 1 pip install
|
||||
pip uninstall -y protobuf
|
||||
pip install --no-binary protobuf protobuf==3.20.*
|
||||
|
||||
- name: Cache kaldifeat
|
||||
id: my-cache
|
||||
uses: actions/cache@v2
|
||||
with:
|
||||
path: |
|
||||
~/tmp/kaldifeat
|
||||
key: cache-tmp-${{ matrix.python-version }}-2023-05-22
|
||||
|
||||
- name: Install kaldifeat
|
||||
if: steps.my-cache.outputs.cache-hit != 'true'
|
||||
shell: bash
|
||||
run: |
|
||||
.github/scripts/install-kaldifeat.sh
|
||||
|
||||
- name: Inference with pre-trained model
|
||||
shell: bash
|
||||
run: |
|
||||
sudo apt-get -qq install git-lfs tree
|
||||
export PYTHONPATH=$PWD:$PYTHONPATH
|
||||
export PYTHONPATH=~/tmp/kaldifeat/kaldifeat/python:$PYTHONPATH
|
||||
export PYTHONPATH=~/tmp/kaldifeat/build/lib:$PYTHONPATH
|
||||
.github/scripts/run-pre-trained-transducer-stateless-modified-2-aishell.sh
|
@ -1,80 +0,0 @@
|
||||
# Copyright 2021 Fangjun Kuang (csukuangfj@gmail.com)
|
||||
|
||||
# See ../../LICENSE for clarification regarding multiple authors
|
||||
#
|
||||
# Licensed under the Apache License, Version 2.0 (the "License");
|
||||
# you may not use this file except in compliance with the License.
|
||||
# You may obtain a copy of the License at
|
||||
#
|
||||
# http://www.apache.org/licenses/LICENSE-2.0
|
||||
#
|
||||
# Unless required by applicable law or agreed to in writing, software
|
||||
# distributed under the License is distributed on an "AS IS" BASIS,
|
||||
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
||||
# See the License for the specific language governing permissions and
|
||||
# limitations under the License.
|
||||
|
||||
name: run-pre-trained-trandsucer-stateless-modified-aishell
|
||||
|
||||
on:
|
||||
push:
|
||||
branches:
|
||||
- master
|
||||
pull_request:
|
||||
types: [labeled]
|
||||
|
||||
concurrency:
|
||||
group: run_pre_trained_transducer_stateless_modified_aishell-${{ github.ref }}
|
||||
cancel-in-progress: true
|
||||
|
||||
jobs:
|
||||
run_pre_trained_transducer_stateless_modified_aishell:
|
||||
if: github.event.label.name == 'ready' || github.event_name == 'push'
|
||||
runs-on: ${{ matrix.os }}
|
||||
strategy:
|
||||
matrix:
|
||||
os: [ubuntu-latest]
|
||||
python-version: [3.8]
|
||||
|
||||
fail-fast: false
|
||||
|
||||
steps:
|
||||
- uses: actions/checkout@v2
|
||||
with:
|
||||
fetch-depth: 0
|
||||
|
||||
- name: Setup Python ${{ matrix.python-version }}
|
||||
uses: actions/setup-python@v2
|
||||
with:
|
||||
python-version: ${{ matrix.python-version }}
|
||||
cache: 'pip'
|
||||
cache-dependency-path: '**/requirements-ci.txt'
|
||||
|
||||
- name: Install Python dependencies
|
||||
run: |
|
||||
grep -v '^#' ./requirements-ci.txt | xargs -n 1 -L 1 pip install
|
||||
pip uninstall -y protobuf
|
||||
pip install --no-binary protobuf protobuf==3.20.*
|
||||
|
||||
- name: Cache kaldifeat
|
||||
id: my-cache
|
||||
uses: actions/cache@v2
|
||||
with:
|
||||
path: |
|
||||
~/tmp/kaldifeat
|
||||
key: cache-tmp-${{ matrix.python-version }}-2023-05-22
|
||||
|
||||
- name: Install kaldifeat
|
||||
if: steps.my-cache.outputs.cache-hit != 'true'
