mirrors/icefall
1
0
Fork 0
mirror of https://github.com/k2-fsa/icefall.git synced 2025-08-26 18:24:18 +00:00
Code Issues Packages Projects Releases Wiki Activity

Merge branch 'k2-fsa:master' into fisher_swbd

Browse Source
This commit is contained in:
Nagendra Goel 2022-12-18 15:13:59 -05:00 committed by GitHub
commit 72b0aa3fbf
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
832 changed files with 74630 additions and 6940 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

5
.flake8
View File

@ -1,7 +1,7 @@
[flake8]
show-source=true
statistics=true
max-line-length = 80
max-line-length = 88
per-file-ignores =
# line too long
icefall/diagnostics.py: E501,
@ -11,7 +11,8 @@ per-file-ignores =
egs/*/ASR/*/scaling.py: E501,
egs/librispeech/ASR/lstm_transducer_stateless*/*.py: E501, E203
egs/librispeech/ASR/conv_emformer_transducer_stateless*/*.py: E501, E203
egs/librispeech/ASR/conformer_ctc2/*py: E501,
egs/librispeech/ASR/conformer_ctc*/*py: E501,
egs/librispeech/ASR/zipformer_mmi/*.py: E501, E203
egs/librispeech/ASR/RESULTS.md: E999,
# invalid escape sequence (cause by tex formular), W605

3
.git-blame-ignore-revs Normal file
View File

@ -0,0 +1,3 @@
# Migrate to 88 characters per line (see: https://github.com/lhotse-speech/lhotse/issues/890)
107df3b115a58f1b68a6458c3f94a130004be34c
d31db010371a4128856480382876acdc0d1739ed

2
.github/scripts/compute-fbank-librispeech-test-clean-and-test-other.sh vendored
View File

@ -4,6 +4,8 @@
# The computed features are saved to ~/tmp/fbank-libri and are
# cached for later runs
set -e
export PYTHONPATH=$PWD:$PYTHONPATH
echo $PYTHONPATH

2
.github/scripts/download-gigaspeech-dev-test-dataset.sh vendored
View File

@ -6,6 +6,8 @@
# You will find directories `~/tmp/giga-dev-dataset-fbank` after running
# this script.
set -e
mkdir -p ~/tmp
cd ~/tmp

2
.github/scripts/download-librispeech-test-clean-and-test-other-dataset.sh vendored
View File

@ -7,6 +7,8 @@
# You will find directories ~/tmp/download/LibriSpeech after running
# this script.
set -e
mkdir ~/tmp/download
cd egs/librispeech/ASR
ln -s ~/tmp/download .

2
.github/scripts/install-kaldifeat.sh vendored
View File

@ -3,6 +3,8 @@
# This script installs kaldifeat into the directory ~/tmp/kaldifeat
# which is cached by GitHub actions for later runs.
set -e
mkdir -p ~/tmp
cd ~/tmp
git clone https://github.com/csukuangfj/kaldifeat

2
.github/scripts/prepare-librispeech-test-clean-and-test-other-manifests.sh vendored
View File

@ -4,6 +4,8 @@
# to egs/librispeech/ASR/download/LibriSpeech and generates manifest
# files in egs/librispeech/ASR/data/manifests
set -e
cd egs/librispeech/ASR
[ ! -e download ] && ln -s ~/tmp/download .
mkdir -p data/manifests

6
.github/scripts/run-aishell-pruned-transducer-stateless3-2022-06-20.sh vendored
View File

@ -1,5 +1,7 @@
#!/usr/bin/env bash
set -e
log() {
# This function is from espnet
local fname=${BASH_SOURCE[1]##*/}
@ -40,7 +42,7 @@ for sym in 1 2 3; do
--lang-dir $repo/data/lang_char \
$repo/test_wavs/BAC009S0764W0121.wav \
$repo/test_wavs/BAC009S0764W0122.wav \
$rep/test_wavs/BAC009S0764W0123.wav
$repo/test_wavs/BAC009S0764W0123.wav
done
for method in modified_beam_search beam_search fast_beam_search; do
@ -53,7 +55,7 @@ for method in modified_beam_search beam_search fast_beam_search; do
--lang-dir $repo/data/lang_char \
$repo/test_wavs/BAC009S0764W0121.wav \
$repo/test_wavs/BAC009S0764W0122.wav \
$rep/test_wavs/BAC009S0764W0123.wav
$repo/test_wavs/BAC009S0764W0123.wav
done
echo "GITHUB_EVENT_NAME: ${GITHUB_EVENT_NAME}"

2
.github/scripts/run-gigaspeech-pruned-transducer-stateless2-2022-05-12.sh vendored
View File

@ -1,5 +1,7 @@
#!/usr/bin/env bash
set -e
log() {
# This function is from espnet
local fname=${BASH_SOURCE[1]##*/}

123
.github/scripts/run-librispeech-conformer-ctc3-2022-11-28.sh vendored Executable file
View File

@ -0,0 +1,123 @@
#!/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/
soxi $repo/test_wavs/*.wav
ls -lh $repo/test_wavs/*.wav
pushd $repo/exp
git lfs pull --include "data/lang_bpe_500/HLG.pt"
git lfs pull --include "data/lang_bpe_500/L.pt"
git lfs pull --include "data/lang_bpe_500/LG.pt"
git lfs pull --include "data/lang_bpe_500/Linv.pt"
git lfs pull --include "data/lang_bpe_500/bpe.model"
git lfs pull --include "data/lm/G_4_gram.pt"
git lfs pull --include "exp/jit_trace.pt"
git lfs pull --include "exp/pretrained.pt"
ln -s pretrained.pt epoch-99.pt
ls -lh *.pt
popd
log "Decode with models exported by torch.jit.trace()"
for m in ctc-decoding 1best; do
./conformer_ctc3/jit_pretrained.py \
--model-filename $repo/exp/jit_trace.pt \
--words-file $repo/data/lang_bpe_500/words.txt \
--HLG $repo/data/lang_bpe_500/HLG.pt \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
--G $repo/data/lm/G_4_gram.pt \
--method $m \
--sample-rate 16000 \
$repo/test_wavs/1089-134686-0001.wav \
$repo/test_wavs/1221-135766-0001.wav \
$repo/test_wavs/1221-135766-0002.wav
done
log "Export to torchscript model"
./conformer_ctc3/export.py \
--exp-dir $repo/exp \
--lang-dir $repo/data/lang_bpe_500 \
--jit-trace 1 \
--epoch 99 \
--avg 1 \
--use-averaged-model 0
ls -lh $repo/exp/*.pt
log "Decode with models exported by torch.jit.trace()"
for m in ctc-decoding 1best; do
./conformer_ctc3/jit_pretrained.py \
--model-filename $repo/exp/jit_trace.pt \
--words-file $repo/data/lang_bpe_500/words.txt \
--HLG $repo/data/lang_bpe_500/HLG.pt \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
--G $repo/data/lm/G_4_gram.pt \
--method $m \
--sample-rate 16000 \
$repo/test_wavs/1089-134686-0001.wav \
$repo/test_wavs/1221-135766-0001.wav \
$repo/test_wavs/1221-135766-0002.wav
done
for m in ctc-decoding 1best; do
./conformer_ctc3/pretrained.py \
--checkpoint $repo/exp/pretrained.pt \
--words-file $repo/data/lang_bpe_500/words.txt \
--HLG $repo/data/lang_bpe_500/HLG.pt \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
--G $repo/data/lm/G_4_gram.pt \
--method $m \
--sample-rate 16000 \
$repo/test_wavs/1089-134686-0001.wav \
$repo/test_wavs/1221-135766-0001.wav \
$repo/test_wavs/1221-135766-0002.wav
done
echo "GITHUB_EVENT_NAME: ${GITHUB_EVENT_NAME}"
echo "GITHUB_EVENT_LABEL_NAME: ${GITHUB_EVENT_LABEL_NAME}"
if [[ x"${GITHUB_EVENT_NAME}" == x"schedule" || x"${GITHUB_EVENT_LABEL_NAME}" == x"run-decode" ]]; then
mkdir -p conformer_ctc3/exp
ln -s $PWD/$repo/exp/pretrained.pt conformer_ctc3/exp/epoch-999.pt
ln -s $PWD/$repo/data/lang_bpe_500 data/
ls -lh data
ls -lh conformer_ctc3/exp
log "Decoding test-clean and test-other"
# use a small value for decoding with CPU
max_duration=100
for method in ctc-decoding 1best; do
log "Decoding with $method"
./conformer_ctc3/decode.py \
--epoch 999 \
--avg 1 \
--use-averaged-model 0 \
--exp-dir conformer_ctc3/exp/ \
--max-duration $max_duration \
--decoding-method $method \
--lm-dir data/lm
done
rm conformer_ctc3/exp/*.pt
fi

79
.github/scripts/run-librispeech-conv-emformer-transducer-stateless2-2022-12-05.sh vendored Executable file
View File

@ -0,0 +1,79 @@
#!/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-conv-emformer-transducer-stateless2-2022-07-05
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-epoch-30-avg-10-averaged.pt"
git lfs pull --include "data/lang_bpe_500/bpe.model"
cd exp
ln -s pretrained-epoch-30-avg-10-averaged.pt epoch-99.pt
popd
log "Display test files"
tree $repo/
soxi $repo/test_wavs/*.wav
ls -lh $repo/test_wavs/*.wav
log "Install ncnn and pnnx"
# We are using a modified ncnn here. Will try to merge it to the official repo
# of ncnn
git clone https://github.com/csukuangfj/ncnn
pushd ncnn
git submodule init
git submodule update python/pybind11
python3 setup.py bdist_wheel
ls -lh dist/
pip install dist/*.whl
cd tools/pnnx
mkdir build
cd build
cmake -D Python3_EXECUTABLE=/opt/hostedtoolcache/Python/3.8.14/x64/bin/python3 ..
make -j4 pnnx
./src/pnnx || echo "pass"
popd
log "Test exporting to pnnx format"
./conv_emformer_transducer_stateless2/export-for-ncnn.py \
--exp-dir $repo/exp \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
--epoch 99 \
--avg 1 \
--use-averaged-model 0 \
\
--num-encoder-layers 12 \
--chunk-length 32 \
--cnn-module-kernel 31 \
--left-context-length 32 \
--right-context-length 8 \
--memory-size 32
./ncnn/tools/pnnx/build/src/pnnx $repo/exp/encoder_jit_trace-pnnx.pt
./ncnn/tools/pnnx/build/src/pnnx $repo/exp/decoder_jit_trace-pnnx.pt
./ncnn/tools/pnnx/build/src/pnnx $repo/exp/joiner_jit_trace-pnnx.pt
./conv_emformer_transducer_stateless2/streaming-ncnn-decode.py \
--tokens $repo/data/lang_bpe_500/tokens.txt \
--encoder-param-filename $repo/exp/encoder_jit_trace-pnnx.ncnn.param \
--encoder-bin-filename $repo/exp/encoder_jit_trace-pnnx.ncnn.bin \
--decoder-param-filename $repo/exp/decoder_jit_trace-pnnx.ncnn.param \
--decoder-bin-filename $repo/exp/decoder_jit_trace-pnnx.ncnn.bin \
--joiner-param-filename $repo/exp/joiner_jit_trace-pnnx.ncnn.param \
--joiner-bin-filename $repo/exp/joiner_jit_trace-pnnx.ncnn.bin \
$repo/test_wavs/1089-134686-0001.wav

128
.github/scripts/run-librispeech-lstm-transducer-stateless2-2022-09-03.yml → .github/scripts/run-librispeech-lstm-transducer-stateless2-2022-09-03.sh vendored
View File

@ -1,4 +1,6 @@
#!/usr/bin/env bash
#
set -e
log() {
# This function is from espnet
@ -14,6 +16,7 @@ log "Downloading pre-trained model from $repo_url"
git lfs install
git clone $repo_url
repo=$(basename $repo_url)
abs_repo=$(realpath $repo)
log "Display test files"
tree $repo/
@ -103,6 +106,47 @@ log "Decode with models exported by torch.jit.trace()"
$repo/test_wavs/1221-135766-0001.wav \
$repo/test_wavs/1221-135766-0002.wav
log "Test exporting to ONNX"
./lstm_transducer_stateless2/export.py \
--exp-dir $repo/exp \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
--epoch 99 \
--avg 1 \
--use-averaged-model 0 \
--onnx 1
log "Decode with ONNX models "
./lstm_transducer_stateless2/streaming-onnx-decode.py \
--bpe-model-filename $repo/data/lang_bpe_500/bpe.model \
--encoder-model-filename $repo//exp/encoder.onnx \
--decoder-model-filename $repo/exp/decoder.onnx \
--joiner-model-filename $repo/exp/joiner.onnx \
--joiner-encoder-proj-model-filename $repo/exp/joiner_encoder_proj.onnx \
--joiner-decoder-proj-model-filename $repo/exp/joiner_decoder_proj.onnx \
$repo/test_wavs/1089-134686-0001.wav
./lstm_transducer_stateless2/streaming-onnx-decode.py \
--bpe-model-filename $repo/data/lang_bpe_500/bpe.model \
--encoder-model-filename $repo//exp/encoder.onnx \
--decoder-model-filename $repo/exp/decoder.onnx \
--joiner-model-filename $repo/exp/joiner.onnx \
--joiner-encoder-proj-model-filename $repo/exp/joiner_encoder_proj.onnx \
--joiner-decoder-proj-model-filename $repo/exp/joiner_decoder_proj.onnx \
$repo/test_wavs/1221-135766-0001.wav
./lstm_transducer_stateless2/streaming-onnx-decode.py \
--bpe-model-filename $repo/data/lang_bpe_500/bpe.model \
--encoder-model-filename $repo//exp/encoder.onnx \
--decoder-model-filename $repo/exp/decoder.onnx \
--joiner-model-filename $repo/exp/joiner.onnx \
--joiner-encoder-proj-model-filename $repo/exp/joiner_encoder_proj.onnx \
--joiner-decoder-proj-model-filename $repo/exp/joiner_decoder_proj.onnx \
$repo/test_wavs/1221-135766-0002.wav
for sym in 1 2 3; do
log "Greedy search with --max-sym-per-frame $sym"
@ -131,7 +175,89 @@ 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"${GITHUB_EVENT_LABEL_NAME}" == x"ncnn" ]]; then
if [[ x"${GITHUB_EVENT_LABEL_NAME}" == x"shallow-fusion" ]]; then
lm_repo_url=https://huggingface.co/ezerhouni/icefall-librispeech-rnn-lm
log "Download pre-trained RNN-LM model from ${lm_repo_url}"
GIT_LFS_SKIP_SMUDGE=1 git clone $lm_repo_url
lm_repo=$(basename $lm_repo_url)
pushd $lm_repo
git lfs pull --include "exp/pretrained.pt"
mv exp/pretrained.pt exp/epoch-88.pt
popd
mkdir -p lstm_transducer_stateless2/exp
ln -sf $PWD/$repo/exp/pretrained.pt lstm_transducer_stateless2/exp/epoch-999.pt
ln -s $PWD/$repo/data/lang_bpe_500 data/
ls -lh data
ls -lh lstm_transducer_stateless2/exp
log "Decoding test-clean and test-other"
./lstm_transducer_stateless2/decode.py \
--use-averaged-model 0 \
--epoch 999 \
--avg 1 \
--exp-dir lstm_transducer_stateless2/exp \
--max-duration 600 \
--decoding-method modified_beam_search_rnnlm_shallow_fusion \
--beam 4 \
--rnn-lm-scale 0.3 \
--rnn-lm-exp-dir $lm_repo/exp \
--rnn-lm-epoch 88 \
--rnn-lm-avg 1 \
--rnn-lm-num-layers 3 \
--rnn-lm-tie-weights 1
fi
if [[ x"${GITHUB_EVENT_LABEL_NAME}" == x"LODR" ]]; then
bigram_repo_url=https://huggingface.co/marcoyang/librispeech_bigram
log "Download bi-gram LM from ${bigram_repo_url}"
GIT_LFS_SKIP_SMUDGE=1 git clone $bigram_repo_url
bigramlm_repo=$(basename $bigram_repo_url)
pushd $bigramlm_repo
git lfs pull --include "2gram.fst.txt"
cp 2gram.fst.txt $abs_repo/data/lang_bpe_500/.
popd
lm_repo_url=https://huggingface.co/ezerhouni/icefall-librispeech-rnn-lm
log "Download pre-trained RNN-LM model from ${lm_repo_url}"
GIT_LFS_SKIP_SMUDGE=1 git clone $lm_repo_url
lm_repo=$(basename $lm_repo_url)
pushd $lm_repo
git lfs pull --include "exp/pretrained.pt"
mv exp/pretrained.pt exp/epoch-88.pt
popd
mkdir -p lstm_transducer_stateless2/exp
ln -sf $PWD/$repo/exp/pretrained.pt lstm_transducer_stateless2/exp/epoch-999.pt
ln -s $PWD/$repo/data/lang_bpe_500 data/
ls -lh data
ls -lh lstm_transducer_stateless2/exp
log "Decoding test-clean and test-other"
./lstm_transducer_stateless2/decode.py \
--use-averaged-model 0 \
--epoch 999 \
--avg 1 \
--exp-dir lstm_transducer_stateless2/exp \
--max-duration 600 \
--decoding-method modified_beam_search_rnnlm_LODR \
--beam 4 \
--rnn-lm-scale 0.3 \
--rnn-lm-exp-dir $lm_repo/exp \
--rnn-lm-epoch 88 \
--rnn-lm-avg 1 \
--rnn-lm-num-layers 3 \
--rnn-lm-tie-weights 1 \
--tokens-ngram 2 \
--ngram-lm-scale -0.16
fi
if [[ x"${GITHUB_EVENT_NAME}" == x"schedule" ]]; then
mkdir -p lstm_transducer_stateless2/exp
ln -s $PWD/$repo/exp/pretrained.pt lstm_transducer_stateless2/exp/epoch-999.pt
ln -s $PWD/$repo/data/lang_bpe_500 data/

2
.github/scripts/run-librispeech-pruned-transducer-stateless-2022-03-12.sh vendored
View File

@ -1,5 +1,7 @@
#!/usr/bin/env bash
set -e
log() {
# This function is from espnet
local fname=${BASH_SOURCE[1]##*/}

11
.github/scripts/run-librispeech-pruned-transducer-stateless2-2022-04-29.sh vendored
View File

@ -1,5 +1,7 @@
#!/usr/bin/env bash
set -e
log() {
# This function is from espnet
local fname=${BASH_SOURCE[1]##*/}
@ -11,10 +13,14 @@ 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 install
git clone $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/
soxi $repo/test_wavs/*.wav
@ -77,4 +83,5 @@ if [[ x"${GITHUB_EVENT_NAME}" == x"schedule" || x"${GITHUB_EVENT_LABEL_NAME}" ==
done
rm pruned_transducer_stateless2/exp/*.pt
rm -r data/lang_bpe_500
fi

10
.github/scripts/run-librispeech-pruned-transducer-stateless3-2022-04-29.sh vendored
View File

@ -1,5 +1,7 @@
#!/usr/bin/env bash
set -e
log() {
# This function is from espnet
local fname=${BASH_SOURCE[1]##*/}
@ -11,9 +13,12 @@ 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 install
git clone $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 $repo/
@ -77,4 +82,5 @@ if [[ x"${GITHUB_EVENT_NAME}" == x"schedule" || x"${GITHUB_EVENT_LABEL_NAME}" ==
done
rm pruned_transducer_stateless3/exp/*.pt
rm -r data/lang_bpe_500
fi

12
.github/scripts/run-librispeech-pruned-transducer-stateless3-2022-05-13.sh vendored
View File