|
||||
shell: bash
|
||||
run: |
|
||||
.github/scripts/install-kaldifeat.sh
|
||||
|
||||
- name: Inference with pre-trained model
|
||||
shell: bash
|
||||
run: |
|
||||
sudo apt-get -qq install git-lfs tree
|
||||
export PYTHONPATH=$PWD:$PYTHONPATH
|
||||
export PYTHONPATH=~/tmp/kaldifeat/kaldifeat/python:$PYTHONPATH
|
||||
export PYTHONPATH=~/tmp/kaldifeat/build/lib:$PYTHONPATH
|
||||
.github/scripts/run-pre-trained-transducer-stateless-modified-aishell.sh
|
@ -1,158 +0,0 @@
|
||||
# Copyright 2021 Fangjun Kuang (csukuangfj@gmail.com)
|
||||
|
||||
# See ../../LICENSE for clarification regarding multiple authors
|
||||
#
|
||||
# Licensed under the Apache License, Version 2.0 (the "License");
|
||||
# you may not use this file except in compliance with the License.
|
||||
# You may obtain a copy of the License at
|
||||
#
|
||||
# http://www.apache.org/licenses/LICENSE-2.0
|
||||
#
|
||||
# Unless required by applicable law or agreed to in writing, software
|
||||
# distributed under the License is distributed on an "AS IS" BASIS,
|
||||
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
||||
# See the License for the specific language governing permissions and
|
||||
# limitations under the License.
|
||||
|
||||
name: run-pre-trained-transducer-stateless
|
||||
|
||||
on:
|
||||
push:
|
||||
branches:
|
||||
- master
|
||||
pull_request:
|
||||
types: [labeled]
|
||||
|
||||
schedule:
|
||||
# minute (0-59)
|
||||
# hour (0-23)
|
||||
# day of the month (1-31)
|
||||
# month (1-12)
|
||||
# day of the week (0-6)
|
||||
# nightly build at 15:50 UTC time every day
|
||||
- cron: "50 15 * * *"
|
||||
|
||||
concurrency:
|
||||
group: run_pre_trained_transducer_stateless-${{ github.ref }}
|
||||
cancel-in-progress: true
|
||||
|
||||
jobs:
|
||||
run_pre_trained_transducer_stateless:
|
||||
if: github.event.label.name == 'ready' || github.event.label.name == 'run-decode' || github.event_name == 'push' || github.event_name == 'schedule'
|
||||
runs-on: ${{ matrix.os }}
|
||||
strategy:
|
||||
matrix:
|
||||
os: [ubuntu-latest]
|
||||
python-version: [3.8]
|
||||
|
||||
fail-fast: false
|
||||
|
||||
steps:
|
||||
- uses: actions/checkout@v2
|
||||
with:
|
||||
fetch-depth: 0
|
||||
|
||||
- name: Setup Python ${{ matrix.python-version }}
|
||||
uses: actions/setup-python@v2
|
||||
with:
|
||||
python-version: ${{ matrix.python-version }}
|
||||
cache: 'pip'
|
||||
cache-dependency-path: '**/requirements-ci.txt'
|
||||
|
||||
- name: Install Python dependencies
|
||||
run: |
|
||||
grep -v '^#' ./requirements-ci.txt | xargs -n 1 -L 1 pip install
|
||||
pip uninstall -y protobuf
|
||||
pip install --no-binary protobuf protobuf==3.20.*
|
||||
|
||||
- name: Cache kaldifeat
|
||||
id: my-cache
|
||||
uses: actions/cache@v2
|
||||
with:
|
||||
path: |
|
||||
~/tmp/kaldifeat
|
||||
key: cache-tmp-${{ matrix.python-version }}-2023-05-22
|
||||
|
||||
- name: Install kaldifeat
|
||||
if: steps.my-cache.outputs.cache-hit != 'true'
|
||||
shell: bash
|
||||
run: |
|
||||
.github/scripts/install-kaldifeat.sh
|
||||
|
||||
- name: Cache LibriSpeech test-clean and test-other datasets
|
||||
id: libri-test-clean-and-test-other-data
|
||||
uses: actions/cache@v2
|
||||
with:
|
||||
path: |
|
||||
~/tmp/download
|
||||
key: cache-libri-test-clean-and-test-other
|
||||
|
||||
- name: Download LibriSpeech test-clean and test-other
|
||||
if: steps.libri-test-clean-and-test-other-data.outputs.cache-hit != 'true'
|
||||
shell: bash
|
||||
run: |
|
||||
.github/scripts/download-librispeech-test-clean-and-test-other-dataset.sh
|
||||
|
||||
- name: Prepare manifests for LibriSpeech test-clean and test-other
|
||||
shell: bash
|
||||
run: |
|
||||
.github/scripts/prepare-librispeech-test-clean-and-test-other-manifests.sh
|
||||
|
||||
- name: Cache LibriSpeech test-clean and test-other fbank features
|
||||
id: libri-test-clean-and-test-other-fbank
|
||||
uses: actions/cache@v2
|
||||
with:
|
||||
path: |
|
||||
~/tmp/fbank-libri
|
||||
key: cache-libri-fbank-test-clean-and-test-other-v2
|
||||
|
||||
- name: Compute fbank for LibriSpeech test-clean and test-other
|
||||
if: steps.libri-test-clean-and-test-other-fbank.outputs.cache-hit != 'true'
|
||||
shell: bash
|
||||
run: |
|
||||
.github/scripts/compute-fbank-librispeech-test-clean-and-test-other.sh
|
||||
|
||||
- name: Inference with pre-trained model
|
||||
shell: bash
|
||||
env:
|
||||
GITHUB_EVENT_NAME: ${{ github.event_name }}
|
||||
GITHUB_EVENT_LABEL_NAME: ${{ github.event.label.name }}
|
||||
run: |
|
||||
mkdir -p egs/librispeech/ASR/data
|
||||
ln -sfv ~/tmp/fbank-libri egs/librispeech/ASR/data/fbank
|
||||
ls -lh egs/librispeech/ASR/data/*
|
||||
|
||||
sudo apt-get -qq install git-lfs tree
|
||||
export PYTHONPATH=$PWD:$PYTHONPATH
|
||||
export PYTHONPATH=~/tmp/kaldifeat/kaldifeat/python:$PYTHONPATH
|
||||
export PYTHONPATH=~/tmp/kaldifeat/build/lib:$PYTHONPATH
|
||||
|
||||
.github/scripts/run-pre-trained-transducer-stateless.sh
|
||||
|
||||
- name: Display decoding results for transducer_stateless
|
||||
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
|
||||
shell: bash
|
||||
run: |
|
||||
cd egs/librispeech/ASR/
|
||||
tree ./transducer_stateless/exp
|
||||
|
||||
cd transducer_stateless
|
||||
echo "results for transducer_stateless"
|
||||
echo "===greedy search==="
|
||||
find exp/greedy_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
|
||||
find exp/greedy_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
|
||||
|
||||
echo "===fast_beam_search==="
|
||||
find exp/fast_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
|
||||
find exp/fast_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
|
||||
|
||||
echo "===modified beam search==="
|
||||
find exp/modified_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
|
||||
find exp/modified_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
|
||||
|
||||
- name: Upload decoding results for transducer_stateless
|
||||
uses: actions/upload-artifact@v2
|
||||
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
|
||||
with:
|
||||
name: torch-${{ matrix.torch }}-python-${{ matrix.python-version }}-ubuntu-latest-cpu-transducer_stateless-2022-02-07
|
||||
path: egs/librispeech/ASR/transducer_stateless/exp/
|
80
.github/workflows/run-pretrained-transducer.yml
vendored
80
.github/workflows/run-pretrained-transducer.yml
vendored
@ -1,80 +0,0 @@
|
||||
# Copyright 2021 Fangjun Kuang (csukuangfj@gmail.com)
|
||||
|
||||
# See ../../LICENSE for clarification regarding multiple authors
|
||||
#
|
||||
# Licensed under the Apache License, Version 2.0 (the "License");
|
||||
# you may not use this file except in compliance with the License.
|
||||
# You may obtain a copy of the License at
|
||||
#
|
||||
# http://www.apache.org/licenses/LICENSE-2.0
|
||||
#
|
||||
# Unless required by applicable law or agreed to in writing, software
|
||||
# distributed under the License is distributed on an "AS IS" BASIS,
|
||||
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
||||
# See the License for the specific language governing permissions and
|
||||
# limitations under the License.
|
||||
|
||||
name: run-pre-trained-transducer
|
||||
|
||||
on:
|
||||
push:
|
||||
branches:
|
||||
- master
|
||||
pull_request:
|
||||
types: [labeled]
|
||||
|
||||
concurrency:
|
||||
group: run_pre_trained_transducer-${{ github.ref }}
|
||||
cancel-in-progress: true
|
||||
|
||||
jobs:
|
||||
run_pre_trained_transducer:
|
||||
if: github.event.label.name == 'ready' || github.event_name == 'push'
|
||||
runs-on: ${{ matrix.os }}
|
||||
strategy:
|
||||
matrix:
|
||||
os: [ubuntu-latest]
|
||||
python-version: [3.8]
|
||||
|
||||
fail-fast: false
|
||||
|
||||
steps:
|
||||
- uses: actions/checkout@v2
|
||||
with:
|
||||
fetch-depth: 0
|
||||
|
||||
- name: Setup Python ${{ matrix.python-version }}
|
||||
uses: actions/setup-python@v2
|
||||
with:
|
||||
python-version: ${{ matrix.python-version }}
|
||||
cache: 'pip'
|
||||
cache-dependency-path: '**/requirements-ci.txt'
|
||||
|
||||
- name: Install Python dependencies
|
||||
run: |
|
||||
grep -v '^#' ./requirements-ci.txt | xargs -n 1 -L 1 pip install
|
||||
pip uninstall -y protobuf
|
||||
pip install --no-binary protobuf protobuf==3.20.*
|
||||
|
||||
- name: Cache kaldifeat
|
||||
id: my-cache
|
||||
uses: actions/cache@v2
|
||||
with:
|
||||
path: |
|
||||
~/tmp/kaldifeat
|
||||
key: cache-tmp-${{ matrix.python-version }}-2023-05-22
|
||||
|
||||
- name: Install kaldifeat
|
||||
if: steps.my-cache.outputs.cache-hit != 'true'
|
||||
shell: bash
|
||||
run: |
|
||||
make -j2 _kaldifeat
|
||||
|
||||
- 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.sh
|
185
.github/workflows/run-yesno-recipe.yml
vendored
185
.github/workflows/run-yesno-recipe.yml
vendored
@ -1,185 +0,0 @@
|
||||
# Copyright 2021 Fangjun Kuang (csukuangfj@gmail.com)
|
||||
|
||||
# See ../../LICENSE for clarification regarding multiple authors
|
||||
#
|
||||
# Licensed under the Apache License, Version 2.0 (the "License");
|
||||
# you may not use this file except in compliance with the License.
|
||||
# You may obtain a copy of the License at
|
||||
#
|
||||
# http://www.apache.org/licenses/LICENSE-2.0
|
||||
#
|
||||
# Unless required by applicable law or agreed to in writing, software
|
||||
# distributed under the License is distributed on an "AS IS" BASIS,
|
||||
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
||||
# See the License for the specific language governing permissions and
|
||||
# limitations under the License.
|
||||
|
||||
name: run-yesno-recipe
|
||||
|
||||
on:
|
||||
push:
|
||||
branches:
|
||||
- master
|
||||
pull_request:
|
||||
branches:
|
||||
- master
|
||||
|
||||
concurrency:
|
||||
group: run-yesno-recipe-${{ github.ref }}
|
||||
cancel-in-progress: true
|
||||
|
||||
jobs:
|
||||
run-yesno-recipe:
|
||||
runs-on: ${{ matrix.os }}
|
||||
strategy:
|
||||
matrix:
|
||||
# os: [ubuntu-latest, macos-10.15]
|
||||
# TODO: enable macOS for CPU testing
|
||||
os: [ubuntu-latest]
|
||||
python-version: [3.8]
|
||||
fail-fast: false
|
||||
|
||||
steps:
|
||||
- uses: actions/checkout@v2
|
||||
with:
|
||||
fetch-depth: 0
|
||||
|
||||
- name: Setup Python ${{ matrix.python-version }}
|
||||
uses: actions/setup-python@v2
|
||||
with:
|
||||
python-version: ${{ matrix.python-version }}
|
||||
cache: 'pip'
|
||||
cache-dependency-path: '**/requirements-ci.txt'
|
||||
|
||||
- name: Install libnsdfile and libsox
|
||||
if: startsWith(matrix.os, 'ubuntu')
|
||||
run: |
|
||||
sudo apt update
|
||||
sudo apt install -q -y libsndfile1-dev libsndfile1 ffmpeg
|
||||
sudo apt install -q -y --fix-missing sox libsox-dev libsox-fmt-all
|
||||
|
||||
- name: Install Python dependencies
|
||||
run: |
|
||||
grep -v '^#' ./requirements-ci.txt | xargs -n 1 -L 1 pip install
|
||||
pip uninstall -y protobuf
|
||||
pip install --no-binary protobuf protobuf==3.20.*
|
||||
|
||||
pip install --no-deps --force-reinstall k2==1.24.4.dev20231021+cpu.torch1.13.1 -f https://k2-fsa.github.io/k2/cpu.html
|
||||
pip install kaldifeat==1.25.1.dev20231022+cpu.torch1.13.1 -f https://csukuangfj.github.io/kaldifeat/cpu.html
|
||||
|
||||
- name: Run yesno recipe
|
||||
shell: bash
|
||||
working-directory: ${{github.workspace}}
|
||||
run: |
|
||||
export PYTHONPATH=$PWD:$PYTHONPATH
|
||||
echo $PYTHONPATH
|
||||
|
||||
cd egs/yesno/ASR
|
||||
./prepare.sh
|
||||
python3 ./tdnn/train.py
|
||||
python3 ./tdnn/decode.py
|
||||
|
||||
- name: Test exporting to pretrained.pt
|
||||
shell: bash
|
||||
working-directory: ${{github.workspace}}
|
||||
run: |
|
||||
export PYTHONPATH=$PWD:$PYTHONPATH
|
||||
echo $PYTHONPATH
|
||||
|
||||
cd egs/yesno/ASR
|
||||
python3 ./tdnn/export.py --epoch 14 --avg 2
|
||||
|
||||
python3 ./tdnn/pretrained.py \
|
||||
--checkpoint ./tdnn/exp/pretrained.pt \
|
||||
--HLG ./data/lang_phone/HLG.pt \
|
||||
--words-file ./data/lang_phone/words.txt \
|
||||
download/waves_yesno/0_0_0_1_0_0_0_1.wav \
|
||||
download/waves_yesno/0_0_1_0_0_0_1_0.wav
|
||||
|
||||
- name: Test exporting to torchscript
|
||||
shell: bash
|
||||
working-directory: ${{github.workspace}}
|
||||
run: |
|
||||
export PYTHONPATH=$PWD:$PYTHONPATH
|
||||
echo $PYTHONPATH
|
||||
|
||||
cd egs/yesno/ASR
|
||||
python3 ./tdnn/export.py --epoch 14 --avg 2 --jit 1
|
||||
|
||||
python3 ./tdnn/jit_pretrained.py \
|
||||
--nn-model ./tdnn/exp/cpu_jit.pt \
|
||||
--HLG ./data/lang_phone/HLG.pt \
|
||||
--words-file ./data/lang_phone/words.txt \
|
||||
download/waves_yesno/0_0_0_1_0_0_0_1.wav \
|
||||
download/waves_yesno/0_0_1_0_0_0_1_0.wav
|
||||
|
||||
- name: Test exporting to onnx
|
||||
shell: bash
|
||||
working-directory: ${{github.workspace}}
|
||||
run: |
|
||||
export PYTHONPATH=$PWD:$PYTHONPATH
|
||||
echo $PYTHONPATH
|
||||
|
||||
cd egs/yesno/ASR
|
||||
python3 ./tdnn/export_onnx.py --epoch 14 --avg 2
|
||||
|
||||
echo "Test float32 model"
|
||||
python3 ./tdnn/onnx_pretrained.py \
|
||||
--nn-model ./tdnn/exp/model-epoch-14-avg-2.onnx \
|
||||
--HLG ./data/lang_phone/HLG.pt \
|
||||
--words-file ./data/lang_phone/words.txt \
|
||||
download/waves_yesno/0_0_0_1_0_0_0_1.wav \
|
||||
download/waves_yesno/0_0_1_0_0_0_1_0.wav
|
||||
|
||||
|
||||
echo "Test int8 model"