@ -1,5 +1,7 @@
#!/usr/bin/env bash
set -e
log() {
# This function is from espnet
local fname=${BASH_SOURCE[1]##*/}
@ -58,17 +60,17 @@ log "Decode with ONNX models"
--jit-filename $repo/exp/cpu_jit.pt \
--onnx-encoder-filename $repo/exp/encoder.onnx \
--onnx-decoder-filename $repo/exp/decoder.onnx \
--onnx-joiner-filename $repo/exp/joiner.onnx
./pruned_transducer_stateless3/onnx_check_all_in_one.py \
--jit-filename $repo/exp/cpu_jit.pt \
--onnx-all-in-one-filename $repo/exp/all_in_one.onnx
--onnx-joiner-filename $repo/exp/joiner.onnx \
--onnx-joiner-encoder-proj-filename $repo/exp/joiner_encoder_proj.onnx \
--onnx-joiner-decoder-proj-filename $repo/exp/joiner_decoder_proj.onnx
./pruned_transducer_stateless3/onnx_pretrained.py \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
--encoder-model-filename $repo/exp/encoder.onnx \
--decoder-model-filename $repo/exp/decoder.onnx \
--joiner-model-filename $repo/exp/joiner.onnx \
--joiner-encoder-proj-model-filename $repo/exp/joiner_encoder_proj.onnx \
--joiner-decoder-proj-model-filename $repo/exp/joiner_decoder_proj.onnx \
$repo/test_wavs/1089-134686-0001.wav \
$repo/test_wavs/1221-135766-0001.wav \
$repo/test_wavs/1221-135766-0002.wav

2
.github/scripts/run-librispeech-pruned-transducer-stateless5-2022-05-13.sh vendored
View File

@ -1,5 +1,7 @@
#!/usr/bin/env bash
set -e
log() {
# This function is from espnet
local fname=${BASH_SOURCE[1]##*/}

107
.github/scripts/run-librispeech-pruned-transducer-stateless7-2022-11-11.sh vendored Executable file
View File

@ -0,0 +1,107 @@
#!/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/
soxi $repo/test_wavs/*.wav
ls -lh $repo/test_wavs/*.wav
pushd $repo/exp
git lfs pull --include "data/lang_bpe_500/bpe.model"
git lfs pull --include "exp/cpu_jit.pt"
git lfs pull --include "exp/pretrained.pt"
ln -s pretrained.pt epoch-99.pt
ls -lh *.pt
popd
log "Export to torchscript model"
./pruned_transducer_stateless7/export.py \
--exp-dir $repo/exp \
--use-averaged-model false \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
--epoch 99 \
--avg 1 \
--jit 1
ls -lh $repo/exp/*.pt
log "Decode with models exported by torch.jit.script()"
./pruned_transducer_stateless7/jit_pretrained.py \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
--nn-model-filename $repo/exp/cpu_jit.pt \
$repo/test_wavs/1089-134686-0001.wav \
$repo/test_wavs/1221-135766-0001.wav \
$repo/test_wavs/1221-135766-0002.wav
for sym in 1 2 3; do
log "Greedy search with --max-sym-per-frame $sym"
./pruned_transducer_stateless7/pretrained.py \
--method greedy_search \
--max-sym-per-frame $sym \
--checkpoint $repo/exp/pretrained.pt \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
$repo/test_wavs/1089-134686-0001.wav \
$repo/test_wavs/1221-135766-0001.wav \
$repo/test_wavs/1221-135766-0002.wav
done
for method in modified_beam_search beam_search fast_beam_search; do
log "$method"
./pruned_transducer_stateless7/pretrained.py \
--method $method \
--beam-size 4 \
--checkpoint $repo/exp/pretrained.pt \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
$repo/test_wavs/1089-134686-0001.wav \
$repo/test_wavs/1221-135766-0001.wav \
$repo/test_wavs/1221-135766-0002.wav
done
echo "GITHUB_EVENT_NAME: ${GITHUB_EVENT_NAME}"
echo "GITHUB_EVENT_LABEL_NAME: ${GITHUB_EVENT_LABEL_NAME}"
if [[ x"${GITHUB_EVENT_NAME}" == x"schedule" || x"${GITHUB_EVENT_LABEL_NAME}" == x"run-decode" ]]; then
mkdir -p pruned_transducer_stateless7/exp
ln -s $PWD/$repo/exp/pretrained.pt pruned_transducer_stateless7/exp/epoch-999.pt
ln -s $PWD/$repo/data/lang_bpe_500 data/
ls -lh data
ls -lh pruned_transducer_stateless7/exp
log "Decoding test-clean and test-other"
# use a small value for decoding with CPU
max_duration=100
for method in greedy_search fast_beam_search modified_beam_search; do
log "Decoding with $method"
./pruned_transducer_stateless7/decode.py \
--decoding-method $method \
--epoch 999 \
--avg 1 \
--use-averaged-model 0 \
--max-duration $max_duration \
--exp-dir pruned_transducer_stateless7/exp
done
rm pruned_transducer_stateless7/exp/*.pt
fi

151
.github/scripts/run-librispeech-pruned-transducer-stateless7-ctc-2022-12-01.sh vendored Executable file
View File

@ -0,0 +1,151 @@
#!/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/
soxi $repo/test_wavs/*.wav
ls -lh $repo/test_wavs/*.wav
pushd $repo/exp
git lfs pull --include "data/lang_bpe_500/HLG.pt"
git lfs pull --include "data/lang_bpe_500/L.pt"
git lfs pull --include "data/lang_bpe_500/LG.pt"
git lfs pull --include "data/lang_bpe_500/Linv.pt"
git lfs pull --include "data/lang_bpe_500/bpe.model"
git lfs pull --include "data/lm/G_4_gram.pt"
git lfs pull --include "exp/cpu_jit.pt"
git lfs pull --include "exp/pretrained.pt"
ln -s pretrained.pt epoch-99.pt
ls -lh *.pt
popd
log "Export to torchscript model"
./pruned_transducer_stateless7_ctc/export.py \
--exp-dir $repo/exp \
--use-averaged-model false \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
--epoch 99 \
--avg 1 \
--jit 1
ls -lh $repo/exp/*.pt
log "Decode with models exported by torch.jit.script()"
./pruned_transducer_stateless7_ctc/jit_pretrained.py \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
--nn-model-filename $repo/exp/cpu_jit.pt \
$repo/test_wavs/1089-134686-0001.wav \
$repo/test_wavs/1221-135766-0001.wav \
$repo/test_wavs/1221-135766-0002.wav
for m in ctc-decoding 1best; do
./pruned_transducer_stateless7_ctc/jit_pretrained_ctc.py \
--model-filename $repo/exp/cpu_jit.pt \
--words-file $repo/data/lang_bpe_500/words.txt \
--HLG $repo/data/lang_bpe_500/HLG.pt \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
--G $repo/data/lm/G_4_gram.pt \
--method $m \
--sample-rate 16000 \
$repo/test_wavs/1089-134686-0001.wav \
$repo/test_wavs/1221-135766-0001.wav \
$repo/test_wavs/1221-135766-0002.wav
done
for sym in 1 2 3; do
log "Greedy search with --max-sym-per-frame $sym"
./pruned_transducer_stateless7_ctc/pretrained.py \
--method greedy_search \
--max-sym-per-frame $sym \
--checkpoint $repo/exp/pretrained.pt \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
$repo/test_wavs/1089-134686-0001.wav \
$repo/test_wavs/1221-135766-0001.wav \
$repo/test_wavs/1221-135766-0002.wav
done
for method in modified_beam_search beam_search fast_beam_search; do
log "$method"
./pruned_transducer_stateless7_ctc/pretrained.py \
--method $method \
--beam-size 4 \
--checkpoint $repo/exp/pretrained.pt \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
$repo/test_wavs/1089-134686-0001.wav \
$repo/test_wavs/1221-135766-0001.wav \
$repo/test_wavs/1221-135766-0002.wav
done
for m in ctc-decoding 1best; do
./pruned_transducer_stateless7_ctc/pretrained_ctc.py \
--checkpoint $repo/exp/pretrained.pt \
--words-file $repo/data/lang_bpe_500/words.txt \
--HLG $repo/data/lang_bpe_500/HLG.pt \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
--G $repo/data/lm/G_4_gram.pt \
--method $m \
--sample-rate 16000 \
$repo/test_wavs/1089-134686-0001.wav \
$repo/test_wavs/1221-135766-0001.wav \
$repo/test_wavs/1221-135766-0002.wav
done
echo "GITHUB_EVENT_NAME: ${GITHUB_EVENT_NAME}"
echo "GITHUB_EVENT_LABEL_NAME: ${GITHUB_EVENT_LABEL_NAME}"
if [[ x"${GITHUB_EVENT_NAME}" == x"schedule" || x"${GITHUB_EVENT_LABEL_NAME}" == x"run-decode" ]]; then
mkdir -p pruned_transducer_stateless7_ctc/exp
ln -s $PWD/$repo/exp/pretrained.pt pruned_transducer_stateless7_ctc/exp/epoch-999.pt
ln -s $PWD/$repo/data/lang_bpe_500 data/
ls -lh data
ls -lh pruned_transducer_stateless7_ctc/exp
log "Decoding test-clean and test-other"
# use a small value for decoding with CPU
max_duration=100
for method in greedy_search fast_beam_search modified_beam_search; do
log "Decoding with $method"
./pruned_transducer_stateless7_ctc/decode.py \
--decoding-method $method \
--epoch 999 \
--avg 1 \
--use-averaged-model 0 \
--max-duration $max_duration \
--exp-dir pruned_transducer_stateless7_ctc/exp
done
for m in ctc-decoding 1best; do
./pruned_transducer_stateless7_ctc/ctc_decode.py \
--epoch 999 \
--avg 1 \
--exp-dir ./pruned_transducer_stateless7_ctc/exp \
--max-duration $max_duration \
--use-averaged-model 0 \
--decoding-method $m \
--hlg-scale 0.6 \
--lm-dir data/lm
done
rm pruned_transducer_stateless7_ctc/exp/*.pt
fi

116
.github/scripts/run-librispeech-pruned-transducer-stateless8-2022-11-14.sh vendored Executable file
View File

@ -0,0 +1,116 @@
#!/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/
soxi $repo/test_wavs/*.wav
ls -lh $repo/test_wavs/*.wav
pushd $repo/exp
git lfs pull --include "data/lang_bpe_500/bpe.model"
git lfs pull --include "exp/cpu_jit.pt"
git lfs pull --include "exp/pretrained.pt"
ln -s pretrained.pt epoch-99.pt
ls -lh *.pt
popd
log "Decode with models exported by torch.jit.script()"
./pruned_transducer_stateless8/jit_pretrained.py \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
--nn-model-filename $repo/exp/cpu_jit.pt \
$repo/test_wavs/1089-134686-0001.wav \
$repo/test_wavs/1221-135766-0001.wav \
$repo/test_wavs/1221-135766-0002.wav
log "Export to torchscript model"
./pruned_transducer_stateless8/export.py \
--exp-dir $repo/exp \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
--use-averaged-model false \
--epoch 99 \
--avg 1 \
--jit 1
ls -lh $repo/exp/*.pt
log "Decode with models exported by torch.jit.script()"
./pruned_transducer_stateless8/jit_pretrained.py \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
--nn-model-filename $repo/exp/cpu_jit.pt \
$repo/test_wavs/1089-134686-0001.wav \
$repo/test_wavs/1221-135766-0001.wav \
$repo/test_wavs/1221-135766-0002.wav
for sym in 1 2 3; do
log "Greedy search with --max-sym-per-frame $sym"
./pruned_transducer_stateless8/pretrained.py \
--method greedy_search \
--max-sym-per-frame $sym \
--checkpoint $repo/exp/pretrained.pt \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
$repo/test_wavs/1089-134686-0001.wav \
$repo/test_wavs/1221-135766-0001.wav \
$repo/test_wavs/1221-135766-0002.wav
done
for method in modified_beam_search beam_search fast_beam_search; do
log "$method"
./pruned_transducer_stateless8/pretrained.py \
--method $method \
--beam-size 4 \
--checkpoint $repo/exp/pretrained.pt \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
$repo/test_wavs/1089-134686-0001.wav \
$repo/test_wavs/1221-135766-0001.wav \
$repo/test_wavs/1221-135766-0002.wav
done
echo "GITHUB_EVENT_NAME: ${GITHUB_EVENT_NAME}"
echo "GITHUB_EVENT_LABEL_NAME: ${GITHUB_EVENT_LABEL_NAME}"
if [[ x"${GITHUB_EVENT_NAME}" == x"schedule" || x"${GITHUB_EVENT_LABEL_NAME}" == x"run-decode" ]]; then
mkdir -p pruned_transducer_stateless8/exp
ln -s $PWD/$repo/exp/pretrained.pt pruned_transducer_stateless8/exp/epoch-999.pt
ln -s $PWD/$repo/data/lang_bpe_500 data/
ls -lh data
ls -lh pruned_transducer_stateless8/exp
log "Decoding test-clean and test-other"
# use a small value for decoding with CPU
max_duration=100
for method in greedy_search fast_beam_search modified_beam_search; do
log "Decoding with $method"
./pruned_transducer_stateless8/decode.py \
--decoding-method $method \
--epoch 999 \
--avg 1 \
--use-averaged-model 0 \
--max-duration $max_duration \
--exp-dir pruned_transducer_stateless8/exp
done
rm pruned_transducer_stateless8/exp/*.pt
fi

2
.github/scripts/run-librispeech-streaming-pruned-transducer-stateless2-2022-06-26.sh vendored
View File

@ -1,5 +1,7 @@
#!/usr/bin/env bash
set -e
log() {
# This function is from espnet
local fname=${BASH_SOURCE[1]##*/}

2
.github/scripts/run-librispeech-transducer-stateless2-2022-04-19.sh vendored
View File

@ -1,5 +1,7 @@
#!/usr/bin/env bash
set -e
log() {
# This function is from espnet
local fname=${BASH_SOURCE[1]##*/}

103
.github/scripts/run-librispeech-zipformer-mmi-2022-12-08.sh vendored Executable file
View File

@ -0,0 +1,103 @@
#!/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/
soxi $repo/test_wavs/*.wav
ls -lh $repo/test_wavs/*.wav
pushd $repo/exp
git lfs pull --include "data/lang_bpe_500/3gram.pt"
git lfs pull --include "data/lang_bpe_500/4gram.pt"
git lfs pull --include "data/lang_bpe_500/L.pt"
git lfs pull --include "data/lang_bpe_500/LG.pt"
git lfs pull --include "data/lang_bpe_500/Linv.pt"
git lfs pull --include "data/lang_bpe_500/bpe.model"
git lfs pull --include "exp/cpu_jit.pt"
git lfs pull --include "exp/pretrained.pt"
ln -s pretrained.pt epoch-99.pt
ls -lh *.pt
popd
log "Export to torchscript model"
./zipformer_mmi/export.py \
--exp-dir $repo/exp \
--use-averaged-model false \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
--epoch 99 \
--avg 1 \
--jit 1
ls -lh $repo/exp/*.pt
log "Decode with models exported by torch.jit.script()"
./zipformer_mmi/jit_pretrained.py \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
--nn-model-filename $repo/exp/cpu_jit.pt \
--lang-dir $repo/data/lang_bpe_500 \
$repo/test_wavs/1089-134686-0001.wav \
$repo/test_wavs/1221-135766-0001.wav \
$repo/test_wavs/1221-135766-0002.wav
for method in 1best nbest nbest-rescoring-LG nbest-rescoring-3-gram nbest-rescoring-4-gram; do
log "$method"
./zipformer_mmi/pretrained.py \
--method $method \
--checkpoint $repo/exp/pretrained.pt \
--lang-dir $repo/data/lang_bpe_500 \
--bpe-model $repo/data/lang_bpe_500/bpe.model \
$repo/test_wavs/1089-134686-0001.wav \
$repo/test_wavs/1221-135766-0001.wav \
$repo/test_wavs/1221-135766-0002.wav
done
echo "GITHUB_EVENT_NAME: ${GITHUB_EVENT_NAME}"
echo "GITHUB_EVENT_LABEL_NAME: ${GITHUB_EVENT_LABEL_NAME}"
if [[ x"${GITHUB_EVENT_NAME}" == x"schedule" || x"${GITHUB_EVENT_LABEL_NAME}" == x"run-decode" ]]; then
mkdir -p zipformer_mmi/exp
ln -s $PWD/$repo/exp/pretrained.pt zipformer_mmi/exp/epoch-999.pt
ln -s $PWD/$repo/data/lang_bpe_500 data/
ls -lh data
ls -lh zipformer_mmi/exp
log "Decoding test-clean and test-other"
# use a small value for decoding with CPU
max_duration=100
for method in 1best nbest nbest-rescoring-LG nbest-rescoring-3-gram nbest-rescoring-4-gram; do
log "Decoding with $method"
./zipformer_mmi/decode.py \
--decoding-method $method \
--epoch 999 \
--avg 1 \
--use-averaged-model 0 \
--nbest-scale 1.2 \
--hp-scale 1.0 \
--max-duration $max_duration \
--lang-dir $repo/data/lang_bpe_500 \
--exp-dir zipformer_mmi/exp
done
rm zipformer_mmi/exp/*.pt
fi

3
.github/scripts/run-pre-trained-conformer-ctc.sh vendored
View File

@ -1,5 +1,7 @@
#!/usr/bin/env bash
set -e
log() {
# This function is from espnet
local fname=${BASH_SOURCE[1]##*/}
@ -10,7 +12,6 @@ cd egs/librispeech/ASR
repo_url=https://github.com/csukuangfj/icefall-asr-conformer-ctc-bpe-500
git lfs install
git clone $repo
log "Downloading pre-trained model from $repo_url"
git clone $repo_url

2
.github/scripts/run-pre-trained-transducer-stateless-librispeech-100h.sh vendored
View File

@ -1,5 +1,7 @@
#!/usr/bin/env bash
set -e
log() {
# This function is from espnet
local fname=${BASH_SOURCE[1]##*/}

2
.github/scripts/run-pre-trained-transducer-stateless-librispeech-960h.sh vendored
View File

@ -1,5 +1,7 @@
#!/usr/bin/env bash
set -e
log() {
# This function is from espnet
local fname=${BASH_SOURCE[1]##*/}

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

@ -1,5 +1,7 @@
#!/usr/bin/env bash
set -e
log() {
# This function is from espnet
local fname=${BASH_SOURCE[1]##*/}

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

@ -1,5 +1,7 @@
#!/usr/bin/env bash
set -e
log() {
# This function is from espnet
local fname=${BASH_SOURCE[1]##*/}

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

@ -1,5 +1,7 @@
#!/usr/bin/env bash
set -e
log() {
# This function is from espnet
local fname=${BASH_SOURCE[1]##*/}

2
.github/scripts/run-pre-trained-transducer.sh vendored
View File

@ -1,5 +1,7 @@
#!/usr/bin/env bash
set -e
log() {
# This function is from espnet
local fname=${BASH_SOURCE[1]##*/}

124
.github/scripts/run-wenetspeech-pruned-transducer-stateless2.sh vendored Executable file
View File