|
||||
python3 ./tdnn/onnx_pretrained.py \
|
||||
--nn-model ./tdnn/exp/model-epoch-14-avg-2.int8.onnx \
|
||||
--HLG ./data/lang_phone/HLG.pt \
|
||||
--words-file ./data/lang_phone/words.txt \
|
||||
download/waves_yesno/0_0_0_1_0_0_0_1.wav \
|
||||
download/waves_yesno/0_0_1_0_0_0_1_0.wav
|
||||
|
||||
- name: Test decoding with H
|
||||
shell: bash
|
||||
working-directory: ${{github.workspace}}
|
||||
run: |
|
||||
export PYTHONPATH=$PWD:$PYTHONPATH
|
||||
echo $PYTHONPATH
|
||||
|
||||
cd egs/yesno/ASR
|
||||
python3 ./tdnn/export.py --epoch 14 --avg 2 --jit 1
|
||||
|
||||
python3 ./tdnn/jit_pretrained_decode_with_H.py \
|
||||
--nn-model ./tdnn/exp/cpu_jit.pt \
|
||||
--H ./data/lang_phone/H.fst \
|
||||
--tokens ./data/lang_phone/tokens.txt \
|
||||
./download/waves_yesno/0_0_0_1_0_0_0_1.wav \
|
||||
./download/waves_yesno/0_0_1_0_0_0_1_0.wav \
|
||||
./download/waves_yesno/0_0_1_0_0_1_1_1.wav
|
||||
|
||||
- name: Test decoding with HL
|
||||
shell: bash
|
||||
working-directory: ${{github.workspace}}
|
||||
run: |
|
||||
export PYTHONPATH=$PWD:$PYTHONPATH
|
||||
echo $PYTHONPATH
|
||||
|
||||
cd egs/yesno/ASR
|
||||
python3 ./tdnn/export.py --epoch 14 --avg 2 --jit 1
|
||||
|
||||
python3 ./tdnn/jit_pretrained_decode_with_HL.py \
|
||||
--nn-model ./tdnn/exp/cpu_jit.pt \
|
||||
--HL ./data/lang_phone/HL.fst \
|
||||
--words ./data/lang_phone/words.txt \
|
||||
./download/waves_yesno/0_0_0_1_0_0_0_1.wav \
|
||||
./download/waves_yesno/0_0_1_0_0_0_1_0.wav \
|
||||
./download/waves_yesno/0_0_1_0_0_1_1_1.wav
|
||||
|
||||
- name: Show generated files
|
||||
shell: bash
|
||||
working-directory: ${{github.workspace}}
|
||||
run: |
|
||||
cd egs/yesno/ASR
|
||||
ls -lh tdnn/exp
|
||||
ls -lh data/lang_phone
|
158
.github/workflows/test.yml
vendored
158
.github/workflows/test.yml
vendored
@ -1,129 +1,109 @@
|
||||
# Copyright 2021 Fangjun Kuang (csukuangfj@gmail.com)
|
||||
|
||||
# See ../../LICENSE for clarification regarding multiple authors
|
||||
#
|
||||
# Licensed under the Apache License, Version 2.0 (the "License");
|
||||
# you may not use this file except in compliance with the License.
|
||||
# You may obtain a copy of the License at
|
||||
#
|
||||
# http://www.apache.org/licenses/LICENSE-2.0
|
||||
#
|
||||
# Unless required by applicable law or agreed to in writing, software
|
||||
# distributed under the License is distributed on an "AS IS" BASIS,
|
||||
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
||||
# See the License for the specific language governing permissions and
|
||||
# limitations under the License.
|
||||
|
||||
name: test
|
||||
|
||||
on:
|
||||
push:
|
||||
branches:
|
||||
- master
|
||||
|
||||
pull_request:
|
||||
branches:
|
||||
- master
|
||||
|
||||
workflow_dispatch:
|
||||
|
||||
concurrency:
|
||||
group: test-${{ github.ref }}
|
||||
cancel-in-progress: true
|
||||
|
||||
jobs:
|
||||
generate_build_matrix:
|
||||
if: github.repository_owner == 'csukuangfj' || github.repository_owner == 'k2-fsa'
|
||||
# see https://github.com/pytorch/pytorch/pull/50633
|
||||
runs-on: ubuntu-latest
|
||||
outputs:
|
||||
matrix: ${{ steps.set-matrix.outputs.matrix }}
|
||||
steps:
|
||||
- uses: actions/checkout@v4
|
||||
with:
|
||||
fetch-depth: 0
|
||||
- name: Generating build matrix
|
||||
id: set-matrix
|
||||
run: |
|
||||
# outputting for debugging purposes
|
||||
python ./.github/scripts/docker/generate_build_matrix.py
|
||||
MATRIX=$(python ./.github/scripts/docker/generate_build_matrix.py)
|
||||
echo "::set-output name=matrix::${MATRIX}"
|
||||
test:
|
||||
runs-on: ${{ matrix.os }}
|
||||
needs: generate_build_matrix
|
||||
name: py${{ matrix.python-version }} torch${{ matrix.torch-version }} v${{ matrix.version }}
|
||||
runs-on: ubuntu-latest
|
||||
strategy:
|
||||
matrix:
|
||||
os: [ubuntu-latest]
|
||||
python-version: ["3.8"]
|
||||
torch: ["1.13.0"]
|
||||
torchaudio: ["0.13.0"]
|
||||
k2-version: ["1.24.3.dev20230719"]
|
||||
|
||||
fail-fast: false
|
||||
matrix:
|
||||
${{ fromJson(needs.generate_build_matrix.outputs.matrix) }}
|
||||
|
||||
steps:
|
||||
- uses: actions/checkout@v2
|
||||
- uses: actions/checkout@v4
|
||||
with:
|
||||
fetch-depth: 0
|
||||
|
||||
- name: Setup Python ${{ matrix.python-version }}
|
||||
uses: actions/setup-python@v1
|
||||
with:
|
||||
python-version: ${{ matrix.python-version }}
|
||||
|
||||
- name: Install libnsdfile and libsox
|
||||
if: startsWith(matrix.os, 'ubuntu')
|
||||
run: |
|
||||
sudo apt update
|
||||
sudo apt install -q -y libsndfile1-dev libsndfile1 ffmpeg
|
||||
sudo apt install -q -y --fix-missing libsox-dev libsox-fmt-all
|
||||
|
||||
- name: Install Python dependencies
|
||||
run: |
|
||||
python3 -m pip install --upgrade pip pytest
|
||||
# numpy 1.20.x does not support python 3.6
|
||||
pip install numpy==1.19
|
||||
pip install torch==${{ matrix.torch }}+cpu -f https://download.pytorch.org/whl/cpu/torch_stable.html
|
||||
pip install torchaudio==${{ matrix.torchaudio }}+cpu -f https://download.pytorch.org/whl/cpu/torch_stable.html
|
||||
|
||||
pip install k2==${{ matrix.k2-version }}+cpu.torch${{ matrix.torch }} -f https://k2-fsa.github.io/k2/cpu.html
|
||||
pip install git+https://github.com/lhotse-speech/lhotse
|
||||
# icefall requirements
|
||||
pip uninstall -y protobuf
|
||||
pip install --no-binary protobuf protobuf==3.20.*
|
||||
|
||||
pip install kaldifst
|
||||
pip install onnxruntime matplotlib
|
||||
pip install -r requirements.txt
|
||||
|
||||
- name: Install graphviz
|
||||
if: startsWith(matrix.os, 'ubuntu')
|
||||
- name: Free space
|
||||
shell: bash
|
||||
run: |
|
||||
python3 -m pip install -qq graphviz
|
||||
sudo apt-get -qq install graphviz
|
||||
df -h
|
||||
rm -rf /opt/hostedtoolcache
|
||||
df -h
|
||||
echo "pwd: $PWD"
|
||||
echo "github.workspace ${{ github.workspace }}"
|
||||
|
||||
- name: Run tests
|
||||
if: startsWith(matrix.os, 'ubuntu')
|
||||
run: |
|
||||
ls -lh
|
||||
export PYTHONPATH=$PWD:$PWD/lhotse:$PYTHONPATH
|
||||
echo $PYTHONPATH
|
||||
pytest -v -s ./test
|
||||
# runt tests for conformer ctc
|
||||
cd egs/librispeech/ASR/conformer_ctc
|
||||
pytest -v -s
|
||||
uses: addnab/docker-run-action@v3
|
||||
with:
|
||||
image: ghcr.io/${{ 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
|
||||
|
||||
cd ../pruned_transducer_stateless
|
||||
pytest -v -s
|
||||
pytest -v -s ./test
|
||||
|
||||
cd ../pruned_transducer_stateless2
|
||||
pytest -v -s
|
||||
# runt tests for conformer ctc
|
||||
cd egs/librispeech/ASR/conformer_ctc
|
||||
pytest -v -s
|
||||
|
||||
cd ../pruned_transducer_stateless3
|
||||
pytest -v -s
|
||||
cd ../pruned_transducer_stateless
|
||||
pytest -v -s
|
||||
|
||||
cd ../pruned_transducer_stateless4
|
||||
pytest -v -s
|
||||
cd ../pruned_transducer_stateless2
|
||||
pytest -v -s
|
||||
|
||||
echo $PYTHONPATH
|
||||
cd ../pruned_transducer_stateless7
|
||||
pytest -v -s
|
||||
cd ../pruned_transducer_stateless3
|
||||
pytest -v -s
|
||||
|
||||
cd ../transducer_stateless
|
||||
pytest -v -s
|
||||
cd ../pruned_transducer_stateless4
|
||||
pytest -v -s
|
||||
|
||||
# cd ../transducer
|
||||
# pytest -v -s
|
||||
echo $PYTHONPATH
|
||||
cd ../pruned_transducer_stateless7
|
||||
pytest -v -s
|
||||
|
||||
cd ../transducer_stateless2
|
||||
pytest -v -s
|
||||
cd ../transducer_stateless
|
||||
pytest -v -s
|
||||
|
||||
cd ../transducer_lstm
|
||||
pytest -v -s
|
||||
# cd ../transducer
|
||||
# pytest -v -s
|
||||
|
||||
cd ../zipformer
|
||||
pytest -v -s
|
||||
cd ../transducer_stateless2
|
||||
pytest -v -s
|
||||
|
||||
cd ../transducer_lstm
|
||||
pytest -v -s
|
||||
|
||||
cd ../zipformer
|
||||
pytest -v -s
|
||||
|
||||
- uses: actions/upload-artifact@v2
|
||||
with:
|
||||
|
62
.github/workflows/yesno.yml
vendored
Normal file
62
.github/workflows/yesno.yml
vendored
Normal file
@ -0,0 +1,62 @@
|
||||
name: yesno
|
||||
|
||||
on:
|
||||
push:
|
||||
branches:
|
||||
- master
|
||||
|
||||
pull_request:
|
||||
branches:
|
||||
- master
|
||||
|
||||
workflow_dispatch:
|
||||
|
||||
concurrency:
|
||||
group: yesno-${{ 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
|
||||
|
||||
.github/scripts/yesno/ASR/run.sh
|
@ -20,7 +20,11 @@ We describe the following items in this section:
|
||||
View available tags
|
||||
===================
|
||||
|
||||
You can use the following command to view available tags:
|
||||
CUDA-enabled docker images
|
||||
--------------------------
|
||||
|
||||
You can use the following command to view available tags for CUDA-enabled
|
||||
docker images:
|
||||
|
||||
.. code-block:: bash
|
||||
|
||||
@ -43,8 +47,25 @@ which will give you something like below:
|
||||
|
||||
Please select an appropriate combination of `torch`_ and CUDA.
|
||||
|
||||
Download a docker image
|
||||
=======================
|
||||
CPU-only docker images
|
||||
----------------------
|
||||
|
||||
To view CPU-only docker images, please visit `<https://github.com/k2-fsa/icefall/pkgs/container/icefall>`_
|
||||
for available tags.
|
||||
|
||||
You can select different combinations of ``Python`` and ``torch``. For instance,
|
||||
to select ``Python 3.8`` and ``torch 2.1.2``, you can use the following tag
|
||||
|
||||
.. code-block:: bash
|
||||
|
||||
cpu-py3.8-torch2.1.2-v1.1
|
||||
|
||||
where ``v1.1`` is the current version of the docker image. You may see
|
||||
``ghcr.io/k2-fsa/icefall:cpu-py3.8-torch2.1.2-v1.2`` or some other versions.
|
||||
We recommend that you always use the latest version.
|
||||
|
||||
Download a docker image (CUDA)
|
||||
==============================
|
||||
|
||||
Suppose that you select the tag ``torch1.13.0-cuda11.6``, you can use
|
||||
the following command to download it:
|
||||
@ -53,6 +74,16 @@ the following command to download it:
|
||||
|
||||
sudo docker image pull k2fsa/icefall:torch1.13.0-cuda11.6
|
||||
|
||||
Download a docker image (CPU)
|
||||
==============================
|
||||
|
||||
Suppose that you select the tag ``cpu-py3.8-torch2.1.2-v1.1``, you can use
|
||||
the following command to download it:
|
||||
|
||||
.. code-block:: bash
|
||||
|
||||
sudo docker pull ghcr.io/k2-fsa/icefall:cpu-py3.8-torch2.1.2-v1.1
|
||||
|
||||
Run a docker image with GPU
|
||||
===========================
|
||||
|
||||
@ -65,7 +96,7 @@ Run a docker image with CPU
|
||||
|
||||
.. code-block:: bash
|
||||
|
||||
sudo docker run --rm -it k2fsa/icefall:torch1.13.0-cuda11.6 /bin/bash
|
||||
sudo docker run --rm -it ghcr.io/k2-fsa/icefall:cpu-py3.8-torch2.1.2-v1.1 /bin/bash
|
||||
|
||||
Run yesno within a docker container
|
||||
===================================
|
||||
@ -74,8 +105,13 @@ After starting the container, the following interface is presented:
|
||||
|
||||
.. code-block:: bash
|
||||
|
||||
# GPU-enabled docker
|
||||
root@60c947eac59c:/workspace/icefall#
|
||||
|
||||
# CPU-only docker
|
||||
root@60c947eac59c:# mkdir /workspace; git clone https://github.com/k2-fsa/icefall
|
||||
root@60c947eac59c:# export PYTHONPATH=/workspace/icefall:$PYTHONPATH
|
||||
|
||||
It shows the current user is ``root`` and the current working directory
|
||||
is ``/workspace/icefall``.