@ -0,0 +1,124 @@
#!/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/wenetspeech/ASR
repo_url=https://huggingface.co/luomingshuang/icefall_asr_wenetspeech_pruned_transducer_stateless2
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/
soxi $repo/test_wavs/*.wav
ls -lh $repo/test_wavs/*.wav
pushd $repo/exp
ln -s pretrained_epoch_10_avg_2.pt pretrained.pt
ln -s pretrained_epoch_10_avg_2.pt epoch-99.pt
popd
log "Test exporting to ONNX format"
./pruned_transducer_stateless2/export.py \
--exp-dir $repo/exp \
--lang-dir $repo/data/lang_char \
--epoch 99 \
--avg 1 \
--onnx 1
log "Export to torchscript model"
./pruned_transducer_stateless2/export.py \
--exp-dir $repo/exp \
--lang-dir $repo/data/lang_char \
--epoch 99 \
--avg 1 \
--jit 1
./pruned_transducer_stateless2/export.py \
--exp-dir $repo/exp \
--lang-dir $repo/data/lang_char \
--epoch 99 \
--avg 1 \
--jit-trace 1
ls -lh $repo/exp/*.onnx
ls -lh $repo/exp/*.pt
log "Decode with ONNX models"
./pruned_transducer_stateless2/onnx_check.py \
--jit-filename $repo/exp/cpu_jit.pt \
--onnx-encoder-filename $repo/exp/encoder.onnx \
--onnx-decoder-filename $repo/exp/decoder.onnx \
--onnx-joiner-filename $repo/exp/joiner.onnx \
--onnx-joiner-encoder-proj-filename $repo/exp/joiner_encoder_proj.onnx \
--onnx-joiner-decoder-proj-filename $repo/exp/joiner_decoder_proj.onnx
./pruned_transducer_stateless2/onnx_pretrained.py \
--tokens $repo/data/lang_char/tokens.txt \
--encoder-model-filename $repo/exp/encoder.onnx \
--decoder-model-filename $repo/exp/decoder.onnx \
--joiner-model-filename $repo/exp/joiner.onnx \
--joiner-encoder-proj-model-filename $repo/exp/joiner_encoder_proj.onnx \
--joiner-decoder-proj-model-filename $repo/exp/joiner_decoder_proj.onnx \
$repo/test_wavs/DEV_T0000000000.wav \
$repo/test_wavs/DEV_T0000000001.wav \
$repo/test_wavs/DEV_T0000000002.wav
log "Decode with models exported by torch.jit.trace()"
./pruned_transducer_stateless2/jit_pretrained.py \
--tokens $repo/data/lang_char/tokens.txt \
--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/DEV_T0000000000.wav \
$repo/test_wavs/DEV_T0000000001.wav \
$repo/test_wavs/DEV_T0000000002.wav
./pruned_transducer_stateless2/jit_pretrained.py \
--tokens $repo/data/lang_char/tokens.txt \
--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/DEV_T0000000000.wav \
$repo/test_wavs/DEV_T0000000001.wav \
$repo/test_wavs/DEV_T0000000002.wav
for sym in 1 2 3; do
log "Greedy search with --max-sym-per-frame $sym"
./pruned_transducer_stateless2/pretrained.py \
--checkpoint $repo/exp/epoch-99.pt \
--lang-dir $repo/data/lang_char \
--decoding-method greedy_search \
--max-sym-per-frame $sym \
$repo/test_wavs/DEV_T0000000000.wav \
$repo/test_wavs/DEV_T0000000001.wav \
$repo/test_wavs/DEV_T0000000002.wav
done
for method in modified_beam_search beam_search fast_beam_search; do
log "$method"
./pruned_transducer_stateless2/pretrained.py \
--decoding-method $method \
--beam-size 4 \
--checkpoint $repo/exp/epoch-99.pt \
--lang-dir $repo/data/lang_char \
$repo/test_wavs/DEV_T0000000000.wav \
$repo/test_wavs/DEV_T0000000001.wav \
$repo/test_wavs/DEV_T0000000002.wav
done

4
.github/workflows/build-doc.yml vendored
View File

@ -26,6 +26,10 @@ on:
pull_request:
types: [labeled]
concurrency:
group: build_doc-${{ github.ref }}
cancel-in-progress: true
jobs:
build-doc:
if: github.event.label.name == 'doc' || github.event_name == 'push'

4
.github/workflows/run-aishell-2022-06-20.yml vendored
View File

@ -34,6 +34,10 @@ on:
# 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'

4
.github/workflows/run-gigaspeech-2022-05-13.yml vendored
View File

@ -33,6 +33,10 @@ on:
# nightly build at 15:50 UTC time every day
- cron: "50 15 * * *"
concurrency:
group: run_gigaspeech_2022_05_13-${{ github.ref }}
cancel-in-progress: true
jobs:
run_gigaspeech_2022_05_13:
if: github.event.label.name == 'ready' || github.event.label.name == 'run-decode' || github.event_name == 'push' || github.event_name == 'schedule'

4
.github/workflows/run-librispeech-2022-03-12.yml vendored
View File

@ -33,6 +33,10 @@ on:
# 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'

4
.github/workflows/run-librispeech-2022-04-29.yml vendored
View File

@ -33,6 +33,10 @@ on:
# 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'

4
.github/workflows/run-librispeech-2022-05-13.yml vendored
View File

@ -33,6 +33,10 @@ on:
# 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'

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

@ -0,0 +1,159 @@
# 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: ${{ matrix.python-version }}
cache: 'pip'
cache-dependency-path: '**/requirements-ci.txt'
- name: Install Python dependencies
run: |
grep -v '^#' ./requirements-ci.txt | xargs -n 1 -L 1 pip install
pip uninstall -y protobuf
pip install --no-binary protobuf protobuf
- name: Cache kaldifeat
id: my-cache
uses: actions/cache@v2
with:
path: |
~/tmp/kaldifeat
key: cache-tmp-${{ matrix.python-version }}-2022-09-25
- name: Install kaldifeat
if: steps.my-cache.outputs.cache-hit != 'true'
shell: bash
run: |
.github/scripts/install-kaldifeat.sh
- name: Cache LibriSpeech test-clean and test-other datasets
id: libri-test-clean-and-test-other-data
uses: actions/cache@v2
with:
path: |
~/tmp/download
key: cache-libri-test-clean-and-test-other
- name: Download LibriSpeech test-clean and test-other
if: steps.libri-test-clean-and-test-other-data.outputs.cache-hit != 'true'
shell: bash
run: |
.github/scripts/download-librispeech-test-clean-and-test-other-dataset.sh
- name: Prepare manifests for LibriSpeech test-clean and test-other
shell: bash
run: |
.github/scripts/prepare-librispeech-test-clean-and-test-other-manifests.sh
- name: Cache LibriSpeech test-clean and test-other fbank features
id: libri-test-clean-and-test-other-fbank
uses: actions/cache@v2
with:
path: |
~/tmp/fbank-libri
key: cache-libri-fbank-test-clean-and-test-other-v2
- name: Compute fbank for LibriSpeech test-clean and test-other
if: steps.libri-test-clean-and-test-other-fbank.outputs.cache-hit != 'true'
shell: bash
run: |
.github/scripts/compute-fbank-librispeech-test-clean-and-test-other.sh
- name: Inference with pre-trained model
shell: bash
env:
GITHUB_EVENT_NAME: ${{ github.event_name }}
GITHUB_EVENT_LABEL_NAME: ${{ github.event.label.name }}
run: |
mkdir -p egs/librispeech/ASR/data
ln -sfv ~/tmp/fbank-libri egs/librispeech/ASR/data/fbank
ls -lh egs/librispeech/ASR/data/*
sudo apt-get -qq install git-lfs tree sox
export PYTHONPATH=$PWD:$PYTHONPATH
export PYTHONPATH=~/tmp/kaldifeat/kaldifeat/python:$PYTHONPATH
export PYTHONPATH=~/tmp/kaldifeat/build/lib:$PYTHONPATH
.github/scripts/run-librispeech-pruned-transducer-stateless7-2022-11-11.sh
- name: Display decoding results for librispeech pruned_transducer_stateless7
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
shell: bash
run: |
cd egs/librispeech/ASR/
tree ./pruned_transducer_stateless7/exp
cd pruned_transducer_stateless7
echo "results for pruned_transducer_stateless7"
echo "===greedy search==="
find exp/greedy_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/greedy_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
echo "===fast_beam_search==="
find exp/fast_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/fast_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
echo "===modified beam search==="
find exp/modified_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/modified_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
- name: Upload decoding results for librispeech pruned_transducer_stateless7
uses: actions/upload-artifact@v2
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
with:
name: torch-${{ matrix.torch }}-python-${{ matrix.python-version }}-ubuntu-18.04-cpu-pruned_transducer_stateless7-2022-11-11
path: egs/librispeech/ASR/pruned_transducer_stateless7/exp/

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

@ -0,0 +1,159 @@
# 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: ${{ matrix.python-version }}
cache: 'pip'
cache-dependency-path: '**/requirements-ci.txt'
- name: Install Python dependencies
run: |
grep -v '^#' ./requirements-ci.txt | xargs -n 1 -L 1 pip install
pip uninstall -y protobuf
pip install --no-binary protobuf protobuf
- name: Cache kaldifeat
id: my-cache
uses: actions/cache@v2
with:
path: |
~/tmp/kaldifeat
key: cache-tmp-${{ matrix.python-version }}-2022-09-25
- name: Install kaldifeat
if: steps.my-cache.outputs.cache-hit != 'true'
shell: bash
run: |
.github/scripts/install-kaldifeat.sh
- name: Cache LibriSpeech test-clean and test-other datasets
id: libri-test-clean-and-test-other-data
uses: actions/cache@v2
with:
path: |
~/tmp/download
key: cache-libri-test-clean-and-test-other
- name: Download LibriSpeech test-clean and test-other
if: steps.libri-test-clean-and-test-other-data.outputs.cache-hit != 'true'
shell: bash
run: |
.github/scripts/download-librispeech-test-clean-and-test-other-dataset.sh
- name: Prepare manifests for LibriSpeech test-clean and test-other
shell: bash
run: |
.github/scripts/prepare-librispeech-test-clean-and-test-other-manifests.sh
- name: Cache LibriSpeech test-clean and test-other fbank features
id: libri-test-clean-and-test-other-fbank
uses: actions/cache@v2
with:
path: |
~/tmp/fbank-libri
key: cache-libri-fbank-test-clean-and-test-other-v2
- name: Compute fbank for LibriSpeech test-clean and test-other
if: steps.libri-test-clean-and-test-other-fbank.outputs.cache-hit != 'true'
shell: bash
run: |
.github/scripts/compute-fbank-librispeech-test-clean-and-test-other.sh
- name: Inference with pre-trained model
shell: bash
env:
GITHUB_EVENT_NAME: ${{ github.event_name }}
GITHUB_EVENT_LABEL_NAME: ${{ github.event.label.name }}
run: |
mkdir -p egs/librispeech/ASR/data
ln -sfv ~/tmp/fbank-libri egs/librispeech/ASR/data/fbank
ls -lh egs/librispeech/ASR/data/*
sudo apt-get -qq install git-lfs tree sox
export PYTHONPATH=$PWD:$PYTHONPATH
export PYTHONPATH=~/tmp/kaldifeat/kaldifeat/python:$PYTHONPATH
export PYTHONPATH=~/tmp/kaldifeat/build/lib:$PYTHONPATH
.github/scripts/run-librispeech-pruned-transducer-stateless8-2022-11-14.sh
- name: Display decoding results for librispeech pruned_transducer_stateless8
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
shell: bash
run: |
cd egs/librispeech/ASR/
tree ./pruned_transducer_stateless8/exp
cd pruned_transducer_stateless8
echo "results for pruned_transducer_stateless8"
echo "===greedy search==="
find exp/greedy_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/greedy_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
echo "===fast_beam_search==="
find exp/fast_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/fast_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
echo "===modified beam search==="
find exp/modified_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/modified_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
- name: Upload decoding results for librispeech pruned_transducer_stateless8
uses: actions/upload-artifact@v2
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
with:
name: torch-${{ matrix.torch }}-python-${{ matrix.python-version }}-ubuntu-18.04-cpu-pruned_transducer_stateless8-2022-11-14
path: egs/librispeech/ASR/pruned_transducer_stateless8/exp/

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

@ -0,0 +1,163 @@
# 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: ${{ matrix.python-version }}
cache: 'pip'
cache-dependency-path: '**/requirements-ci.txt'
- name: Install Python dependencies
run: |
grep -v '^#' ./requirements-ci.txt | xargs -n 1 -L 1 pip install
pip uninstall -y protobuf
pip install --no-binary protobuf protobuf
- name: Cache kaldifeat
id: my-cache
uses: actions/cache@v2
with:
path: |
~/tmp/kaldifeat
key: cache-tmp-${{ matrix.python-version }}-2022-09-25
- name: Install kaldifeat
if: steps.my-cache.outputs.cache-hit != 'true'
shell: bash
run: |
.github/scripts/install-kaldifeat.sh
- name: Cache LibriSpeech test-clean and test-other datasets
id: libri-test-clean-and-test-other-data
uses: actions/cache@v2
with:
path: |
~/tmp/download
key: cache-libri-test-clean-and-test-other
- name: Download LibriSpeech test-clean and test-other
if: steps.libri-test-clean-and-test-other-data.outputs.cache-hit != 'true'
shell: bash
run: |
.github/scripts/download-librispeech-test-clean-and-test-other-dataset.sh
- name: Prepare manifests for LibriSpeech test-clean and test-other
shell: bash
run: |
.github/scripts/prepare-librispeech-test-clean-and-test-other-manifests.sh
- name: Cache LibriSpeech test-clean and test-other fbank features
id: libri-test-clean-and-test-other-fbank
uses: actions/cache@v2
with:
path: |
~/tmp/fbank-libri
key: cache-libri-fbank-test-clean-and-test-other-v2
- name: Compute fbank for LibriSpeech test-clean and test-other
if: steps.libri-test-clean-and-test-other-fbank.outputs.cache-hit != 'true'
shell: bash
run: |
.github/scripts/compute-fbank-librispeech-test-clean-and-test-other.sh
- name: Inference with pre-trained model
shell: bash
env:
GITHUB_EVENT_NAME: ${{ github.event_name }}
GITHUB_EVENT_LABEL_NAME: ${{ github.event.label.name }}
run: |
mkdir -p egs/librispeech/ASR/data
ln -sfv ~/tmp/fbank-libri egs/librispeech/ASR/data/fbank
ls -lh egs/librispeech/ASR/data/*
sudo apt-get -qq install git-lfs tree sox
export PYTHONPATH=$PWD:$PYTHONPATH
export PYTHONPATH=~/tmp/kaldifeat/kaldifeat/python:$PYTHONPATH
export PYTHONPATH=~/tmp/kaldifeat/build/lib:$PYTHONPATH
.github/scripts/run-librispeech-pruned-transducer-stateless7-ctc-2022-12-01.sh
- name: Display decoding results for librispeech pruned_transducer_stateless7_ctc
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
shell: bash
run: |
cd egs/librispeech/ASR/
tree ./pruned_transducer_stateless7_ctc/exp
cd pruned_transducer_stateless7_ctc
echo "results for pruned_transducer_stateless7_ctc"
echo "===greedy search==="
find exp/greedy_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/greedy_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
echo "===fast_beam_search==="
find exp/fast_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/fast_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
echo "===modified beam search==="
find exp/modified_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/modified_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
echo "===ctc decoding==="
find exp/ctc-decoding -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/ctc-decoding -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
echo "===1best==="
find exp/1best -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/1best -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
- name: Upload decoding results for librispeech pruned_transducer_stateless7_ctc
uses: actions/upload-artifact@v2
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
with:
name: torch-${{ matrix.torch }}-python-${{ matrix.python-version }}-ubuntu-18.04-cpu-pruned_transducer_stateless7-ctc-2022-12-01
path: egs/librispeech/ASR/pruned_transducer_stateless7_ctc/exp/

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

@ -0,0 +1,167 @@
# 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: ${{ matrix.python-version }}
cache: 'pip'
cache-dependency-path: '**/requirements-ci.txt'
- name: Install Python dependencies
run: |
grep -v '^#' ./requirements-ci.txt | xargs -n 1 -L 1 pip install
pip uninstall -y protobuf
pip install --no-binary protobuf protobuf
- name: Cache kaldifeat
id: my-cache
uses: actions/cache@v2
with:
path: |
~/tmp/kaldifeat
key: cache-tmp-${{ matrix.python-version }}-2022-09-25
- name: Install kaldifeat
if: steps.my-cache.outputs.cache-hit != 'true'
shell: bash
run: |
.github/scripts/install-kaldifeat.sh
- name: Cache LibriSpeech test-clean and test-other datasets
id: libri-test-clean-and-test-other-data
uses: actions/cache@v2
with:
path: |
~/tmp/download
key: cache-libri-test-clean-and-test-other
- name: Download LibriSpeech test-clean and test-other
if: steps.libri-test-clean-and-test-other-data.outputs.cache-hit != 'true'
shell: bash
run: |
.github/scripts/download-librispeech-test-clean-and-test-other-dataset.sh
- name: Prepare manifests for LibriSpeech test-clean and test-other
shell: bash
run: |
.github/scripts/prepare-librispeech-test-clean-and-test-other-manifests.sh
- name: Cache LibriSpeech test-clean and test-other fbank features
id: libri-test-clean-and-test-other-fbank
uses: actions/cache@v2
with:
path: |
~/tmp/fbank-libri
key: cache-libri-fbank-test-clean-and-test-other-v2
- name: Compute fbank for LibriSpeech test-clean and test-other
if: steps.libri-test-clean-and-test-other-fbank.outputs.cache-hit != 'true'
shell: bash
run: |
.github/scripts/compute-fbank-librispeech-test-clean-and-test-other.sh
- name: Inference with pre-trained model
shell: bash
env:
GITHUB_EVENT_NAME: ${{ github.event_name }}
GITHUB_EVENT_LABEL_NAME: ${{ github.event.label.name }}
run: |
mkdir -p egs/librispeech/ASR/data
ln -sfv ~/tmp/fbank-libri egs/librispeech/ASR/data/fbank
ls -lh egs/librispeech/ASR/data/*
sudo apt-get -qq install git-lfs tree sox
export PYTHONPATH=$PWD:$PYTHONPATH
export PYTHONPATH=~/tmp/kaldifeat/kaldifeat/python:$PYTHONPATH
export PYTHONPATH=~/tmp/kaldifeat/build/lib:$PYTHONPATH
.github/scripts/run-librispeech-zipformer-mmi-2022-12-08.sh
- name: Display decoding results for librispeech zipformer-mmi
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
shell: bash
run: |
cd egs/librispeech/ASR/
tree ./zipformer-mmi/exp
cd zipformer-mmi
echo "results for zipformer-mmi"
echo "===1best==="
find exp/1best -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/1best -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
echo "===nbest==="
find exp/nbest -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/nbest -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
echo "===nbest-rescoring-LG==="
find exp/nbest-rescoring-LG -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/nbest-rescoring-LG -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
echo "===nbest-rescoring-3-gram==="
find exp/nbest-rescoring-3-gram -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/nbest-rescoring-3-gram -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
echo "===nbest-rescoring-4-gram==="
find exp/nbest-rescoring-4-gram -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/nbest-rescoring-4-gram -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
- name: Upload decoding results for librispeech zipformer-mmi
uses: actions/upload-artifact@v2
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
with:
name: torch-${{ matrix.torch }}-python-${{ matrix.python-version }}-ubuntu-18.04-cpu-zipformer_mmi-2022-12-08
path: egs/librispeech/ASR/zipformer_mmi/exp/

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

@ -0,0 +1,155 @@
# 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: ${{ matrix.python-version }}
cache: 'pip'
cache-dependency-path: '**/requirements-ci.txt'
- name: Install Python dependencies
run: |
grep -v '^#' ./requirements-ci.txt | xargs -n 1 -L 1 pip install
pip uninstall -y protobuf
pip install --no-binary protobuf protobuf
- name: Cache kaldifeat
id: my-cache
uses: actions/cache@v2
with:
path: |
~/tmp/kaldifeat
key: cache-tmp-${{ matrix.python-version }}-2022-09-25
- name: Install kaldifeat
if: steps.my-cache.outputs.cache-hit != 'true'
shell: bash
run: |
.github/scripts/install-kaldifeat.sh
- name: Cache LibriSpeech test-clean and test-other datasets
id: libri-test-clean-and-test-other-data
uses: actions/cache@v2
with:
path: |
~/tmp/download
key: cache-libri-test-clean-and-test-other
- name: Download LibriSpeech test-clean and test-other
if: steps.libri-test-clean-and-test-other-data.outputs.cache-hit != 'true'
shell: bash
run: |
.github/scripts/download-librispeech-test-clean-and-test-other-dataset.sh
- name: Prepare manifests for LibriSpeech test-clean and test-other
shell: bash
run: |
.github/scripts/prepare-librispeech-test-clean-and-test-other-manifests.sh
- name: Cache LibriSpeech test-clean and test-other fbank features
id: libri-test-clean-and-test-other-fbank
uses: actions/cache@v2
with:
path: |
~/tmp/fbank-libri
key: cache-libri-fbank-test-clean-and-test-other-v2
- name: Compute fbank for LibriSpeech test-clean and test-other
if: steps.libri-test-clean-and-test-other-fbank.outputs.cache-hit != 'true'
shell: bash
run: |
.github/scripts/compute-fbank-librispeech-test-clean-and-test-other.sh
- name: Inference with pre-trained model
shell: bash
env:
GITHUB_EVENT_NAME: ${{ github.event_name }}
GITHUB_EVENT_LABEL_NAME: ${{ github.event.label.name }}
run: |
mkdir -p egs/librispeech/ASR/data
ln -sfv ~/tmp/fbank-libri egs/librispeech/ASR/data/fbank
ls -lh egs/librispeech/ASR/data/*
sudo apt-get -qq install git-lfs tree sox
export PYTHONPATH=$PWD:$PYTHONPATH
export PYTHONPATH=~/tmp/kaldifeat/kaldifeat/python:$PYTHONPATH
export PYTHONPATH=~/tmp/kaldifeat/build/lib:$PYTHONPATH
.github/scripts/run-librispeech-conformer-ctc3-2022-11-28.sh
- name: Display decoding results for librispeech conformer_ctc3
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
shell: bash
run: |
cd egs/librispeech/ASR/
tree ./conformer_ctc3/exp
cd conformer_ctc3
echo "results for conformer_ctc3"
echo "===ctc-decoding==="
find exp/ctc-decoding -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/ctc-decoding -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
echo "===1best==="
find exp/1best -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find exp/1best -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
- name: Upload decoding results for librispeech conformer_ctc3
uses: actions/upload-artifact@v2
if: github.event_name == 'schedule' || github.event.label.name == 'run-decode'
with:
name: torch-${{ matrix.torch }}-python-${{ matrix.python-version }}-ubuntu-18.04-cpu-conformer_ctc3-2022-11-28
path: egs/librispeech/ASR/conformer_ctc3/exp/

77
.github/workflows/run-librispeech-conv-emformer-transducer-stateless2-2022-12-05.yml vendored Normal file
View File

@ -0,0 +1,77 @@
name: run-librispeech-conv-emformer-transducer-stateless2-2022-12-05
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_conv_emformer_transducer_stateless2_2022_12_05:
if: github.event.label.name == 'ready' || github.event.label.name == 'ncnn' || 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: ${{ matrix.python-version }}
cache: 'pip'
cache-dependency-path: '**/requirements-ci.txt'
- name: Install Python dependencies
run: |
grep -v '^#' ./requirements-ci.txt | grep -v kaldifst | xargs -n 1 -L 1 pip install
pip uninstall -y protobuf
pip install --no-binary protobuf protobuf
- name: Cache kaldifeat
id: my-cache
uses: actions/cache@v2
with:
path: |
~/tmp/kaldifeat
key: cache-tmp-${{ matrix.python-version }}-2022-09-25
- name: Install kaldifeat
if: steps.my-cache.outputs.cache-hit != 'true'
shell: bash
run: |
.github/scripts/install-kaldifeat.sh
- name: Inference with pre-trained model
shell: bash
env:
GITHUB_EVENT_NAME: ${{ github.event_name }}
GITHUB_EVENT_LABEL_NAME: ${{ github.event.label.name }}
run: |
mkdir -p egs/librispeech/ASR/data
ln -sfv ~/tmp/fbank-libri egs/librispeech/ASR/data/fbank
ls -lh egs/librispeech/ASR/data/*
sudo apt-get -qq install git-lfs tree sox
export PYTHONPATH=$PWD:$PYTHONPATH
export PYTHONPATH=~/tmp/kaldifeat/kaldifeat/python:$PYTHONPATH
export PYTHONPATH=~/tmp/kaldifeat/build/lib:$PYTHONPATH
.github/scripts/run-librispeech-conv-emformer-transducer-stateless2-2022-12-05.sh

38
.github/workflows/run-librispeech-lstm-transducer-stateless2-2022-09-03.yml vendored
View File

@ -1,4 +1,4 @@
name: run-librispeech-lstm-transducer-2022-09-03
name: run-librispeech-lstm-transducer2-2022-09-03
on:
push:
@ -16,9 +16,13 @@ on:
# nightly build at 15:50 UTC time every day
- cron: "50 15 * * *"
concurrency:
group: run_librispeech_lstm_transducer_stateless2_2022_09_03-${{ github.ref }}
cancel-in-progress: true
jobs:
run_librispeech_pruned_transducer_stateless3_2022_05_13:
if: github.event.label.name == 'ncnn' || github.event_name == 'push' || github.event_name == 'schedule'
run_librispeech_lstm_transducer_stateless2_2022_09_03:
if: github.event.label.name == 'ready' || github.event.label.name == 'LODR' || github.event.label.name == 'shallow-fusion' || github.event.label.name == 'ncnn' || github.event.label.name == 'onnx' || github.event_name == 'push' || github.event_name == 'schedule'
runs-on: ${{ matrix.os }}
strategy:
matrix:
@ -107,10 +111,10 @@ jobs:
export PYTHONPATH=~/tmp/kaldifeat/kaldifeat/python:$PYTHONPATH
export PYTHONPATH=~/tmp/kaldifeat/build/lib:$PYTHONPATH
.github/scripts/run-librispeech-lstm-transducer-stateless2-2022-09-03.yml
.github/scripts/run-librispeech-lstm-transducer-stateless2-2022-09-03.sh
- name: Display decoding results for lstm_transducer_stateless2
if: github.event_name == 'schedule' || github.event.label.name == 'ncnn'
if: github.event_name == 'schedule'
shell: bash
run: |
cd egs/librispeech/ASR
@ -128,9 +132,31 @@ jobs:
find modified_beam_search -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find modified_beam_search -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
- name: Display decoding results for lstm_transducer_stateless2
if: github.event.label.name == 'shallow-fusion'
shell: bash
run: |
cd egs/librispeech/ASR
tree lstm_transducer_stateless2/exp
cd lstm_transducer_stateless2/exp
echo "===modified_beam_search_rnnlm_shallow_fusion==="
find modified_beam_search_rnnlm_shallow_fusion -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find modified_beam_search_rnnlm_shallow_fusion -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
- name: Display decoding results for lstm_transducer_stateless2
if: github.event.label.name == 'LODR'
shell: bash
run: |
cd egs/librispeech/ASR
tree lstm_transducer_stateless2/exp
cd lstm_transducer_stateless2/exp
echo "===modified_beam_search_rnnlm_LODR==="
find modified_beam_search_rnnlm_LODR -name "log-*" -exec grep -n --color "best for test-clean" {} + | sort -n -k2
find modified_beam_search_rnnlm_LODR -name "log-*" -exec grep -n --color "best for test-other" {} + | sort -n -k2
- name: Upload decoding results for lstm_transducer_stateless2
uses: actions/upload-artifact@v2
if: github.event_name == 'schedule' || github.event.label.name == 'ncnn'
if: github.event_name == 'schedule' || github.event.label.name == 'shallow-fusion' || github.event.label.name == 'LODR'
with:
name: torch-${{ matrix.torch }}-python-${{ matrix.python-version }}-ubuntu-18.04-cpu-lstm_transducer_stateless2-2022-09-03
path: egs/librispeech/ASR/lstm_transducer_stateless2/exp/

4
.github/workflows/run-librispeech-pruned-transducer-stateless3-2022-05-13.yml vendored
View File

@ -33,6 +33,10 @@ on:
# 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 == 'onnx' || github.event.label.name == 'ready' || github.event.label.name == 'run-decode' || github.event_name == 'push' || github.event_name == 'schedule'

4
.github/workflows/run-librispeech-streaming-transducer-stateless2-2022-06-26.yml vendored
View File

@ -33,6 +33,10 @@ on:
# 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'

4
.github/workflows/run-librispeech-transducer-stateless2-2022-04-19.yml vendored
View File

@ -33,6 +33,10 @@ on:
# 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'

4
.github/workflows/run-pretrained-conformer-ctc.yml vendored
View File

@ -23,6 +23,10 @@ on:
pull_request:
types: [labeled]
concurrency:
group: run_pre_trained_conformer_ctc-${{ github.ref }}
cancel-in-progress: true
jobs:
run_pre_trained_conformer_ctc:
if: github.event.label.name == 'ready' || github.event_name == 'push'

4
.github/workflows/run-pretrained-transducer-stateless-librispeech-100h.yml vendored
View File

@ -32,6 +32,10 @@ on:
# 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'

4
.github/workflows/run-pretrained-transducer-stateless-librispeech-multi-datasets.yml vendored
View File

@ -32,6 +32,10 @@ on:
# 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'

4
.github/workflows/run-pretrained-transducer-stateless-modified-2-aishell.yml vendored
View File

@ -23,6 +23,10 @@ on:
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'

4
.github/workflows/run-pretrained-transducer-stateless-modified-aishell.yml vendored
View File

@ -23,6 +23,10 @@ on:
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'

4
.github/workflows/run-pretrained-transducer-stateless.yml vendored
View File

@ -32,6 +32,10 @@ on:
# 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'

4
.github/workflows/run-pretrained-transducer.yml vendored
View File

@ -23,6 +23,10 @@ on:
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'

71
.github/workflows/run-ptb-rnn-lm.yml vendored Normal file
View File

@ -0,0 +1,71 @@
name: run-ptb-rnn-lm-training
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_ptb_rnn_lm_training-${{ github.ref }}
cancel-in-progress: true
jobs:
run_ptb_rnn_lm_training:
if: github.event.label.name == 'ready' || github.event.label.name == 'rnnlm' || 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: ${{ matrix.python-version }}
cache: 'pip'
cache-dependency-path: '**/requirements-ci.txt'
- name: Install Python dependencies
run: |
grep -v '^#' ./requirements-ci.txt | grep -v kaldifst | xargs -n 1 -L 1 pip install
pip uninstall -y protobuf
pip install --no-binary protobuf protobuf
- name: Prepare data
shell: bash
run: |
export PYTHONPATH=$PWD:$PYTHONPATH
cd egs/ptb/LM
./prepare.sh
- name: Run training
shell: bash
run: |
export PYTHONPATH=$PWD:$PYTHONPATH
cd egs/ptb/LM
./train-rnn-lm.sh --world-size 1 --num-epochs 5 --use-epoch 4 --use-avg 2
- name: Upload pretrained models
uses: actions/upload-artifact@v2
if: github.event.label.name == 'ready' || github.event.label.name == 'rnnlm' || github.event_name == 'push' || github.event_name == 'schedule'
with:
name: python-${{ matrix.python-version }}-ubuntu-rnn-lm-ptb
path: egs/ptb/LM/my-rnnlm-exp/

84
.github/workflows/run-wenetspeech-pruned-transducer-stateless2.yml vendored Normal file
View File

@ -0,0 +1,84 @@
# 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-wenetspeech-pruned-transducer-stateless2
on:
push:
branches:
- master
pull_request:
types: [labeled]
concurrency:
group: run_wenetspeech_pruned_transducer_stateless2-${{ github.ref }}
cancel-in-progress: true
jobs:
run_wenetspeech_pruned_transducer_stateless2:
if: github.event.label.name == 'onnx' || github.event.label.name == 'ready' || github.event_name == 'push' || github.event.label.name == 'wenetspeech'
runs-on: ${{ matrix.os }}
strategy:
matrix:
os: [ubuntu-18.04]
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: ${{ matrix.python-version }}
cache: 'pip'
cache-dependency-path: '**/requirements-ci.txt'
- name: Install Python dependencies
run: |
grep -v '^#' ./requirements-ci.txt | xargs -n 1 -L 1 pip install
pip uninstall -y protobuf
pip install --no-binary protobuf protobuf
- name: Cache kaldifeat
id: my-cache
uses: actions/cache@v2
with:
path: |
~/tmp/kaldifeat
key: cache-tmp-${{ matrix.python-version }}-2022-09-25
- name: Install kaldifeat
if: steps.my-cache.outputs.cache-hit != 'true'
shell: bash
run: |
.github/scripts/install-kaldifeat.sh
- name: Inference with pre-trained model
shell: bash
env:
GITHUB_EVENT_NAME: ${{ github.event_name }}
GITHUB_EVENT_LABEL_NAME: ${{ github.event.label.name }}
run: |
sudo apt-get -qq install git-lfs tree sox
export PYTHONPATH=$PWD:$PYTHONPATH
export PYTHONPATH=~/tmp/kaldifeat/kaldifeat/python:$PYTHONPATH
export PYTHONPATH=~/tmp/kaldifeat/build/lib:$PYTHONPATH
.github/scripts/run-wenetspeech-pruned-transducer-stateless2.sh

10
.github/workflows/run-yesno-recipe.yml vendored
View File

@ -21,11 +21,15 @@ on:
branches:
- master
pull_request:
types: [labeled]
branches:
- master
concurrency:
group: run-yesno-recipe-${{ github.ref }}
cancel-in-progress: true
jobs:
run-yesno-recipe:
if: github.event.label.name == 'ready' || github.event_name == 'push'
runs-on: ${{ matrix.os }}
strategy:
matrix:
@ -61,7 +65,7 @@ jobs:
- name: Install Python dependencies
run: |
grep -v '^#' ./requirements-ci.txt | xargs -n 1 -L 1 pip install
grep -v '^#' ./requirements-ci.txt | grep -v kaldifst | xargs -n 1 -L 1 pip install
pip uninstall -y protobuf
pip install --no-binary protobuf protobuf

19
.github/workflows/style_check.yml vendored
View File

@ -24,13 +24,17 @@ on:
branches:
- master
concurrency:
group: style_check-${{ github.ref }}
cancel-in-progress: true
jobs:
style_check:
runs-on: ${{ matrix.os }}
strategy:
matrix:
os: [ubuntu-18.04, macos-latest]
python-version: [3.7, 3.9]
os: [ubuntu-latest]
python-version: [3.8]
fail-fast: false
steps:
@ -45,17 +49,18 @@ jobs:
- name: Install Python dependencies
run: |
python3 -m pip install --upgrade pip black==21.6b0 flake8==3.9.2 click==8.0.4
# See https://github.com/psf/black/issues/2964
# The version of click should be selected from 8.0.0, 8.0.1, 8.0.2, 8.0.3, and 8.0.4
python3 -m pip install --upgrade pip black==22.3.0 flake8==5.0.4 click==8.1.0
# Click issue fixed in https://github.com/psf/black/pull/2966
- name: Run flake8
shell: bash
working-directory: ${{github.workspace}}
run: |
# stop the build if there are Python syntax errors or undefined names
flake8 . --count --show-source --statistics
flake8 .
flake8 . --count --select=E9,F63,F7,F82 --show-source --statistics
# exit-zero treats all errors as warnings. The GitHub editor is 127 chars wide
flake8 . --count --exit-zero --max-complexity=10 --max-line-length=127 \
--statistics --extend-ignore=E203,E266,E501,F401,E402,F403,F841,W503
- name: Run black
shell: bash

64
.github/workflows/test.yml vendored
View File

@ -21,26 +21,23 @@ on:
branches:
- master
pull_request:
types: [labeled]
branches:
- master
concurrency:
group: test-${{ github.ref }}
cancel-in-progress: true
jobs:
test:
if: github.event.label.name == 'ready' || github.event_name == 'push'
runs-on: ${{ matrix.os }}
strategy:
matrix:
# os: [ubuntu-18.04, macos-10.15]
# disable macOS test for now.
os: [ubuntu-18.04]
python-version: [3.7, 3.8]
torch: ["1.8.0", "1.11.0"]
torchaudio: ["0.8.0", "0.11.0"]
k2-version: ["1.15.1.dev20220427"]
exclude:
- torch: "1.8.0"
torchaudio: "0.11.0"
- torch: "1.11.0"
torchaudio: "0.8.0"
os: [ubuntu-latest]
python-version: ["3.8"]
torch: ["1.10.0"]
torchaudio: ["0.10.0"]
k2-version: ["1.23.2.dev20221201"]
fail-fast: false
@ -67,11 +64,7 @@ jobs:
# numpy 1.20.x does not support python 3.6
pip install numpy==1.19
pip install torch==${{ matrix.torch }}+cpu -f https://download.pytorch.org/whl/cpu/torch_stable.html
if [[ ${{ matrix.torchaudio }} == "0.11.0" ]]; then
pip install torchaudio==${{ matrix.torchaudio }}+cpu -f https://download.pytorch.org/whl/cpu/torch_stable.html
else
pip install torchaudio==${{ matrix.torchaudio }}
fi
pip install torchaudio==${{ 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.org/nightly/
pip install git+https://github.com/lhotse-speech/lhotse
@ -79,6 +72,8 @@ jobs:
pip uninstall -y protobuf
pip install --no-binary protobuf protobuf
pip install kaldifst
pip install onnxruntime
pip install -r requirements.txt
- name: Install graphviz
@ -118,19 +113,20 @@ jobs:
cd ../pruned_transducer_stateless4
pytest -v -s
cd ../pruned_transducer_stateless7
pytest -v -s
cd ../transducer_stateless
pytest -v -s
if [[ ${{ matrix.torchaudio }} == "0.10.0" ]]; then
cd ../transducer
pytest -v -s
cd ../transducer
pytest -v -s
cd ../transducer_stateless2
pytest -v -s
cd ../transducer_stateless2
pytest -v -s
cd ../transducer_lstm
pytest -v -s
fi
cd ../transducer_lstm
pytest -v -s
- name: Run tests
if: startsWith(matrix.os, 'macos')
@ -161,13 +157,11 @@ jobs:
cd ../transducer_stateless
pytest -v -s
if [[ ${{ matrix.torchaudio }} == "0.10.0" ]]; then
cd ../transducer
pytest -v -s
cd ../transducer
pytest -v -s
cd ../transducer_stateless2
pytest -v -s
cd ../transducer_stateless2
pytest -v -s
cd ../transducer_lstm
pytest -v -s
fi
cd ../transducer_lstm
pytest -v -s

20
.gitignore vendored
View File

@ -11,5 +11,25 @@ log
*.bak
*-bak
*bak.py
# Ignore Mac system files
.DS_store
# Ignore node_modules folder
node_modules
# ignore .nfs
.nfs*
# Ignore all text files
*.txt
# Ignore files related to API keys
.env
# Ignore SASS config files
.sass-cache
*.param
*.bin

28
.pre-commit-config.yaml
View File

@ -1,26 +1,38 @@
repos:
- repo: https://github.com/psf/black
rev: 21.6b0
rev: 22.3.0
hooks:
- id: black
args: [--line-length=80]
additional_dependencies: ['click==8.0.1']
args: ["--line-length=88"]
additional_dependencies: ['click==8.1.0']
exclude: icefall\/__init__\.py
- repo: https://github.com/PyCQA/flake8
rev: 3.9.2
rev: 5.0.4
hooks:
- id: flake8
args: [--max-line-length=80]
args: ["--max-line-length=88", "--extend-ignore=E203,E266,E501,F401,E402,F403,F841,W503"]
# What are we ignoring here?
# E203: whitespace before ':'
# E266: too many leading '#' for block comment
# E501: line too long
# F401: module imported but unused
# E402: module level import not at top of file
# F403: 'from module import *' used; unable to detect undefined names
# F841: local variable is assigned to but never used
# W503: line break before binary operator
# In addition, the default ignore list is:
# E121,E123,E126,E226,E24,E704,W503,W504
- repo: https://github.com/pycqa/isort
rev: 5.9.2
rev: 5.10.1
hooks:
- id: isort
args: [--profile=black, --line-length=80]
args: ["--profile=black"]
- repo: https://github.com/pre-commit/pre-commit-hooks
rev: v4.0.1
rev: v4.2.0
hooks:
- id: check-executables-have-shebangs
- id: end-of-file-fixer