|
||||
|
||||
@ -107,7 +143,7 @@ to switch to the ``yesno`` recipe and run
|
||||
|
||||
.. hint::
|
||||
|
||||
If you are running without GPU, it may report the following error:
|
||||
If you are running without GPU with a GPU-enabled docker, it may report the following error:
|
||||
|
||||
.. code-block:: bash
|
||||
|
||||
|
@ -66,13 +66,13 @@ to install dependencies of `icefall`_:
|
||||
|
||||
pip install torch==2.0.0+cpu torchaudio==2.0.0+cpu -f https://download.pytorch.org/whl/torch_stable.html
|
||||
|
||||
# If you are using macOS or Windows, please use the following command to install torch and torchaudio
|
||||
# If you are using macOS, please use the following command to install torch and torchaudio
|
||||
# pip install torch==2.0.0 torchaudio==2.0.0 -f https://download.pytorch.org/whl/torch_stable.html
|
||||
|
||||
# Now install k2
|
||||
# Please refer to https://k2-fsa.github.io/k2/installation/from_wheels.html#linux-cpu-example
|
||||
|
||||
pip install k2==1.24.3.dev20230726+cpu.torch2.0.0 -f https://k2-fsa.github.io/k2/cpu.html
|
||||
pip install k2==1.24.4.dev20231220+cpu.torch2.0.0 -f https://k2-fsa.github.io/k2/cpu.html
|
||||
|
||||
# Install the latest version of lhotse
|
||||
|
||||
|
@ -85,7 +85,7 @@ We can also use it to decode files with the following command:
|
||||
# Please refer to https://csukuangfj.github.io/kaldifeat/installation/from_wheels.html
|
||||
# for how to install kaldifeat
|
||||
|
||||
pip install kaldifeat==1.25.0.dev20230726+cpu.torch2.0.0 -f https://csukuangfj.github.io/kaldifeat/cpu.html
|
||||
pip install kaldifeat==1.25.3.dev20231221+cpu.torch2.0.0 -f https://csukuangfj.github.io/kaldifeat/cpu.html
|
||||
|
||||
./tdnn/pretrained.py \
|
||||
--checkpoint ./tdnn/exp/pretrained.pt \
|
||||
@ -162,7 +162,7 @@ To use ``tdnn/exp/cpu_jit.pt`` with `icefall`_ to decode files, we can use:
|
||||
# Please refer to https://csukuangfj.github.io/kaldifeat/installation/from_wheels.html
|
||||
# for how to install kaldifeat
|
||||
|
||||
pip install kaldifeat==1.25.0.dev20230726+cpu.torch2.0.0 -f https://csukuangfj.github.io/kaldifeat/cpu.html
|
||||
pip install kaldifeat==1.25.3.dev20231221+cpu.torch2.0.0 -f https://csukuangfj.github.io/kaldifeat/cpu.html
|
||||
|
||||
|
||||
./tdnn/jit_pretrained.py \
|
||||
@ -249,7 +249,7 @@ To use the generated ONNX model files for decoding with `onnxruntime`_, we can u
|
||||
# Please refer to https://csukuangfj.github.io/kaldifeat/installation/from_wheels.html
|
||||
# for how to install kaldifeat
|
||||
|
||||
pip install kaldifeat==1.25.0.dev20230726+cpu.torch2.0.0 -f https://csukuangfj.github.io/kaldifeat/cpu.html
|
||||
pip install kaldifeat==1.25.3.dev20231221+cpu.torch2.0.0 -f https://csukuangfj.github.io/kaldifeat/cpu.html
|
||||
|
||||
./tdnn/onnx_pretrained.py \
|
||||
--nn-model ./tdnn/exp/model-epoch-14-avg-2.onnx \
|
||||
|
@ -4,7 +4,7 @@ Train an RNN language model
|
||||
======================================
|
||||
|
||||
If you have enough text data, you can train a neural network language model (NNLM) to improve
|
||||
the WER of your E2E ASR system. This tutorial shows you how to train an RNNLM from
|
||||
the WER of your E2E ASR system. This tutorial shows you how to train an RNNLM from
|
||||
scratch.
|
||||
|
||||
.. HINT::
|
||||
@ -15,23 +15,23 @@ scratch.
|
||||
.. note::
|
||||
|
||||
This tutorial is based on the LibriSpeech recipe. Please check it out for the necessary
|
||||
python scripts for this tutorial. We use the LibriSpeech LM-corpus as the LM training set
|
||||
python scripts for this tutorial. We use the LibriSpeech LM-corpus as the LM training set
|
||||
for illustration purpose. You can also collect your own data. The data format is quite simple:
|
||||
each line should contain a complete sentence, and words should be separated by space.
|
||||
|
||||
First, let's download the training data for the RNNLM. This can be done via the
|
||||
First, let's download the training data for the RNNLM. This can be done via the
|
||||
following command:
|
||||
|
||||
.. code-block:: bash
|
||||
|
||||
$ wget https://www.openslr.org/resources/11/librispeech-lm-norm.txt.gz
|
||||
$ wget https://www.openslr.org/resources/11/librispeech-lm-norm.txt.gz
|
||||
$ gzip -d librispeech-lm-norm.txt.gz
|
||||
|
||||
As we are training a BPE-level RNNLM, we need to tokenize the training text, which requires a
|
||||
BPE tokenizer. This can be achieved by executing the following command:
|
||||
|
||||
.. code-block:: bash
|
||||
|
||||
|
||||
$ # if you don't have the BPE
|
||||
$ GIT_LFS_SKIP_SMUDGE=1 git clone https://huggingface.co/Zengwei/icefall-asr-librispeech-zipformer-2023-05-15
|
||||
$ cd icefall-asr-librispeech-zipformer-2023-05-15/data/lang_bpe_500
|
||||
@ -56,11 +56,11 @@ sentence length.
|
||||
--out-statistics data/lang_bpe_500/lm_data_stats.txt
|
||||
|
||||
|
||||
The aforementioned steps can be repeated to create a a validation set for you RNNLM. Let's say
|
||||
you have a validation set in ``valid.txt``, you can just set ``--lm-data valid.txt``
|
||||
The aforementioned steps can be repeated to create a a validation set for you RNNLM. Let's say
|
||||
you have a validation set in ``valid.txt``, you can just set ``--lm-data valid.txt``
|
||||
and ``--lm-archive data/lang_bpe_500/lm-data-valid.pt`` when calling ``./local/prepare_lm_training_data.py``.
|
||||
|
||||
After completing the previous steps, the training and testing sets for training RNNLM are ready.
|
||||
After completing the previous steps, the training and testing sets for training RNNLM are ready.
|
||||
The next step is to train the RNNLM model. The training command is as follows:
|
||||
|
||||
.. code-block:: bash
|
||||
@ -77,7 +77,7 @@ The next step is to train the RNNLM model. The training command is as follows:
|
||||
--use-fp16 0 \
|
||||
--tie-weights 1 \
|
||||
--embedding-dim 2048 \
|
||||
--hidden_dim 2048 \
|
||||
--hidden-dim 2048 \
|
||||
--num-layers 3 \
|
||||
--batch-size 300 \
|
||||
--lm-data rnn_lm/data/lang_bpe_500/sorted_lm_data.pt \
|
||||
@ -93,12 +93,3 @@ The next step is to train the RNNLM model. The training command is as follows:
|
||||
.. note::
|
||||
|
||||
The training of RNNLM can take a long time (usually a couple of days).