4
README.md
View File

@ -82,7 +82,7 @@ The WER for this model is:
|-----|------------|------------|
| WER | 6.59 | 17.69 |
We provide a Colab notebook to run a pre-trained TDNN LSTM CTC model: [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1kNmDXNMwREi0rZGAOIAOJo93REBuOTcd?usp=sharing)
We provide a Colab notebook to run a pre-trained TDNN LSTM CTC model: [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1-iSfQMp2So-We_Uu49N4AAcMInB72u9z?usp=sharing)
#### Transducer: Conformer encoder + LSTM decoder
@ -162,7 +162,7 @@ The CER for this model is:
|-----|-------|
| CER | 10.16 |
We provide a Colab notebook to run a pre-trained TDNN LSTM CTC model: [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1qULaGvXq7PCu_P61oubfz9b53JzY4H3z?usp=sharing)
We provide a Colab notebook to run a pre-trained TDNN LSTM CTC model: [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/drive/1jbyzYq3ytm6j2nlEt-diQm-6QVWyDDEa?usp=sharing)
### TIMIT

26
docker/README.md
View File

@ -2,7 +2,7 @@
2 sets of configuration are provided - (a) Ubuntu18.04-pytorch1.12.1-cuda11.3-cudnn8, and (b) Ubuntu18.04-pytorch1.7.1-cuda11.0-cudnn8.
If your NVIDIA driver supports CUDA Version: 11.3, please go for case (a) Ubuntu18.04-pytorch1.12.1-cuda11.3-cudnn8.
If your NVIDIA driver supports CUDA Version: 11.3, please go for case (a) Ubuntu18.04-pytorch1.12.1-cuda11.3-cudnn8.
Otherwise, since the older PyTorch images are not updated with the [apt-key rotation by NVIDIA](https://developer.nvidia.com/blog/updating-the-cuda-linux-gpg-repository-key), you have to go for case (b) Ubuntu18.04-pytorch1.7.1-cuda11.0-cudnn8. Ensure that your NVDIA driver supports at least CUDA 11.0.
@ -10,7 +10,7 @@ You can check the highest CUDA version within your NVIDIA driver's support with
```bash
$ nvidia-smi
Tue Sep 20 00:26:13 2022
Tue Sep 20 00:26:13 2022
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 450.119.03 Driver Version: 450.119.03 CUDA Version: 11.0 |
|-------------------------------+----------------------+----------------------+
@ -26,7 +26,7 @@ Tue Sep 20 00:26:13 2022
| 41% 30C P8 11W / 280W | 6MiB / 24220MiB | 0% Default |
| | | N/A |
+-------------------------------+----------------------+----------------------+
+-----------------------------------------------------------------------------+
| Processes: |
| GPU GI CI PID Type Process name GPU Memory |
@ -40,15 +40,15 @@ Tue Sep 20 00:26:13 2022
```
## Building images locally
If your environment requires a proxy to access the Internet, remember to add those information into the Dockerfile directly.
For most cases, you can uncomment these lines in the Dockerfile and add in your proxy details.
If your environment requires a proxy to access the Internet, remember to add those information into the Dockerfile directly.
For most cases, you can uncomment these lines in the Dockerfile and add in your proxy details.
```dockerfile
ENV http_proxy=http://aaa.bb.cc.net:8080 \
https_proxy=http://aaa.bb.cc.net:8080
```
Then, proceed with these commands.
Then, proceed with these commands.
### If you are case (a), i.e. your NVIDIA driver supports CUDA version >= 11.3:
@ -72,23 +72,23 @@ docker run -it --runtime=nvidia --shm-size=2gb --name=icefall --gpus all icefall
```
### Tips:
1. Since your data and models most probably won't be in the docker, you must use the -v flag to access the host machine. Do this by specifying `-v {/path/in/docker}:{/path/in/host/machine}`.
1. Since your data and models most probably won't be in the docker, you must use the -v flag to access the host machine. Do this by specifying `-v {/path/in/host/machine}:{/path/in/docker}`.
2. Also, if your environment requires a proxy, this would be a good time to add it in too: `-e http_proxy=http://aaa.bb.cc.net:8080 -e https_proxy=http://aaa.bb.cc.net:8080`.
Overall, your docker run command should look like this.
Overall, your docker run command should look like this.
```bash
docker run -it --runtime=nvidia --shm-size=2gb --name=icefall --gpus all -v {/path/in/docker}:{/path/in/host/machine} -e http_proxy=http://aaa.bb.cc.net:8080 -e https_proxy=http://aaa.bb.cc.net:8080 icefall/pytorch1.12.1
docker run -it --runtime=nvidia --shm-size=2gb --name=icefall --gpus all -v {/path/in/host/machine}:{/path/in/docker} -e http_proxy=http://aaa.bb.cc.net:8080 -e https_proxy=http://aaa.bb.cc.net:8080 icefall/pytorch1.12.1
```
You can explore more docker run options [here](https://docs.docker.com/engine/reference/commandline/run/) to suit your environment.
### Linking to icefall in your host machine
If you already have icefall downloaded onto your host machine, you can use that repository instead so that changes in your code are visible inside and outside of the container.
If you already have icefall downloaded onto your host machine, you can use that repository instead so that changes in your code are visible inside and outside of the container.
Note: Remember to set the -v flag above during the first run of the container, as that is the only way for your container to access your host machine.
Note: Remember to set the -v flag above during the first run of the container, as that is the only way for your container to access your host machine.
Warning: Check that the icefall in your host machine is visible from within your container before proceeding to the commands below.
Use these commands once you are inside the container.
@ -103,7 +103,7 @@ ln -s {/path/in/docker/to/icefall} /workspace/icefall
docker exec -it icefall /bin/bash
```
## Restarting a killed container that has been run before.
## Restarting a killed container that has been run before.
```bash
docker start -ai icefall
```
@ -111,4 +111,4 @@ docker start -ai icefall
## Sample usage of the CPU based images:
```bash
docker run -it icefall /bin/bash
```
```

18
docker/Ubuntu18.04-pytorch1.12.1-cuda11.3-cudnn8/Dockerfile
View File

@ -1,7 +1,7 @@
FROM pytorch/pytorch:1.12.1-cuda11.3-cudnn8-devel
# ENV http_proxy=http://aaa.bbb.cc.net:8080 \
# https_proxy=http://aaa.bbb.cc.net:8080
# https_proxy=http://aaa.bbb.cc.net:8080
# install normal source
RUN apt-get update && \
@ -38,10 +38,10 @@ RUN wget -P /opt https://cmake.org/files/v3.18/cmake-3.18.0.tar.gz && \
rm -rf cmake-3.18.0.tar.gz && \
find /opt/cmake-3.18.0 -type f \( -name "*.o" -o -name "*.la" -o -name "*.a" \) -exec rm {} \; && \
cd -
# flac
# flac
RUN wget -P /opt https://downloads.xiph.org/releases/flac/flac-1.3.2.tar.xz && \
cd /opt && \
cd /opt && \
xz -d flac-1.3.2.tar.xz && \
tar -xvf flac-1.3.2.tar && \
cd flac-1.3.2 && \
@ -49,10 +49,11 @@ RUN wget -P /opt https://downloads.xiph.org/releases/flac/flac-1.3.2.tar.xz &&
make && make install && \
rm -rf flac-1.3.2.tar && \
find /opt/flac-1.3.2 -type f \( -name "*.o" -o -name "*.la" -o -name "*.a" \) -exec rm {} \; && \
cd -
cd -
RUN conda install -y -c pytorch torchaudio=0.12 && \
pip install graphviz
RUN pip install kaldiio graphviz && \
conda install -y -c pytorch torchaudio
#install k2 from source
RUN git clone https://github.com/k2-fsa/k2.git /opt/k2 && \
@ -67,6 +68,7 @@ RUN git clone https://github.com/k2-fsa/icefall /workspace/icefall && \
cd /workspace/icefall && \
pip install -r requirements.txt
RUN pip install kaldifeat
ENV PYTHONPATH /workspace/icefall:$PYTHONPATH
WORKDIR /workspace/icefall
WORKDIR /workspace/icefall

21
docker/Ubuntu18.04-pytorch1.7.1-cuda11.0-cudnn8/Dockerfile
View File

@ -1,12 +1,12 @@
FROM pytorch/pytorch:1.7.1-cuda11.0-cudnn8-devel
# ENV http_proxy=http://aaa.bbb.cc.net:8080 \
# https_proxy=http://aaa.bbb.cc.net:8080
# https_proxy=http://aaa.bbb.cc.net:8080
RUN rm /etc/apt/sources.list.d/cuda.list && \
rm /etc/apt/sources.list.d/nvidia-ml.list && \
apt-key del 7fa2af80
# install normal source
RUN apt-get update && \
apt-get install -y --no-install-recommends \
@ -36,7 +36,7 @@ RUN curl -fsSL https://developer.download.nvidia.com/compute/cuda/repos/ubuntu18
curl -fsSL https://developer.download.nvidia.com/compute/machine-learning/repos/ubuntu1804/x86_64/7fa2af80.pub | apt-key add - && \
echo "deb https://developer.download.nvidia.com/compute/cuda/repos/ubuntu1804/x86_64 /" > /etc/apt/sources.list.d/cuda.list && \
echo "deb https://developer.download.nvidia.com/compute/machine-learning/repos/ubuntu1804/x86_64 /" > /etc/apt/sources.list.d/nvidia-ml.list && \
rm -rf /var/lib/apt/lists/* && \
rm -rf /var/lib/apt/lists/* && \
mv /opt/conda/lib/libcufft.so.10 /opt/libcufft.so.10.bak && \
mv /opt/conda/lib/libcurand.so.10 /opt/libcurand.so.10.bak && \
mv /opt/conda/lib/libcublas.so.11 /opt/libcublas.so.11.bak && \
@ -56,10 +56,10 @@ RUN wget -P /opt https://cmake.org/files/v3.18/cmake-3.18.0.tar.gz && \
rm -rf cmake-3.18.0.tar.gz && \
find /opt/cmake-3.18.0 -type f \( -name "*.o" -o -name "*.la" -o -name "*.a" \) -exec rm {} \; && \
cd -
# flac
# flac
RUN wget -P /opt https://downloads.xiph.org/releases/flac/flac-1.3.2.tar.xz && \
cd /opt && \
cd /opt && \
xz -d flac-1.3.2.tar.xz && \
tar -xvf flac-1.3.2.tar && \
cd flac-1.3.2 && \
@ -67,10 +67,10 @@ RUN wget -P /opt https://downloads.xiph.org/releases/flac/flac-1.3.2.tar.xz &&
make && make install && \
rm -rf flac-1.3.2.tar && \
find /opt/flac-1.3.2 -type f \( -name "*.o" -o -name "*.la" -o -name "*.a" \) -exec rm {} \; && \
cd -
cd -
RUN pip install kaldiio graphviz && \
conda install -y -c pytorch torchaudio=0.7.1
RUN conda install -y -c pytorch torchaudio=0.7.1 && \
pip install graphviz
#install k2 from source
RUN git clone https://github.com/k2-fsa/k2.git /opt/k2 && \
@ -79,7 +79,7 @@ RUN git clone https://github.com/k2-fsa/k2.git /opt/k2 && \
cd -
# install lhotse
RUN pip install git+https://github.com/lhotse-speech/lhotse
RUN pip install git+https://github.com/lhotse-speech/lhotse
RUN git clone https://github.com/k2-fsa/icefall /workspace/icefall && \
cd /workspace/icefall && \
@ -88,4 +88,3 @@ RUN git clone https://github.com/k2-fsa/icefall /workspace/icefall && \
ENV PYTHONPATH /workspace/icefall:$PYTHONPATH
WORKDIR /workspace/icefall

2
docs/source/conf.py
View File

@ -74,7 +74,7 @@ html_context = {
"github_user": "k2-fsa",
"github_repo": "icefall",
"github_version": "master",
"conf_py_path": "/icefall/docs/source/",
"conf_py_path": "/docs/source/",
}
todo_include_todos = True

15
docs/source/contributing/code-style.rst
View File

@ -11,9 +11,9 @@ We use the following tools to make the code style to be as consistent as possibl
The following versions of the above tools are used:
- ``black == 12.6b0``
- ``flake8 == 3.9.2``
- ``isort == 5.9.2``
- ``black == 22.3.0``
- ``flake8 == 5.0.4``
- ``isort == 5.10.1``
After running the following commands:
@ -54,10 +54,17 @@ it should succeed this time:
If you want to check the style of your code before ``git commit``, you
can do the following:
.. code-block:: bash
$ pre-commit install
$ pre-commit run
Or without installing the pre-commit hooks:
.. code-block:: bash
$ cd icefall
$ pip install black==21.6b0 flake8==3.9.2 isort==5.9.2
$ pip install black==22.3.0 flake8==5.0.4 isort==5.10.1
$ black --check your_changed_file.py
$ black your_changed_file.py # modify it in-place
$

9
docs/source/index.rst
View File

@ -21,6 +21,15 @@ speech recognition recipes using `k2 <https://github.com/k2-fsa/k2>`_.
:caption: Contents:
installation/index
model-export/index
.. toctree::
:maxdepth: 3
recipes/index
.. toctree::
:maxdepth: 2
contributing/index
huggingface/index

2
docs/source/installation/images/k2-gt-v1.9-blueviolet.svg
View File

@ -1 +1 @@
<svg xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" width="80" height="20" role="img" aria-label="k2: &gt;= v1.9"><title>k2: &gt;= v1.9</title><linearGradient id="s" x2="0" y2="100%"><stop offset="0" stop-color="#bbb" stop-opacity=".1"/><stop offset="1" stop-opacity=".1"/></linearGradient><clipPath id="r"><rect width="80" height="20" rx="3" fill="#fff"/></clipPath><g clip-path="url(#r)"><rect width="23" height="20" fill="#555"/><rect x="23" width="57" height="20" fill="blueviolet"/><rect width="80" height="20" fill="url(#s)"/></g><g fill="#fff" text-anchor="middle" font-family="Verdana,Geneva,DejaVu Sans,sans-serif" text-rendering="geometricPrecision" font-size="110"><text aria-hidden="true" x="125" y="150" fill="#010101" fill-opacity=".3" transform="scale(.1)" textLength="130">k2</text><text x="125" y="140" transform="scale(.1)" fill="#fff" textLength="130">k2</text><text aria-hidden="true" x="505" y="150" fill="#010101" fill-opacity=".3" transform="scale(.1)" textLength="470">&gt;= v1.9</text><text x="505" y="140" transform="scale(.1)" fill="#fff" textLength="470">&gt;= v1.9</text></g></svg>
<svg xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" width="80" height="20" role="img" aria-label="k2: &gt;= v1.9"><title>k2: &gt;= v1.9</title><linearGradient id="s" x2="0" y2="100%"><stop offset="0" stop-color="#bbb" stop-opacity=".1"/><stop offset="1" stop-opacity=".1"/></linearGradient><clipPath id="r"><rect width="80" height="20" rx="3" fill="#fff"/></clipPath><g clip-path="url(#r)"><rect width="23" height="20" fill="#555"/><rect x="23" width="57" height="20" fill="blueviolet"/><rect width="80" height="20" fill="url(#s)"/></g><g fill="#fff" text-anchor="middle" font-family="Verdana,Geneva,DejaVu Sans,sans-serif" text-rendering="geometricPrecision" font-size="110"><text aria-hidden="true" x="125" y="150" fill="#010101" fill-opacity=".3" transform="scale(.1)" textLength="130">k2</text><text x="125" y="140" transform="scale(.1)" fill="#fff" textLength="130">k2</text><text aria-hidden="true" x="505" y="150" fill="#010101" fill-opacity=".3" transform="scale(.1)" textLength="470">&gt;= v1.9</text><text x="505" y="140" transform="scale(.1)" fill="#fff" textLength="470">&gt;= v1.9</text></g></svg>

Side by Side Swipe Overlay

Before

Width:  |  Height:  |  Size: 1.1 KiB

After

Width:  |  Height:  |  Size: 1.1 KiB

2
docs/source/installation/images/python-gt-v3.6-blue.svg
View File

@ -1 +1 @@
<svg xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" width="98" height="20" role="img" aria-label="python: &gt;= 3.6"><title>python: &gt;= 3.6</title><linearGradient id="s" x2="0" y2="100%"><stop offset="0" stop-color="#bbb" stop-opacity=".1"/><stop offset="1" stop-opacity=".1"/></linearGradient><clipPath id="r"><rect width="98" height="20" rx="3" fill="#fff"/></clipPath><g clip-path="url(#r)"><rect width="49" height="20" fill="#555"/><rect x="49" width="49" height="20" fill="#007ec6"/><rect width="98" height="20" fill="url(#s)"/></g><g fill="#fff" text-anchor="middle" font-family="Verdana,Geneva,DejaVu Sans,sans-serif" text-rendering="geometricPrecision" font-size="110"><text aria-hidden="true" x="255" y="150" fill="#010101" fill-opacity=".3" transform="scale(.1)" textLength="390">python</text><text x="255" y="140" transform="scale(.1)" fill="#fff" textLength="390">python</text><text aria-hidden="true" x="725" y="150" fill="#010101" fill-opacity=".3" transform="scale(.1)" textLength="390">&gt;= 3.6</text><text x="725" y="140" transform="scale(.1)" fill="#fff" textLength="390">&gt;= 3.6</text></g></svg>
<svg xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" width="98" height="20" role="img" aria-label="python: &gt;= 3.6"><title>python: &gt;= 3.6</title><linearGradient id="s" x2="0" y2="100%"><stop offset="0" stop-color="#bbb" stop-opacity=".1"/><stop offset="1" stop-opacity=".1"/></linearGradient><clipPath id="r"><rect width="98" height="20" rx="3" fill="#fff"/></clipPath><g clip-path="url(#r)"><rect width="49" height="20" fill="#555"/><rect x="49" width="49" height="20" fill="#007ec6"/><rect width="98" height="20" fill="url(#s)"/></g><g fill="#fff" text-anchor="middle" font-family="Verdana,Geneva,DejaVu Sans,sans-serif" text-rendering="geometricPrecision" font-size="110"><text aria-hidden="true" x="255" y="150" fill="#010101" fill-opacity=".3" transform="scale(.1)" textLength="390">python</text><text x="255" y="140" transform="scale(.1)" fill="#fff" textLength="390">python</text><text aria-hidden="true" x="725" y="150" fill="#010101" fill-opacity=".3" transform="scale(.1)" textLength="390">&gt;= 3.6</text><text x="725" y="140" transform="scale(.1)" fill="#fff" textLength="390">&gt;= 3.6</text></g></svg>

Side by Side Swipe Overlay

Before

Width:  |  Height:  |  Size: 1.1 KiB

After

Width:  |  Height:  |  Size: 1.1 KiB

2
docs/source/installation/images/torch-gt-v1.6.0-green.svg
View File

@ -1 +1 @@
<svg xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" width="100" height="20" role="img" aria-label="torch: &gt;= 1.6.0"><title>torch: &gt;= 1.6.0</title><linearGradient id="s" x2="0" y2="100%"><stop offset="0" stop-color="#bbb" stop-opacity=".1"/><stop offset="1" stop-opacity=".1"/></linearGradient><clipPath id="r"><rect width="100" height="20" rx="3" fill="#fff"/></clipPath><g clip-path="url(#r)"><rect width="39" height="20" fill="#555"/><rect x="39" width="61" height="20" fill="#97ca00"/><rect width="100" height="20" fill="url(#s)"/></g><g fill="#fff" text-anchor="middle" font-family="Verdana,Geneva,DejaVu Sans,sans-serif" text-rendering="geometricPrecision" font-size="110"><text aria-hidden="true" x="205" y="150" fill="#010101" fill-opacity=".3" transform="scale(.1)" textLength="290">torch</text><text x="205" y="140" transform="scale(.1)" fill="#fff" textLength="290">torch</text><text aria-hidden="true" x="685" y="150" fill="#010101" fill-opacity=".3" transform="scale(.1)" textLength="510">&gt;= 1.6.0</text><text x="685" y="140" transform="scale(.1)" fill="#fff" textLength="510">&gt;= 1.6.0</text></g></svg>
<svg xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" width="100" height="20" role="img" aria-label="torch: &gt;= 1.6.0"><title>torch: &gt;= 1.6.0</title><linearGradient id="s" x2="0" y2="100%"><stop offset="0" stop-color="#bbb" stop-opacity=".1"/><stop offset="1" stop-opacity=".1"/></linearGradient><clipPath id="r"><rect width="100" height="20" rx="3" fill="#fff"/></clipPath><g clip-path="url(#r)"><rect width="39" height="20" fill="#555"/><rect x="39" width="61" height="20" fill="#97ca00"/><rect width="100" height="20" fill="url(#s)"/></g><g fill="#fff" text-anchor="middle" font-family="Verdana,Geneva,DejaVu Sans,sans-serif" text-rendering="geometricPrecision" font-size="110"><text aria-hidden="true" x="205" y="150" fill="#010101" fill-opacity=".3" transform="scale(.1)" textLength="290">torch</text><text x="205" y="140" transform="scale(.1)" fill="#fff" textLength="290">torch</text><text aria-hidden="true" x="685" y="150" fill="#010101" fill-opacity=".3" transform="scale(.1)" textLength="510">&gt;= 1.6.0</text><text x="685" y="140" transform="scale(.1)" fill="#fff" textLength="510">&gt;= 1.6.0</text></g></svg>

Side by Side Swipe Overlay

Before

Width:  |  Height:  |  Size: 1.1 KiB

After

Width:  |  Height:  |  Size: 1.1 KiB

11
docs/source/installation/index.rst
View File

@ -393,6 +393,17 @@ Now let us run the training part:
We use ``export CUDA_VISIBLE_DEVICES=""`` so that ``icefall`` uses CPU
even if there are GPUs available.
.. hint::
In case you get a ``Segmentation fault (core dump)`` error, please use:
.. code-block:: bash
export PROTOCOL_BUFFERS_PYTHON_IMPLEMENTATION=python
See more at `<https://github.com/k2-fsa/icefall/issues/674>` if you are
interested.
The training log is given below:
.. code-block::

21
docs/source/model-export/code/export-model-state-dict-pretrained-out.txt Normal file
View File

@ -0,0 +1,21 @@
2022-10-13 19:09:02,233 INFO [pretrained.py:265] {'best_train_loss': inf, 'best_valid_loss': inf, 'best_train_epoch': -1, 'best_valid_epoch': -1, 'batch_idx_train': 0, 'log_interval': 50, 'reset_interval': 200, 'valid_interval': 3000, 'feature_dim': 80, 'subsampling_factor': 4, 'encoder_dim': 512, 'nhead': 8, 'dim_feedforward': 2048, 'num_encoder_layers': 12, 'decoder_dim': 512, 'joiner_dim': 512, 'model_warm_step': 3000, 'env_info': {'k2-version': '1.21', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': '4810e00d8738f1a21278b0156a42ff396a2d40ac', 'k2-git-date': 'Fri Oct 7 19:35:03 2022', 'lhotse-version': '1.3.0.dev+missing.version.file', 'torch-version': '1.10.0+cu102', 'torch-cuda-available': False, 'torch-cuda-version': '10.2', 'python-version': '3.8', 'icefall-git-branch': 'onnx-doc-1013', 'icefall-git-sha1': 'c39cba5-dirty', 'icefall-git-date': 'Thu Oct 13 15:17:20 2022', 'icefall-path': '/k2-dev/fangjun/open-source/icefall-master', 'k2-path': '/k2-dev/fangjun/open-source/k2-master/k2/python/k2/__init__.py', 'lhotse-path': '/ceph-fj/fangjun/open-source-2/lhotse-jsonl/lhotse/__init__.py', 'hostname': 'de-74279-k2-test-4-0324160024-65bfd8b584-jjlbn', 'IP address': '10.177.74.203'}, 'checkpoint': './icefall-asr-librispeech-pruned-transducer-stateless3-2022-05-13/exp/pretrained-iter-1224000-avg-14.pt', 'bpe_model': './icefall-asr-librispeech-pruned-transducer-stateless3-2022-05-13/data/lang_bpe_500/bpe.model', 'method': 'greedy_search', 'sound_files': ['./icefall-asr-librispeech-pruned-transducer-stateless3-2022-05-13/test_wavs/1089-134686-0001.wav', './icefall-asr-librispeech-pruned-transducer-stateless3-2022-05-13/test_wavs/1221-135766-0001.wav', './icefall-asr-librispeech-pruned-transducer-stateless3-2022-05-13/test_wavs/1221-135766-0002.wav'], 'sample_rate': 16000, 'beam_size': 4, 'beam': 4, 'max_contexts': 4, 'max_states': 8, 'context_size': 2, 'max_sym_per_frame': 1, 'simulate_streaming': False, 'decode_chunk_size': 16, 'left_context': 64, 'dynamic_chunk_training': False, 'causal_convolution': False, 'short_chunk_size': 25, 'num_left_chunks': 4, 'blank_id': 0, 'unk_id': 2, 'vocab_size': 500}
2022-10-13 19:09:02,233 INFO [pretrained.py:271] device: cpu
2022-10-13 19:09:02,233 INFO [pretrained.py:273] Creating model
2022-10-13 19:09:02,612 INFO [train.py:458] Disable giga
2022-10-13 19:09:02,623 INFO [pretrained.py:277] Number of model parameters: 78648040
2022-10-13 19:09:02,951 INFO [pretrained.py:285] Constructing Fbank computer
2022-10-13 19:09:02,952 INFO [pretrained.py:295] Reading sound files: ['./icefall-asr-librispeech-pruned-transducer-stateless3-2022-05-13/test_wavs/1089-134686-0001.wav', './icefall-asr-librispeech-pruned-transducer-stateless3-2022-05-13/test_wavs/1221-135766-0001.wav', './icefall-asr-librispeech-pruned-transducer-stateless3-2022-05-13/test_wavs/1221-135766-0002.wav']
2022-10-13 19:09:02,957 INFO [pretrained.py:301] Decoding started
2022-10-13 19:09:06,700 INFO [pretrained.py:329] Using greedy_search
2022-10-13 19:09:06,912 INFO [pretrained.py:388]
./icefall-asr-librispeech-pruned-transducer-stateless3-2022-05-13/test_wavs/1089-134686-0001.wav:
AFTER EARLY NIGHTFALL THE YELLOW LAMPS WOULD LIGHT UP HERE AND THERE THE SQUALID QUARTER OF THE BROTHELS
./icefall-asr-librispeech-pruned-transducer-stateless3-2022-05-13/test_wavs/1221-135766-0001.wav:
GOD AS A DIRECT CONSEQUENCE OF THE SIN WHICH MAN THUS PUNISHED HAD GIVEN HER A LOVELY CHILD WHOSE PLACE WAS ON THAT SAME DISHONORED BOSOM TO CONNECT HER PARENT FOREVER WITH THE RACE AND DESCENT OF MORTALS AND TO BE FINALLY A BLESSED SOUL IN HEAVEN
./icefall-asr-librispeech-pruned-transducer-stateless3-2022-05-13/test_wavs/1221-135766-0002.wav:
YET THESE THOUGHTS AFFECTED HESTER PRYNNE LESS WITH HOPE THAN APPREHENSION
2022-10-13 19:09:06,912 INFO [pretrained.py:390] Decoding Done

135
docs/source/model-export/export-model-state-dict.rst Normal file
View File

@ -0,0 +1,135 @@
Export model.state_dict()
=========================
When to use it
--------------
During model training, we save checkpoints periodically to disk.
A checkpoint contains the following information:
- ``model.state_dict()``
- ``optimizer.state_dict()``
- and some other information related to training
When we need to resume the training process from some point, we need a checkpoint.
However, if we want to publish the model for inference, then only
``model.state_dict()`` is needed. In this case, we need to strip all other information
except ``model.state_dict()`` to reduce the file size of the published model.
How to export
-------------
Every recipe contains a file ``export.py`` that you can use to
export ``model.state_dict()`` by taking some checkpoints as inputs.
.. hint::
Each ``export.py`` contains well-documented usage information.
In the following, we use
`<https://github.com/k2-fsa/icefall/blob/master/egs/librispeech/ASR/pruned_transducer_stateless3/export.py>`_
as an example.
.. note::
The steps for other recipes are almost the same.
.. code-block:: bash
cd egs/librispeech/ASR
./pruned_transducer_stateless3/export.py \
--exp-dir ./pruned_transducer_stateless3/exp \
--bpe-model data/lang_bpe_500/bpe.model \
--epoch 20 \
--avg 10
will generate a file ``pruned_transducer_stateless3/exp/pretrained.pt``, which
is a dict containing ``{"model": model.state_dict()}`` saved by ``torch.save()``.
How to use the exported model
-----------------------------
For each recipe, we provide pretrained models hosted on huggingface.
You can find links to pretrained models in ``RESULTS.md`` of each dataset.
In the following, we demonstrate how to use the pretrained model from
`<https://huggingface.co/csukuangfj/icefall-asr-librispeech-pruned-transducer-stateless3-2022-05-13>`_.
.. code-block:: bash
cd egs/librispeech/ASR
git lfs install
git clone https://huggingface.co/csukuangfj/icefall-asr-librispeech-pruned-transducer-stateless3-2022-05-13
After cloning the repo with ``git lfs``, you will find several files in the folder
``icefall-asr-librispeech-pruned-transducer-stateless3-2022-05-13/exp``
that have a prefix ``pretrained-``. Those files contain ``model.state_dict()``
exported by the above ``export.py``.
In each recipe, there is also a file ``pretrained.py``, which can use
``pretrained-xxx.pt`` to decode waves. The following is an example:
.. code-block:: bash
cd egs/librispeech/ASR
./pruned_transducer_stateless3/pretrained.py \
--checkpoint ./icefall-asr-librispeech-pruned-transducer-stateless3-2022-05-13/exp/pretrained-iter-1224000-avg-14.pt \
--bpe-model ./icefall-asr-librispeech-pruned-transducer-stateless3-2022-05-13/data/lang_bpe_500/bpe.model \
--method greedy_search \
./icefall-asr-librispeech-pruned-transducer-stateless3-2022-05-13/test_wavs/1089-134686-0001.wav \
./icefall-asr-librispeech-pruned-transducer-stateless3-2022-05-13/test_wavs/1221-135766-0001.wav \
./icefall-asr-librispeech-pruned-transducer-stateless3-2022-05-13/test_wavs/1221-135766-0002.wav
The above commands show how to use the exported model with ``pretrained.py`` to
decode multiple sound files. Its output is given as follows for reference:
.. literalinclude:: ./code/export-model-state-dict-pretrained-out.txt
Use the exported model to run decode.py
---------------------------------------
When we publish the model, we always note down its WERs on some test
dataset in ``RESULTS.md``. This section describes how to use the
pretrained model to reproduce the WER.
.. code-block:: bash
cd egs/librispeech/ASR
git lfs install
git clone https://huggingface.co/csukuangfj/icefall-asr-librispeech-pruned-transducer-stateless3-2022-05-13
cd icefall-asr-librispeech-pruned-transducer-stateless3-2022-05-13/exp
ln -s pretrained-iter-1224000-avg-14.pt epoch-9999.pt
cd ../..
We create a symlink with name ``epoch-9999.pt`` to ``pretrained-iter-1224000-avg-14.pt``,
so that we can pass ``--epoch 9999 --avg 1`` to ``decode.py`` in the following
command:
.. code-block:: bash
./pruned_transducer_stateless3/decode.py \
--epoch 9999 \
--avg 1 \
--exp-dir ./icefall-asr-librispeech-pruned-transducer-stateless3-2022-05-13/exp \
--lang-dir ./icefall-asr-librispeech-pruned-transducer-stateless3-2022-05-13/data/lang_bpe_500 \
--max-duration 600 \
--decoding-method greedy_search
You will find the decoding results in
``./icefall-asr-librispeech-pruned-transducer-stateless3-2022-05-13/exp/greedy_search``.
.. caution::
For some recipes, you also need to pass ``--use-averaged-model False``
to ``decode.py``. The reason is that the exported pretrained model is already
the averaged one.
.. hint::
Before running ``decode.py``, we assume that you have already run
``prepare.sh`` to prepare the test dataset.

12
docs/source/model-export/export-ncnn.rst Normal file
View File

@ -0,0 +1,12 @@
Export to ncnn
==============
We support exporting LSTM transducer models to `ncnn <https://github.com/tencent/ncnn>`_.
Please refer to :ref:`export-model-for-ncnn` for details.
We also provide `<https://github.com/k2-fsa/sherpa-ncnn>`_
performing speech recognition using ``ncnn`` with exported models.
It has been tested on Linux, macOS, Windows, and Raspberry Pi. The project is
self-contained and can be statically linked to produce a binary containing
everything needed.

69
docs/source/model-export/export-onnx.rst Normal file
View File

@ -0,0 +1,69 @@
Export to ONNX
==============
In this section, we describe how to export models to ONNX.
.. hint::
Only non-streaming conformer transducer models are tested.
When to use it
--------------
It you want to use an inference framework that supports ONNX
to run the pretrained model.
How to export
-------------
We use
`<https://github.com/k2-fsa/icefall/tree/master/egs/librispeech/ASR/pruned_transducer_stateless3>`_
as an example in the following.
.. code-block:: bash
cd egs/librispeech/ASR
epoch=14
avg=2
./pruned_transducer_stateless3/export.py \
--exp-dir ./pruned_transducer_stateless3/exp \
--bpe-model data/lang_bpe_500/bpe.model \
--epoch $epoch \
--avg $avg \
--onnx 1
It will generate the following files inside ``pruned_transducer_stateless3/exp``:
- ``encoder.onnx``
- ``decoder.onnx``
- ``joiner.onnx``
- ``joiner_encoder_proj.onnx``
- ``joiner_decoder_proj.onnx``
You can use ``./pruned_transducer_stateless3/exp/onnx_pretrained.py`` to decode
waves with the generated files:
.. code-block:: bash
./pruned_transducer_stateless3/onnx_pretrained.py \
--bpe-model ./data/lang_bpe_500/bpe.model \
--encoder-model-filename ./pruned_transducer_stateless3/exp/encoder.onnx \
--decoder-model-filename ./pruned_transducer_stateless3/exp/decoder.onnx \
--joiner-model-filename ./pruned_transducer_stateless3/exp/joiner.onnx \
--joiner-encoder-proj-model-filename ./pruned_transducer_stateless3/exp/joiner_encoder_proj.onnx \
--joiner-decoder-proj-model-filename ./pruned_transducer_stateless3/exp/joiner_decoder_proj.onnx \
/path/to/foo.wav \
/path/to/bar.wav \
/path/to/baz.wav
How to use the exported model
-----------------------------
We also provide `<https://github.com/k2-fsa/sherpa-onnx>`_
performing speech recognition using `onnxruntime <https://github.com/microsoft/onnxruntime>`_
with exported models.
It has been tested on Linux, macOS, and Windows.

58
docs/source/model-export/export-with-torch-jit-script.rst Normal file
View File

@ -0,0 +1,58 @@
.. _export-model-with-torch-jit-script:
Export model with torch.jit.script()
===================================
In this section, we describe how to export a model via
``torch.jit.script()``.
When to use it
--------------
If we want to use our trained model with torchscript,
we can use ``torch.jit.script()``.
.. hint::
See :ref:`export-model-with-torch-jit-trace`
if you want to use ``torch.jit.trace()``.
How to export
-------------
We use
`<https://github.com/k2-fsa/icefall/tree/master/egs/librispeech/ASR/pruned_transducer_stateless3>`_
as an example in the following.
.. code-block:: bash
cd egs/librispeech/ASR
epoch=14
avg=1
./pruned_transducer_stateless3/export.py \
--exp-dir ./pruned_transducer_stateless3/exp \
--bpe-model data/lang_bpe_500/bpe.model \
--epoch $epoch \
--avg $avg \
--jit 1
It will generate a file ``cpu_jit.pt`` in ``pruned_transducer_stateless3/exp``.
.. caution::
Don't be confused by ``cpu`` in ``cpu_jit.pt``. We move all parameters
to CPU before saving it into a ``pt`` file; that's why we use ``cpu``
in the filename.
How to use the exported model
-----------------------------
Please refer to the following pages for usage:
- `<https://k2-fsa.github.io/sherpa/python/streaming_asr/emformer/index.html>`_
- `<https://k2-fsa.github.io/sherpa/python/streaming_asr/conv_emformer/index.html>`_
- `<https://k2-fsa.github.io/sherpa/python/streaming_asr/conformer/index.html>`_
- `<https://k2-fsa.github.io/sherpa/python/offline_asr/conformer/index.html>`_
- `<https://k2-fsa.github.io/sherpa/cpp/offline_asr/gigaspeech.html>`_
- `<https://k2-fsa.github.io/sherpa/cpp/offline_asr/wenetspeech.html>`_

69
docs/source/model-export/export-with-torch-jit-trace.rst Normal file
View File

@ -0,0 +1,69 @@
.. _export-model-with-torch-jit-trace:
Export model with torch.jit.trace()
===================================
In this section, we describe how to export a model via
``torch.jit.trace()``.
When to use it
--------------
If we want to use our trained model with torchscript,
we can use ``torch.jit.trace()``.
.. hint::
See :ref:`export-model-with-torch-jit-script`
if you want to use ``torch.jit.script()``.
How to export
-------------
We use
`<https://github.com/k2-fsa/icefall/tree/master/egs/librispeech/ASR/lstm_transducer_stateless2>`_
as an example in the following.
.. code-block:: bash
iter=468000
avg=16
cd egs/librispeech/ASR
./lstm_transducer_stateless2/export.py \
--exp-dir ./lstm_transducer_stateless2/exp \
--bpe-model data/lang_bpe_500/bpe.model \
--iter $iter \
--avg $avg \
--jit-trace 1
It will generate three files inside ``lstm_transducer_stateless2/exp``:
- ``encoder_jit_trace.pt``
- ``decoder_jit_trace.pt``
- ``joiner_jit_trace.pt``
You can use
`<https://github.com/k2-fsa/icefall/blob/master/egs/librispeech/ASR/lstm_transducer_stateless2/jit_pretrained.py>`_
to decode sound files with the following commands:
.. code-block:: bash
cd egs/librispeech/ASR
./lstm_transducer_stateless2/jit_pretrained.py \
--bpe-model ./data/lang_bpe_500/bpe.model \
--encoder-model-filename ./lstm_transducer_stateless2/exp/encoder_jit_trace.pt \
--decoder-model-filename ./lstm_transducer_stateless2/exp/decoder_jit_trace.pt \
--joiner-model-filename ./lstm_transducer_stateless2/exp/joiner_jit_trace.pt \
/path/to/foo.wav \
/path/to/bar.wav \
/path/to/baz.wav
How to use the exported models
------------------------------
Please refer to
`<https://k2-fsa.github.io/sherpa/python/streaming_asr/lstm/index.html>`_
for its usage in `sherpa <https://k2-fsa.github.io/sherpa/python/streaming_asr/lstm/index.html>`_.
You can also find pretrained models there.

14
docs/source/model-export/index.rst Normal file
View File

@ -0,0 +1,14 @@
Model export
============
In this section, we describe various ways to export models.
.. toctree::
export-model-state-dict
export-with-torch-jit-trace
export-with-torch-jit-script
export-onnx
export-ncnn

20
docs/source/recipes/aishell/conformer_ctc.rst → docs/source/recipes/Non-streaming-ASR/aishell/conformer_ctc.rst
View File

@ -422,7 +422,7 @@ The information of the test sound files is listed below:
.. code-block:: bash
$ soxi tmp/icefall_asr_aishell_conformer_ctc/test_wavs/*.wav
$ soxi tmp/icefall_asr_aishell_conformer_ctc/test_waves/*.wav
Input File : 'tmp/icefall_asr_aishell_conformer_ctc/test_waves/BAC009S0764W0121.wav'