|
||||
|
||||
|
||||
|
||||
|
||||
|
||||
|
||||
|
||||
|
||||
|
||||
|
@ -242,6 +242,7 @@ def main():
|
||||
opts.frame_opts.snip_edges = False
|
||||
opts.frame_opts.samp_freq = params.sample_rate
|
||||
opts.mel_opts.num_bins = params.feature_dim
|
||||
opts.mel_opts.high_freq = -400
|
||||
|
||||
fbank = kaldifeat.Fbank(opts)
|
||||
|
||||
|
@ -261,6 +261,7 @@ def main():
|
||||
opts.frame_opts.snip_edges = False
|
||||
opts.frame_opts.samp_freq = params.sample_rate
|
||||
opts.mel_opts.num_bins = params.feature_dim
|
||||
opts.mel_opts.high_freq = -400
|
||||
|
||||
fbank = kaldifeat.Fbank(opts)
|
||||
|
||||
|
@ -240,6 +240,7 @@ def main():
|
||||
opts.frame_opts.snip_edges = False
|
||||
opts.frame_opts.samp_freq = params.sample_rate
|
||||
opts.mel_opts.num_bins = params.feature_dim
|
||||
opts.mel_opts.high_freq = -400
|
||||
|
||||
fbank = kaldifeat.Fbank(opts)
|
||||
|
||||
|
@ -241,6 +241,7 @@ def main():
|
||||
opts.frame_opts.snip_edges = False
|
||||
opts.frame_opts.samp_freq = params.sample_rate
|
||||
opts.mel_opts.num_bins = params.feature_dim
|
||||
opts.mel_opts.high_freq = -400
|
||||
|
||||
fbank = kaldifeat.Fbank(opts)
|
||||
|
||||
|
@ -230,6 +230,7 @@ def main():
|
||||
opts.frame_opts.snip_edges = False
|
||||
opts.frame_opts.samp_freq = 16000
|
||||
opts.mel_opts.num_bins = 80
|
||||
opts.mel_opts.high_freq = -400
|
||||
|
||||
fbank = kaldifeat.Fbank(opts)
|
||||
|
||||
|
@ -369,6 +369,7 @@ def main():
|
||||
opts.frame_opts.snip_edges = False
|
||||
opts.frame_opts.samp_freq = args.sample_rate
|
||||
opts.mel_opts.num_bins = 80
|
||||
opts.mel_opts.high_freq = -400
|
||||
|
||||
fbank = kaldifeat.Fbank(opts)
|
||||
|
||||
|
@ -227,6 +227,7 @@ def main():
|
||||
opts.frame_opts.snip_edges = False
|
||||
opts.frame_opts.samp_freq = 16000
|
||||
opts.mel_opts.num_bins = 80
|
||||
opts.mel_opts.high_freq = -400
|
||||
|
||||
fbank = kaldifeat.Fbank(opts)
|
||||
|
||||
|
@ -250,6 +250,7 @@ def main():
|
||||
opts.frame_opts.snip_edges = False
|
||||
opts.frame_opts.samp_freq = params.sample_rate
|
||||
opts.mel_opts.num_bins = params.feature_dim
|
||||
opts.mel_opts.high_freq = -400
|
||||
|
||||
fbank = kaldifeat.Fbank(opts)
|
||||
|
||||
|
@ -317,6 +317,7 @@ def decode_dataset(
|
||||
opts.frame_opts.snip_edges = False
|
||||
opts.frame_opts.samp_freq = 16000
|
||||
opts.mel_opts.num_bins = 80
|
||||
opts.mel_opts.high_freq = -400
|
||||
|
||||
log_interval = 50
|
||||
|
||||
@ -341,7 +342,12 @@ def decode_dataset(
|
||||
assert audio.dtype == np.float32, audio.dtype
|
||||
|
||||
# The trained model is using normalized samples
|
||||
assert audio.max() <= 1, "Should be normalized to [-1, 1])"
|
||||
# - this is to avoid sending [-32k,+32k] signal in...
|
||||
# - some lhotse AudioTransform classes can make the signal
|
||||
# be out of range [-1, 1], hence the tolerance 10
|
||||
assert (
|
||||
np.abs(audio).max() <= 10
|
||||
), "Should be normalized to [-1, 1], 10 for tolerance..."
|
||||
|
||||
samples = torch.from_numpy(audio).squeeze(0)
|
||||
|
||||
|
@ -158,6 +158,7 @@ def main():
|
||||
opts.frame_opts.snip_edges = False
|
||||
opts.frame_opts.samp_freq = params.sample_rate
|
||||
opts.mel_opts.num_bins = params.feature_dim
|
||||
opts.mel_opts.high_freq = -400
|
||||
|
||||
fbank = kaldifeat.Fbank(opts)
|
||||
|
||||
|
@ -258,6 +258,7 @@ def main():
|
||||
opts.frame_opts.snip_edges = False
|
||||
opts.frame_opts.samp_freq = params.sample_rate
|
||||
opts.mel_opts.num_bins = params.feature_dim
|
||||
opts.mel_opts.high_freq = -400
|
||||
|
||||
fbank = kaldifeat.Fbank(opts)
|
||||
|
||||
|
@ -238,6 +238,7 @@ def main():
|
||||
opts.frame_opts.snip_edges = False
|
||||
opts.frame_opts.samp_freq = params.sample_rate
|
||||
opts.mel_opts.num_bins = params.feature_dim
|
||||
opts.mel_opts.high_freq = -400
|
||||
|
||||
fbank = kaldifeat.Fbank(opts)
|
||||
|
||||
|
@ -238,6 +238,7 @@ def main():
|
||||
opts.frame_opts.snip_edges = False
|
||||
opts.frame_opts.samp_freq = params.sample_rate
|
||||
opts.mel_opts.num_bins = params.feature_dim
|
||||
opts.mel_opts.high_freq = -400
|
||||
|
||||
fbank = kaldifeat.Fbank(opts)
|
||||
|
||||
|
@ -572,6 +572,7 @@ def decode_dataset(
|
||||
opts.frame_opts.snip_edges = False
|
||||
opts.frame_opts.samp_freq = 16000
|
||||
opts.mel_opts.num_bins = 80
|
||||
opts.mel_opts.high_freq = -400
|
||||
|
||||
log_interval = 100
|
||||
|
||||
@ -596,12 +597,12 @@ def decode_dataset(
|
||||
assert audio.dtype == np.float32, audio.dtype
|
||||
|
||||
# The trained model is using normalized samples
|
||||
if audio.max() > 1:
|
||||
logging.warning(
|
||||
f"The audio should be normalized to [-1, 1], audio.max : {audio.max()}."
|
||||
f"Clipping to [-1, 1]."
|
||||
)
|
||||
audio = np.clip(audio, -1, 1)
|
||||
# - this is to avoid sending [-32k,+32k] signal in...
|
||||
# - some lhotse AudioTransform classes can make the signal
|
||||
# be out of range [-1, 1], hence the tolerance 10
|
||||
assert (
|
||||
np.abs(audio).max() <= 10
|
||||
), "Should be normalized to [-1, 1], 10 for tolerance..."
|
||||
|
||||
samples = torch.from_numpy(audio).squeeze(0)
|
||||
|
||||
|
@ -239,6 +239,7 @@ def main():
|
||||
opts.frame_opts.snip_edges = False
|
||||
opts.frame_opts.samp_freq = params.sample_rate
|
||||
opts.mel_opts.num_bins = params.feature_dim
|
||||
opts.mel_opts.high_freq = -400
|
||||
|
||||
fbank = kaldifeat.Fbank(opts)
|
||||
|
||||
|
@ -251,6 +251,7 @@ def main():
|
||||
opts.frame_opts.snip_edges = False
|
||||
opts.frame_opts.samp_freq = params.sample_rate
|
||||
opts.mel_opts.num_bins = params.feature_dim
|
||||
opts.mel_opts.high_freq = -400
|
||||
|
||||
fbank = kaldifeat.Fbank(opts)
|
||||
|
||||
|
@ -242,6 +242,7 @@ def main():
|
||||
opts.frame_opts.snip_edges = False
|
||||
opts.frame_opts.samp_freq = params.sample_rate
|
||||
opts.mel_opts.num_bins = params.feature_dim
|
||||
opts.mel_opts.high_freq = -400
|
||||
|
||||
fbank = kaldifeat.Fbank(opts)
|
||||
|
||||
|
@ -109,10 +109,10 @@ def compute_fbank_commonvoice_splits(args):