Channels : 1
@ -485,9 +485,9 @@ The command to run CTC decoding is:
--checkpoint ./tmp/icefall_asr_aishell_conformer_ctc/exp/pretrained.pt \
--tokens-file ./tmp/icefall_asr_aishell_conformer_ctc/data/lang_char/tokens.txt \
--method ctc-decoding \
./tmp/icefall_asr_aishell_conformer_ctc/test_wavs/BAC009S0764W0121.wav \
./tmp/icefall_asr_aishell_conformer_ctc/test_wavs/BAC009S0764W0122.wav \
./tmp/icefall_asr_aishell_conformer_ctc/test_wavs/BAC009S0764W0123.wav
./tmp/icefall_asr_aishell_conformer_ctc/test_waves/BAC009S0764W0121.wav \
./tmp/icefall_asr_aishell_conformer_ctc/test_waves/BAC009S0764W0122.wav \
./tmp/icefall_asr_aishell_conformer_ctc/test_waves/BAC009S0764W0123.wav
The output is given below:
@ -529,9 +529,9 @@ The command to run HLG decoding is:
--words-file ./tmp/icefall_asr_aishell_conformer_ctc/data/lang_char/words.txt \
--HLG ./tmp/icefall_asr_aishell_conformer_ctc/data/lang_char/HLG.pt \
--method 1best \
./tmp/icefall_asr_aishell_conformer_ctc/test_wavs/BAC009S0764W0121.wav \
./tmp/icefall_asr_aishell_conformer_ctc/test_wavs/BAC009S0764W0122.wav \
./tmp/icefall_asr_aishell_conformer_ctc/test_wavs/BAC009S0764W0123.wav
./tmp/icefall_asr_aishell_conformer_ctc/test_waves/BAC009S0764W0121.wav \
./tmp/icefall_asr_aishell_conformer_ctc/test_waves/BAC009S0764W0122.wav \
./tmp/icefall_asr_aishell_conformer_ctc/test_waves/BAC009S0764W0123.wav
The output is given below:
@ -575,9 +575,9 @@ The command to run HLG decoding + attention decoder rescoring is:
--words-file ./tmp/icefall_asr_aishell_conformer_ctc/data/lang_char/words.txt \
--HLG ./tmp/icefall_asr_aishell_conformer_ctc/data/lang_char/HLG.pt \
--method attention-decoder \
./tmp/icefall_asr_aishell_conformer_ctc/test_wavs/BAC009S0764W0121.wav \
./tmp/icefall_asr_aishell_conformer_ctc/test_wavs/BAC009S0764W0122.wav \
./tmp/icefall_asr_aishell_conformer_ctc/test_wavs/BAC009S0764W0123.wav
./tmp/icefall_asr_aishell_conformer_ctc/test_waves/BAC009S0764W0121.wav \
./tmp/icefall_asr_aishell_conformer_ctc/test_waves/BAC009S0764W0122.wav \
./tmp/icefall_asr_aishell_conformer_ctc/test_waves/BAC009S0764W0123.wav
The output is below:

0
docs/source/recipes/aishell/images/aishell-conformer-ctc-tensorboard-log.jpg → docs/source/recipes/Non-streaming-ASR/aishell/images/aishell-conformer-ctc-tensorboard-log.jpg
View File

Side by Side Swipe Overlay

Before

Width:  |  Height:  |  Size: 334 KiB

After

Width:  |  Height:  |  Size: 334 KiB

0
docs/source/recipes/aishell/images/aishell-tdnn-lstm-ctc-tensorboard-log.jpg → docs/source/recipes/Non-streaming-ASR/aishell/images/aishell-tdnn-lstm-ctc-tensorboard-log.jpg
View File

Side by Side Swipe Overlay

Before

Width:  |  Height:  |  Size: 426 KiB

After

Width:  |  Height:  |  Size: 426 KiB

0
docs/source/recipes/aishell/images/aishell-transducer_stateless_modified-tensorboard-log.png → docs/source/recipes/Non-streaming-ASR/aishell/images/aishell-transducer_stateless_modified-tensorboard-log.png
View File

Side by Side Swipe Overlay

Before

Width:  |  Height:  |  Size: 441 KiB

After

Width:  |  Height:  |  Size: 441 KiB

1
docs/source/recipes/aishell/index.rst → docs/source/recipes/Non-streaming-ASR/aishell/index.rst
View File

@ -19,4 +19,3 @@ It can be downloaded from `<https://www.openslr.org/33/>`_
tdnn_lstm_ctc
conformer_ctc
stateless_transducer

0
docs/source/recipes/aishell/stateless_transducer.rst → docs/source/recipes/Non-streaming-ASR/aishell/stateless_transducer.rst
View File

10
docs/source/recipes/aishell/tdnn_lstm_ctc.rst → docs/source/recipes/Non-streaming-ASR/aishell/tdnn_lstm_ctc.rst
View File

@ -402,7 +402,7 @@ The information of the test sound files is listed below:
.. code-block:: bash
$ soxi tmp/icefall_asr_aishell_tdnn_lstm_ctc/test_wavs/*.wav
$ soxi tmp/icefall_asr_aishell_tdnn_lstm_ctc/test_waves/*.wav
Input File : 'tmp/icefall_asr_aishell_tdnn_lstm_ctc/test_waves/BAC009S0764W0121.wav'
Channels : 1
@ -461,9 +461,9 @@ The command to run HLG decoding is:
--words-file ./tmp/icefall_asr_aishell_tdnn_lstm_ctc/data/lang_phone/words.txt \
--HLG ./tmp/icefall_asr_aishell_tdnn_lstm_ctc/data/lang_phone/HLG.pt \
--method 1best \
./tmp/icefall_asr_aishell_tdnn_lstm_ctc/test_wavs/BAC009S0764W0121.wav \
./tmp/icefall_asr_aishell_tdnn_lstm_ctc/test_wavs/BAC009S0764W0122.wav \
./tmp/icefall_asr_aishell_tdnn_lstm_ctc/test_wavs/BAC009S0764W0123.wav
./tmp/icefall_asr_aishell_tdnn_lstm_ctc/test_waves/BAC009S0764W0121.wav \
./tmp/icefall_asr_aishell_tdnn_lstm_ctc/test_waves/BAC009S0764W0122.wav \
./tmp/icefall_asr_aishell_tdnn_lstm_ctc/test_waves/BAC009S0764W0123.wav
The output is given below:
@ -498,7 +498,7 @@ We do provide a colab notebook for this recipe showing how to use a pre-trained
|aishell asr conformer ctc colab notebook|
.. |aishell asr conformer ctc colab notebook| image:: https://colab.research.google.com/assets/colab-badge.svg
:target: https://colab.research.google.com/drive/1qULaGvXq7PCu_P61oubfz9b53JzY4H3z
:target: https://colab.research.google.com/drive/1jbyzYq3ytm6j2nlEt-diQm-6QVWyDDEa?usp=sharing
**Congratulations!** You have finished the aishell ASR recipe with
TDNN-LSTM CTC models in ``icefall``.

10
docs/source/recipes/Non-streaming-ASR/index.rst Normal file
View File

@ -0,0 +1,10 @@
Non Streaming ASR
=================
.. toctree::
:maxdepth: 2
aishell/index
librispeech/index
timit/index
yesno/index

0
docs/source/recipes/librispeech/conformer_ctc.rst → docs/source/recipes/Non-streaming-ASR/librispeech/conformer_ctc.rst
View File

0
docs/source/recipes/librispeech/images/librispeech-conformer-ctc-tensorboard-log.png → docs/source/recipes/Non-streaming-ASR/librispeech/images/librispeech-conformer-ctc-tensorboard-log.png
View File

Side by Side Swipe Overlay

Before

Width:  |  Height:  |  Size: 422 KiB

After

Width:  |  Height:  |  Size: 422 KiB

BIN
docs/source/recipes/Non-streaming-ASR/librispeech/images/librispeech-pruned-transducer-tensorboard-log.jpg Normal file
View File

Binary file not shown.
Side by Side

After

Width:  |  Height:  |  Size: 554 KiB

11
docs/source/recipes/Non-streaming-ASR/librispeech/index.rst Normal file
View File

@ -0,0 +1,11 @@
LibriSpeech
===========
.. toctree::
:maxdepth: 1
tdnn_lstm_ctc
conformer_ctc
pruned_transducer_stateless
lstm_pruned_stateless_transducer
zipformer_mmi

545
docs/source/recipes/Non-streaming-ASR/librispeech/pruned_transducer_stateless.rst Normal file
View File

@ -0,0 +1,545 @@
Pruned transducer statelessX
============================
This tutorial shows you how to run a conformer transducer model
with the `LibriSpeech <https://www.openslr.org/12>`_ dataset.
.. Note::
The tutorial is suitable for `pruned_transducer_stateless <https://github.com/k2-fsa/icefall/tree/master/egs/librispeech/ASR/pruned_transducer_stateless>`_,
`pruned_transducer_stateless2 <https://github.com/k2-fsa/icefall/tree/master/egs/librispeech/ASR/pruned_transducer_stateless2>`_,
`pruned_transducer_stateless4 <https://github.com/k2-fsa/icefall/tree/master/egs/librispeech/ASR/pruned_transducer_stateless4>`_,
`pruned_transducer_stateless5 <https://github.com/k2-fsa/icefall/tree/master/egs/librispeech/ASR/pruned_transducer_stateless5>`_,
We will take pruned_transducer_stateless4 as an example in this tutorial.
.. HINT::
We assume you have read the page :ref:`install icefall` and have setup
the environment for ``icefall``.
.. HINT::
We recommend you to use a GPU or several GPUs to run this recipe.
.. hint::
Please scroll down to the bottom of this page to find download links
for pretrained models if you don't want to train a model from scratch.
We use pruned RNN-T to compute the loss.
.. note::
You can find the paper about pruned RNN-T at the following address:
`<https://arxiv.org/abs/2206.13236>`_
The transducer model consists of 3 parts:
- Encoder, a.k.a, the transcription network. We use a Conformer model (the reworked version by Daniel Povey)
- Decoder, a.k.a, the prediction network. We use a stateless model consisting of
``nn.Embedding`` and ``nn.Conv1d``
- Joiner, a.k.a, the joint network.
.. caution::
Contrary to the conventional RNN-T models, we use a stateless decoder.
That is, it has no recurrent connections.
Data preparation
----------------
.. hint::
The data preparation is the same as other recipes on LibriSpeech dataset,
if you have finished this step, you can skip to ``Training`` directly.
.. code-block:: bash
$ cd egs/librispeech/ASR
$ ./prepare.sh
The script ``./prepare.sh`` handles the data preparation for you, **automagically**.
All you need to do is to run it.
The data preparation contains several stages, you can use the following two
options:
- ``--stage``
- ``--stop-stage``
to control which stage(s) should be run. By default, all stages are executed.
For example,
.. code-block:: bash
$ cd egs/librispeech/ASR
$ ./prepare.sh --stage 0 --stop-stage 0
means to run only stage 0.
To run stage 2 to stage 5, use:
.. code-block:: bash
$ ./prepare.sh --stage 2 --stop-stage 5
.. HINT::
If you have pre-downloaded the `LibriSpeech <https://www.openslr.org/12>`_
dataset and the `musan <http://www.openslr.org/17/>`_ dataset, say,
they are saved in ``/tmp/LibriSpeech`` and ``/tmp/musan``, you can modify
the ``dl_dir`` variable in ``./prepare.sh`` to point to ``/tmp`` so that
``./prepare.sh`` won't re-download them.
.. NOTE::
All generated files by ``./prepare.sh``, e.g., features, lexicon, etc,
are saved in ``./data`` directory.
We provide the following YouTube video showing how to run ``./prepare.sh``.
.. note::
To get the latest news of `next-gen Kaldi <https://github.com/k2-fsa>`_, please subscribe
the following YouTube channel by `Nadira Povey <https://www.youtube.com/channel/UC_VaumpkmINz1pNkFXAN9mw>`_:
`<https://www.youtube.com/channel/UC_VaumpkmINz1pNkFXAN9mw>`_
.. youtube:: ofEIoJL-mGM
Training
--------
Configurable options
~~~~~~~~~~~~~~~~~~~~
.. code-block:: bash
$ cd egs/librispeech/ASR
$ ./pruned_transducer_stateless4/train.py --help
shows you the training options that can be passed from the commandline.
The following options are used quite often:
- ``--exp-dir``
The directory to save checkpoints, training logs and tensorboard.
- ``--full-libri``
If it's True, the training part uses all the training data, i.e.,
960 hours. Otherwise, the training part uses only the subset
``train-clean-100``, which has 100 hours of training data.
.. CAUTION::
The training set is perturbed by speed with two factors: 0.9 and 1.1.
If ``--full-libri`` is True, each epoch actually processes
``3x960 == 2880`` hours of data.
- ``--num-epochs``
It is the number of epochs to train. For instance,
``./pruned_transducer_stateless4/train.py --num-epochs 30`` trains for 30 epochs
and generates ``epoch-1.pt``, ``epoch-2.pt``, ..., ``epoch-30.pt``
in the folder ``./pruned_transducer_stateless4/exp``.
- ``--start-epoch``
It's used to resume training.
``./pruned_transducer_stateless4/train.py --start-epoch 10`` loads the
checkpoint ``./pruned_transducer_stateless4/exp/epoch-9.pt`` and starts
training from epoch 10, based on the state from epoch 9.
- ``--world-size``
It is used for multi-GPU single-machine DDP training.
- (a) If it is 1, then no DDP training is used.
- (b) If it is 2, then GPU 0 and GPU 1 are used for DDP training.
The following shows some use cases with it.
**Use case 1**: You have 4 GPUs, but you only want to use GPU 0 and
GPU 2 for training. You can do the following:
.. code-block:: bash
$ cd egs/librispeech/ASR
$ export CUDA_VISIBLE_DEVICES="0,2"
$ ./pruned_transducer_stateless4/train.py --world-size 2
**Use case 2**: You have 4 GPUs and you want to use all of them
for training. You can do the following:
.. code-block:: bash
$ cd egs/librispeech/ASR
$ ./pruned_transducer_stateless4/train.py --world-size 4
**Use case 3**: You have 4 GPUs but you only want to use GPU 3
for training. You can do the following:
.. code-block:: bash
$ cd egs/librispeech/ASR
$ export CUDA_VISIBLE_DEVICES="3"
$ ./pruned_transducer_stateless4/train.py --world-size 1
.. caution::
Only multi-GPU single-machine DDP training is implemented at present.
Multi-GPU multi-machine DDP training will be added later.
- ``--max-duration``
It specifies the number of seconds over all utterances in a
batch, before **padding**.
If you encounter CUDA OOM, please reduce it.
.. HINT::
Due to padding, the number of seconds of all utterances in a
batch will usually be larger than ``--max-duration``.
A larger value for ``--max-duration`` may cause OOM during training,
while a smaller value may increase the training time. You have to
tune it.
- ``--use-fp16``
If it is True, the model will train with half precision, from our experiment
results, by using half precision you can train with two times larger ``--max-duration``
so as to get almost 2X speed up.
Pre-configured options
~~~~~~~~~~~~~~~~~~~~~~
There are some training options, e.g., number of encoder layers,
encoder dimension, decoder dimension, number of warmup steps etc,
that are not passed from the commandline.
They are pre-configured by the function ``get_params()`` in
`pruned_transducer_stateless4/train.py <https://github.com/k2-fsa/icefall/blob/master/egs/librispeech/ASR/pruned_transducer_stateless4/train.py>`_
You don't need to change these pre-configured parameters. If you really need to change
them, please modify ``./pruned_transducer_stateless4/train.py`` directly.
.. NOTE::
The options for `pruned_transducer_stateless5 <https://github.com/k2-fsa/icefall/blob/master/egs/librispeech/ASR/pruned_transducer_stateless5/train.py>`_ are a little different from
other recipes. It allows you to configure ``--num-encoder-layers``, ``--dim-feedforward``, ``--nhead``, ``--encoder-dim``, ``--decoder-dim``, ``--joiner-dim`` from commandline, so that you can train models with different size with pruned_transducer_stateless5.
Training logs
~~~~~~~~~~~~~
Training logs and checkpoints are saved in ``--exp-dir`` (e.g. ``pruned_transducer_stateless4/exp``.
You will find the following files in that directory:
- ``epoch-1.pt``, ``epoch-2.pt``, ...
These are checkpoint files saved at the end of each epoch, containing model
``state_dict`` and optimizer ``state_dict``.
To resume training from some checkpoint, say ``epoch-10.pt``, you can use:
.. code-block:: bash
$ ./pruned_transducer_stateless4/train.py --start-epoch 11
- ``checkpoint-436000.pt``, ``checkpoint-438000.pt``, ...
These are checkpoint files saved every ``--save-every-n`` batches,
containing model ``state_dict`` and optimizer ``state_dict``.
To resume training from some checkpoint, say ``checkpoint-436000``, you can use:
.. code-block:: bash
$ ./pruned_transducer_stateless4/train.py --start-batch 436000
- ``tensorboard/``
This folder contains tensorBoard logs. Training loss, validation loss, learning
rate, etc, are recorded in these logs. You can visualize them by:
.. code-block:: bash
$ cd pruned_transducer_stateless4/exp/tensorboard
$ tensorboard dev upload --logdir . --description "pruned transducer training for LibriSpeech with icefall"
It will print something like below:
.. code-block::
TensorFlow installation not found - running with reduced feature set.
Upload started and will continue reading any new data as it's added to the logdir.
To stop uploading, press Ctrl-C.
New experiment created. View your TensorBoard at: https://tensorboard.dev/experiment/QOGSPBgsR8KzcRMmie9JGw/
[2022-11-20T15:50:50] Started scanning logdir.
Uploading 4468 scalars...
[2022-11-20T15:53:02] Total uploaded: 210171 scalars, 0 tensors, 0 binary objects
Listening for new data in logdir...
Note there is a URL in the above output. Click it and you will see
the following screenshot:
.. figure:: images/librispeech-pruned-transducer-tensorboard-log.jpg
:width: 600
:alt: TensorBoard screenshot
:align: center
:target: https://tensorboard.dev/experiment/QOGSPBgsR8KzcRMmie9JGw/
TensorBoard screenshot.
.. hint::
If you don't have access to google, you can use the following command
to view the tensorboard log locally:
.. code-block:: bash
cd pruned_transducer_stateless4/exp/tensorboard
tensorboard --logdir . --port 6008
It will print the following message:
.. code-block::
Serving TensorBoard on localhost; to expose to the network, use a proxy or pass --bind_all
TensorBoard 2.8.0 at http://localhost:6008/ (Press CTRL+C to quit)
Now start your browser and go to `<http://localhost:6008>`_ to view the tensorboard
logs.
- ``log/log-train-xxxx``
It is the detailed training log in text format, same as the one
you saw printed to the console during training.
Usage example
~~~~~~~~~~~~~
You can use the following command to start the training using 6 GPUs:
.. code-block:: bash
export CUDA_VISIBLE_DEVICES="0,1,2,3,4,5"
./pruned_transducer_stateless4/train.py \
--world-size 6 \
--num-epochs 30 \
--start-epoch 1 \
--exp-dir pruned_transducer_stateless4/exp \
--full-libri 1 \
--max-duration 300
Decoding
--------
The decoding part uses checkpoints saved by the training part, so you have
to run the training part first.
.. hint::
There are two kinds of checkpoints:
- (1) ``epoch-1.pt``, ``epoch-2.pt``, ..., which are saved at the end
of each epoch. You can pass ``--epoch`` to
``pruned_transducer_stateless4/decode.py`` to use them.
- (2) ``checkpoints-436000.pt``, ``epoch-438000.pt``, ..., which are saved
every ``--save-every-n`` batches. You can pass ``--iter`` to
``pruned_transducer_stateless4/decode.py`` to use them.
We suggest that you try both types of checkpoints and choose the one
that produces the lowest WERs.