|
||||
extractor = KaldifeatFbank(KaldifeatFbankConfig(device=device))
|
||||
logging.info(f"device: {device}")
|
||||
|
||||
set_audio_duration_mismatch_tolerance(0.01) # 10ms tolerance
|
||||
set_audio_duration_mismatch_tolerance(0.05) # 50ms tolerance
|
||||
set_caching_enabled(False)
|
||||
for i in range(start, stop):
|
||||
idx = f"{i + 1}".zfill(num_digits)
|
||||
idx = f"{i}".zfill(num_digits)
|
||||
logging.info(f"Processing {idx}/{num_splits}")
|
||||
|
||||
cuts_path = output_dir / f"cv-{language}_cuts_{subset}.{idx}.jsonl.gz"
|
||||
|
@ -370,6 +370,7 @@ def main():
|
||||
opts.frame_opts.snip_edges = False
|
||||
opts.frame_opts.samp_freq = args.sample_rate
|
||||
opts.mel_opts.num_bins = 80
|
||||
opts.mel_opts.high_freq = -400
|
||||
|
||||
fbank = kaldifeat.Fbank(opts)
|
||||
|
||||
|
@ -260,6 +260,7 @@ def main():
|
||||
opts.frame_opts.snip_edges = False
|
||||
opts.frame_opts.samp_freq = params.sample_rate
|
||||
opts.mel_opts.num_bins = params.feature_dim
|
||||
opts.mel_opts.high_freq = -400
|
||||
|
||||
fbank = kaldifeat.Fbank(opts)
|
||||
|
||||
|
@ -320,6 +320,7 @@ def decode_dataset(
|
||||
opts.frame_opts.snip_edges = False
|
||||
opts.frame_opts.samp_freq = 16000
|
||||
opts.mel_opts.num_bins = 80
|
||||
opts.mel_opts.high_freq = -400
|
||||
|
||||
log_interval = 50
|
||||
|
||||
|
@ -177,6 +177,7 @@ def create_streaming_feature_extractor(sample_rate) -> OnlineFeature:
|
||||
opts.frame_opts.snip_edges = False
|
||||
opts.frame_opts.samp_freq = sample_rate
|
||||
opts.mel_opts.num_bins = 80
|
||||
opts.mel_opts.high_freq = -400
|
||||
return OnlineFbank(opts)
|
||||
|
||||
|
||||
|
@ -252,6 +252,7 @@ def main():
|
||||
opts.frame_opts.snip_edges = False
|
||||
opts.frame_opts.samp_freq = params.sample_rate
|
||||
opts.mel_opts.num_bins = params.feature_dim
|
||||
opts.mel_opts.high_freq = -400
|
||||
|
||||
fbank = kaldifeat.Fbank(opts)
|
||||
|
||||
|
@ -337,6 +337,7 @@ def decode_dataset(
|
||||
opts.frame_opts.snip_edges = False
|
||||
opts.frame_opts.samp_freq = 16000
|
||||
opts.mel_opts.num_bins = 80
|
||||
opts.mel_opts.high_freq = -400
|
||||
|
||||
log_interval = 50
|
||||
|
||||
@ -361,7 +362,12 @@ def decode_dataset(
|
||||
assert audio.dtype == np.float32, audio.dtype
|
||||
|
||||
# The trained model is using normalized samples
|
||||
assert audio.max() <= 1, "Should be normalized to [-1, 1])"
|
||||
# - this is to avoid sending [-32k,+32k] signal in...
|
||||
# - some lhotse AudioTransform classes can make the signal
|
||||
# be out of range [-1, 1], hence the tolerance 10
|
||||
assert (
|
||||
np.abs(audio).max() <= 10
|
||||
), "Should be normalized to [-1, 1], 10 for tolerance..."
|
||||
|
||||
samples = torch.from_numpy(audio).squeeze(0)
|
||||
|
||||
|
1
egs/gigaspeech/ASR/zipformer/my_profile.py
Symbolic link
1
egs/gigaspeech/ASR/zipformer/my_profile.py
Symbolic link
@ -0,0 +1 @@
|
||||
../../../librispeech/ASR/zipformer/my_profile.py
|
@ -1 +0,0 @@
|
||||
../../../librispeech/ASR/zipformer/profile.py
|
@ -553,6 +553,7 @@ def decode_dataset(
|
||||
opts.frame_opts.snip_edges = False
|
||||
opts.frame_opts.samp_freq = 16000
|
||||
opts.mel_opts.num_bins = 80
|
||||
opts.mel_opts.high_freq = -400
|
||||
|
||||
log_interval = 100
|
||||
|
||||
@ -577,7 +578,12 @@ def decode_dataset(
|
||||
assert audio.dtype == np.float32, audio.dtype
|
||||
|
||||
# The trained model is using normalized samples
|
||||
assert audio.max() <= 1, "Should be normalized to [-1, 1])"
|
||||
# - this is to avoid sending [-32k,+32k] signal in...
|
||||
# - some lhotse AudioTransform classes can make the signal
|
||||
# be out of range [-1, 1], hence the tolerance 10
|
||||
assert (
|
||||
np.abs(audio).max() <= 10
|
||||
), "Should be normalized to [-1, 1], 10 for tolerance..."
|
||||
|
||||
samples = torch.from_numpy(audio).squeeze(0)
|
||||
|
||||
|
@ -248,7 +248,7 @@ if [ $stage -le 10 ] && [ $stop_stage -ge 10 ]; then
|
||||
| jq '.supervisions[].text' | sed 's/"//;s/\\//g;s/"$//' > data/punc_texts
|
||||
fi
|
||||
for vocab_size in ${vocab_sizes[@]}; do
|
||||
new_vacab_size = $(($vocab_size + 256))
|
||||
new_vocab_size=$(($vocab_size + 256))
|
||||
lang_dir=data/lang_punc_bpe_${new_vocab_size}
|
||||
mkdir -p $lang_dir
|
||||
|
||||
|
@ -264,6 +264,7 @@ def main():
|
||||
opts.frame_opts.snip_edges = False
|
||||
opts.frame_opts.samp_freq = params.sample_rate
|
||||
opts.mel_opts.num_bins = params.feature_dim
|
||||
opts.mel_opts.high_freq = -400
|
||||
|
||||
fbank = kaldifeat.Fbank(opts)
|
||||
|
||||
|
@ -195,6 +195,7 @@ def main():
|
||||
opts.frame_opts.snip_edges = False
|
||||
opts.frame_opts.samp_freq = sample_rate
|
||||
opts.mel_opts.num_bins = 80
|
||||
opts.mel_opts.high_freq = -400
|
||||
|
||||
fbank = kaldifeat.Fbank(opts)
|
||||
|
||||
|
@ -192,6 +192,7 @@ def main():
|
||||
opts.frame_opts.snip_edges = False
|
||||
opts.frame_opts.samp_freq = sample_rate
|
||||
opts.mel_opts.num_bins = 80
|
||||
opts.mel_opts.high_freq = -400
|
||||
|
||||
fbank = kaldifeat.Fbank(opts)
|
||||
|
||||
|
@ -191,6 +191,7 @@ def main():
|
||||
opts.frame_opts.snip_edges = False
|
||||
opts.frame_opts.samp_freq = sample_rate
|
||||
opts.mel_opts.num_bins = 80
|
||||
opts.mel_opts.high_freq = -400
|
||||
|
||||
fbank = kaldifeat.Fbank(opts)
|
||||
|
||||
|
@ -283,6 +283,7 @@ def main():
|
||||
opts.frame_opts.snip_edges = False
|
||||
opts.frame_opts.samp_freq = params.sample_rate
|
||||
opts.mel_opts.num_bins = params.feature_dim
|
||||
opts.mel_opts.high_freq = -400
|
||||
|
||||
fbank = kaldifeat.Fbank(opts)
|
||||
|
||||
|
@ -271,6 +271,7 @@ def main():
|
||||
opts.frame_opts.snip_edges = False
|
||||
opts.frame_opts.samp_freq = params.sample_rate
|
||||
opts.mel_opts.num_bins = params.feature_dim
|
||||
opts.mel_opts.high_freq = -400
|
||||
|
||||
fbank = kaldifeat.Fbank(opts)
|
||||
|
||||
|
Some files were not shown because too many files have changed in this diff Show More
Loading…
x
Reference in New Issue
Block a user