.. code-block:: bash
$ cd egs/librispeech/ASR
$ ./pruned_transducer_stateless4/decode.py --help
shows the options for decoding.
The following shows two examples (for two types of checkpoints):
.. code-block:: bash
for m in greedy_search fast_beam_search modified_beam_search; do
for epoch in 25 20; do
for avg in 7 5 3 1; do
./pruned_transducer_stateless4/decode.py \
--epoch $epoch \
--avg $avg \
--exp-dir pruned_transducer_stateless4/exp \
--max-duration 600 \
--decoding-method $m
done
done
done
.. code-block:: bash
for m in greedy_search fast_beam_search modified_beam_search; do
for iter in 474000; do
for avg in 8 10 12 14 16 18; do
./pruned_transducer_stateless4/decode.py \
--iter $iter \
--avg $avg \
--exp-dir pruned_transducer_stateless4/exp \
--max-duration 600 \
--decoding-method $m
done
done
done
.. Note::
Supporting decoding methods are as follows:
- ``greedy_search`` : It takes the symbol with largest posterior probability
of each frame as the decoding result.
- ``beam_search`` : It implements Algorithm 1 in https://arxiv.org/pdf/1211.3711.pdf and
`espnet/nets/beam_search_transducer.py <https://github.com/espnet/espnet/blob/master/espnet/nets/beam_search_transducer.py#L247>`_
is used as a reference. Basicly, it keeps topk states for each frame, and expands the kept states with their own contexts to
next frame.
- ``modified_beam_search`` : It implements the same algorithm as ``beam_search`` above, but it
runs in batch mode with ``--max-sym-per-frame=1`` being hardcoded.
- ``fast_beam_search`` : It implements graph composition between the output ``log_probs`` and
given ``FSAs``. It is hard to describe the details in several lines of texts, you can read
our paper in https://arxiv.org/pdf/2211.00484.pdf or our `rnnt decode code in k2 <https://github.com/k2-fsa/k2/blob/master/k2/csrc/rnnt_decode.h>`_. ``fast_beam_search`` can decode with ``FSAs`` on GPU efficiently.
- ``fast_beam_search_LG`` : The same as ``fast_beam_search`` above, ``fast_beam_search`` uses
an trivial graph that has only one state, while ``fast_beam_search_LG`` uses an LG graph
(with N-gram LM).
- ``fast_beam_search_nbest`` : It produces the decoding results as follows:
- (1) Use ``fast_beam_search`` to get a lattice
- (2) Select ``num_paths`` paths from the lattice using ``k2.random_paths()``
- (3) Unique the selected paths
- (4) Intersect the selected paths with the lattice and compute the
shortest path from the intersection result
- (5) The path with the largest score is used as the decoding output.
- ``fast_beam_search_nbest_LG`` : It implements same logic as ``fast_beam_search_nbest``, the
only difference is that it uses ``fast_beam_search_LG`` to generate the lattice.
Export Model
------------
`pruned_transducer_stateless4/export.py <https://github.com/k2-fsa/icefall/blob/master/egs/librispeech/ASR/pruned_transducer_stateless4/export.py>`_ supports exporting checkpoints from ``pruned_transducer_stateless4/exp`` in the following ways.
Export ``model.state_dict()``
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Checkpoints saved by ``pruned_transducer_stateless4/train.py`` also include
``optimizer.state_dict()``. It is useful for resuming training. But after training,
we are interested only in ``model.state_dict()``. You can use the following
command to extract ``model.state_dict()``.
.. code-block:: bash
# Assume that --epoch 25 --avg 3 produces the smallest WER
# (You can get such information after running ./pruned_transducer_stateless4/decode.py)
epoch=25
avg=3
./pruned_transducer_stateless4/export.py \
--exp-dir ./pruned_transducer_stateless4/exp \
--bpe-model data/lang_bpe_500/bpe.model \
--epoch $epoch \
--avg $avg
It will generate a file ``./pruned_transducer_stateless4/exp/pretrained.pt``.
.. hint::
To use the generated ``pretrained.pt`` for ``pruned_transducer_stateless4/decode.py``,
you can run:
.. code-block:: bash
cd pruned_transducer_stateless4/exp
ln -s pretrained.pt epoch-999.pt
And then pass ``--epoch 999 --avg 1 --use-averaged-model 0`` to
``./pruned_transducer_stateless4/decode.py``.
To use the exported model with ``./pruned_transducer_stateless4/pretrained.py``, you
can run:
.. code-block:: bash
./pruned_transducer_stateless4/pretrained.py \
--checkpoint ./pruned_transducer_stateless4/exp/pretrained.pt \
--bpe-model ./data/lang_bpe_500/bpe.model \
--method greedy_search \
/path/to/foo.wav \
/path/to/bar.wav
Export model using ``torch.jit.script()``
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
.. code-block:: bash
./pruned_transducer_stateless4/export.py \
--exp-dir ./pruned_transducer_stateless4/exp \
--bpe-model data/lang_bpe_500/bpe.model \
--epoch 25 \
--avg 3 \
--jit 1
It will generate a file ``cpu_jit.pt`` in the given ``exp_dir``. You can later
load it by ``torch.jit.load("cpu_jit.pt")``.
Note ``cpu`` in the name ``cpu_jit.pt`` means the parameters when loaded into Python
are on CPU. You can use ``to("cuda")`` to move them to a CUDA device.
.. NOTE::
You will need this ``cpu_jit.pt`` when deploying with Sherpa framework.
Download pretrained models
--------------------------
If you don't want to train from scratch, you can download the pretrained models
by visiting the following links:
- `pruned_transducer_stateless <https://huggingface.co/csukuangfj/icefall-asr-librispeech-pruned-transducer-stateless-2022-03-12>`_
- `pruned_transducer_stateless2 <https://huggingface.co/csukuangfj/icefall-asr-librispeech-pruned-transducer-stateless2-2022-04-29>`_
- `pruned_transducer_stateless4 <https://huggingface.co/Zengwei/icefall-asr-librispeech-pruned-transducer-stateless4-2022-06-03>`_
- `pruned_transducer_stateless5 <https://huggingface.co/Zengwei/icefall-asr-librispeech-pruned-transducer-stateless5-2022-07-07>`_
See `<https://github.com/k2-fsa/icefall/blob/master/egs/librispeech/ASR/RESULTS.md>`_
for the details of the above pretrained models
Deploy with Sherpa
------------------
Please see `<https://k2-fsa.github.io/sherpa/python/offline_asr/conformer/librispeech.html#>`_
for how to deploy the models in ``sherpa``.

2
docs/source/recipes/librispeech/tdnn_lstm_ctc.rst → docs/source/recipes/Non-streaming-ASR/librispeech/tdnn_lstm_ctc.rst
View File

@ -398,7 +398,7 @@ We provide a colab notebook for decoding with pre-trained model.
|librispeech tdnn_lstm_ctc colab notebook|
.. |librispeech tdnn_lstm_ctc colab notebook| image:: https://colab.research.google.com/assets/colab-badge.svg
:target: https://colab.research.google.com/drive/1kNmDXNMwREi0rZGAOIAOJo93REBuOTcd
:target: https://colab.research.google.com/drive/1-iSfQMp2So-We_Uu49N4AAcMInB72u9z?usp=sharing
**Congratulations!** You have finished the TDNN-LSTM-CTC recipe on librispeech in ``icefall``.

422
docs/source/recipes/Non-streaming-ASR/librispeech/zipformer_mmi.rst Normal file
View File

@ -0,0 +1,422 @@
Zipformer MMI
===============
.. hint::
Please scroll down to the bottom of this page to find download links
for pretrained models if you don't want to train a model from scratch.
This tutorial shows you how to train an Zipformer MMI model
with the `LibriSpeech <https://www.openslr.org/12>`_ dataset.
We use LF-MMI to compute the loss.
.. note::
You can find the document about LF-MMI training at the following address:
`<https://github.com/k2-fsa/next-gen-kaldi-wechat/blob/master/pdf/LF-MMI-training-and-decoding-in-k2-Part-I.pdf>`_
Data preparation
----------------
.. code-block:: bash
$ cd egs/librispeech/ASR
$ ./prepare.sh
The script ``./prepare.sh`` handles the data preparation for you, **automagically**.
All you need to do is to run it.
.. note::
We encourage you to read ``./prepare.sh``.
The data preparation contains several stages. You can use the following two
options:
- ``--stage``
- ``--stop-stage``
to control which stage(s) should be run. By default, all stages are executed.
For example,
.. code-block:: bash
$ cd egs/librispeech/ASR
$ ./prepare.sh --stage 0 --stop-stage 0
means to run only stage 0.
To run stage 2 to stage 5, use:
.. code-block:: bash
$ ./prepare.sh --stage 2 --stop-stage 5
.. hint::
If you have pre-downloaded the `LibriSpeech <https://www.openslr.org/12>`_
dataset and the `musan <http://www.openslr.org/17/>`_ dataset, say,
they are saved in ``/tmp/LibriSpeech`` and ``/tmp/musan``, you can modify
the ``dl_dir`` variable in ``./prepare.sh`` to point to ``/tmp`` so that
``./prepare.sh`` won't re-download them.
.. note::
All generated files by ``./prepare.sh``, e.g., features, lexicon, etc,
are saved in ``./data`` directory.
We provide the following YouTube video showing how to run ``./prepare.sh``.
.. note::
To get the latest news of `next-gen Kaldi <https://github.com/k2-fsa>`_, please subscribe
the following YouTube channel by `Nadira Povey <https://www.youtube.com/channel/UC_VaumpkmINz1pNkFXAN9mw>`_:
`<https://www.youtube.com/channel/UC_VaumpkmINz1pNkFXAN9mw>`_
.. youtube:: ofEIoJL-mGM
Training
--------
For stability, it uses CTC loss for model warm-up and then switches to MMI loss.
Configurable options
~~~~~~~~~~~~~~~~~~~~
.. code-block:: bash
$ cd egs/librispeech/ASR
$ ./zipformer_mmi/train.py --help
shows you the training options that can be passed from the commandline.
The following options are used quite often:
- ``--full-libri``
If it's True, the training part uses all the training data, i.e.,
960 hours. Otherwise, the training part uses only the subset
``train-clean-100``, which has 100 hours of training data.
.. CAUTION::
The training set is perturbed by speed with two factors: 0.9 and 1.1.
If ``--full-libri`` is True, each epoch actually processes
``3x960 == 2880`` hours of data.
- ``--num-epochs``
It is the number of epochs to train. For instance,
``./zipformer_mmi/train.py --num-epochs 30`` trains for 30 epochs
and generates ``epoch-1.pt``, ``epoch-2.pt``, ..., ``epoch-30.pt``
in the folder ``./zipformer_mmi/exp``.
- ``--start-epoch``
It's used to resume training.
``./zipformer_mmi/train.py --start-epoch 10`` loads the
checkpoint ``./zipformer_mmi/exp/epoch-9.pt`` and starts
training from epoch 10, based on the state from epoch 9.
- ``--world-size``
It is used for multi-GPU single-machine DDP training.
- (a) If it is 1, then no DDP training is used.
- (b) If it is 2, then GPU 0 and GPU 1 are used for DDP training.
The following shows some use cases with it.
**Use case 1**: You have 4 GPUs, but you only want to use GPU 0 and
GPU 2 for training. You can do the following:
.. code-block:: bash
$ cd egs/librispeech/ASR
$ export CUDA_VISIBLE_DEVICES="0,2"
$ ./zipformer_mmi/train.py --world-size 2
**Use case 2**: You have 4 GPUs and you want to use all of them
for training. You can do the following:
.. code-block:: bash
$ cd egs/librispeech/ASR
$ ./zipformer_mmi/train.py --world-size 4
**Use case 3**: You have 4 GPUs but you only want to use GPU 3
for training. You can do the following:
.. code-block:: bash
$ cd egs/librispeech/ASR
$ export CUDA_VISIBLE_DEVICES="3"
$ ./zipformer_mmi/train.py --world-size 1
.. caution::
Only multi-GPU single-machine DDP training is implemented at present.
Multi-GPU multi-machine DDP training will be added later.
- ``--max-duration``
It specifies the number of seconds over all utterances in a
batch, before **padding**.
If you encounter CUDA OOM, please reduce it.
.. HINT::
Due to padding, the number of seconds of all utterances in a
batch will usually be larger than ``--max-duration``.
A larger value for ``--max-duration`` may cause OOM during training,
while a smaller value may increase the training time. You have to
tune it.
Pre-configured options
~~~~~~~~~~~~~~~~~~~~~~
There are some training options, e.g., weight decay,
number of warmup steps, results dir, etc,
that are not passed from the commandline.
They are pre-configured by the function ``get_params()`` in
`zipformer_mmi/train.py <https://github.com/k2-fsa/icefall/blob/master/egs/librispeech/ASR/zipformer_mmi/train.py>`_
You don't need to change these pre-configured parameters. If you really need to change
them, please modify ``./zipformer_mmi/train.py`` directly.
Training logs
~~~~~~~~~~~~~
Training logs and checkpoints are saved in ``zipformer_mmi/exp``.
You will find the following files in that directory:
- ``epoch-1.pt``, ``epoch-2.pt``, ...
These are checkpoint files saved at the end of each epoch, containing model
``state_dict`` and optimizer ``state_dict``.
To resume training from some checkpoint, say ``epoch-10.pt``, you can use:
.. code-block:: bash
$ ./zipformer_mmi/train.py --start-epoch 11
- ``checkpoint-436000.pt``, ``checkpoint-438000.pt``, ...
These are checkpoint files saved every ``--save-every-n`` batches,
containing model ``state_dict`` and optimizer ``state_dict``.
To resume training from some checkpoint, say ``checkpoint-436000``, you can use:
.. code-block:: bash
$ ./zipformer_mmi/train.py --start-batch 436000
- ``tensorboard/``
This folder contains tensorBoard logs. Training loss, validation loss, learning
rate, etc, are recorded in these logs. You can visualize them by:
.. code-block:: bash
$ cd zipformer_mmi/exp/tensorboard
$ tensorboard dev upload --logdir . --description "Zipformer MMI training for LibriSpeech with icefall"
It will print something like below:
.. code-block::
TensorFlow installation not found - running with reduced feature set.
Upload started and will continue reading any new data as it's added to the logdir.
To stop uploading, press Ctrl-C.
New experiment created. View your TensorBoard at: https://tensorboard.dev/experiment/xyOZUKpEQm62HBIlUD4uPA/
Note there is a URL in the above output. Click it and you will see
tensorboard.
.. hint::
If you don't have access to google, you can use the following command
to view the tensorboard log locally:
.. code-block:: bash
cd zipformer_mmi/exp/tensorboard
tensorboard --logdir . --port 6008
It will print the following message:
.. code-block::
Serving TensorBoard on localhost; to expose to the network, use a proxy or pass --bind_all
TensorBoard 2.8.0 at http://localhost:6008/ (Press CTRL+C to quit)
Now start your browser and go to `<http://localhost:6008>`_ to view the tensorboard
logs.
- ``log/log-train-xxxx``
It is the detailed training log in text format, same as the one
you saw printed to the console during training.
Usage example
~~~~~~~~~~~~~
You can use the following command to start the training using 8 GPUs:
.. code-block:: bash
export CUDA_VISIBLE_DEVICES="0,1,2,3"
./zipformer_mmi/train.py \
--world-size 4 \
--num-epochs 30 \
--start-epoch 1 \
--full-libri 1 \
--exp-dir zipformer_mmi/exp \
--max-duration 500 \
--use-fp16 1 \
--num-workers 2
Decoding
--------
The decoding part uses checkpoints saved by the training part, so you have
to run the training part first.
.. hint::
There are two kinds of checkpoints:
- (1) ``epoch-1.pt``, ``epoch-2.pt``, ..., which are saved at the end
of each epoch. You can pass ``--epoch`` to
``zipformer_mmi/decode.py`` to use them.
- (2) ``checkpoints-436000.pt``, ``epoch-438000.pt``, ..., which are saved
every ``--save-every-n`` batches. You can pass ``--iter`` to
``zipformer_mmi/decode.py`` to use them.
We suggest that you try both types of checkpoints and choose the one
that produces the lowest WERs.
.. code-block:: bash
$ cd egs/librispeech/ASR
$ ./zipformer_mmi/decode.py --help
shows the options for decoding.
The following shows the example using ``epoch-*.pt``:
.. code-block:: bash
for m in nbest nbest-rescoring-LG nbest-rescoring-3-gram nbest-rescoring-4-gram; do
./zipformer_mmi/decode.py \
--epoch 30 \
--avg 10 \
--exp-dir ./zipformer_mmi/exp/ \
--max-duration 100 \
--lang-dir data/lang_bpe_500 \
--nbest-scale 1.2 \
--hp-scale 1.0 \
--decoding-method $m
done
Export models
-------------
`zipformer_mmi/export.py <https://github.com/k2-fsa/icefall/blob/master/egs/librispeech/ASR/zipformer_mmi/export.py>`_ supports exporting checkpoints from ``zipformer_mmi/exp`` in the following ways.
Export ``model.state_dict()``
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Checkpoints saved by ``zipformer_mmi/train.py`` also include
``optimizer.state_dict()``. It is useful for resuming training. But after training,
we are interested only in ``model.state_dict()``. You can use the following
command to extract ``model.state_dict()``.
.. code-block:: bash
./zipformer_mmi/export.py \
--exp-dir ./zipformer_mmi/exp \
--bpe-model data/lang_bpe_500/bpe.model \
--epoch 30 \
--avg 9 \
--jit 0
It will generate a file ``./zipformer_mmi/exp/pretrained.pt``.
.. hint::
To use the generated ``pretrained.pt`` for ``zipformer_mmi/decode.py``,
you can run:
.. code-block:: bash
cd zipformer_mmi/exp
ln -s pretrained epoch-9999.pt
And then pass ``--epoch 9999 --avg 1 --use-averaged-model 0`` to
``./zipformer_mmi/decode.py``.
To use the exported model with ``./zipformer_mmi/pretrained.py``, you
can run:
.. code-block:: bash
./zipformer_mmi/pretrained.py \
--checkpoint ./zipformer_mmi/exp/pretrained.pt \
--bpe-model ./data/lang_bpe_500/bpe.model \
--method 1best \
/path/to/foo.wav \
/path/to/bar.wav
Export model using ``torch.jit.script()``
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
.. code-block:: bash
./zipformer_mmi/export.py \
--exp-dir ./zipformer_mmi/exp \
--bpe-model data/lang_bpe_500/bpe.model \
--epoch 30 \
--avg 9 \
--jit 1
It will generate a file ``cpu_jit.pt`` in the given ``exp_dir``. You can later
load it by ``torch.jit.load("cpu_jit.pt")``.
Note ``cpu`` in the name ``cpu_jit.pt`` means the parameters when loaded into Python
are on CPU. You can use ``to("cuda")`` to move them to a CUDA device.
To use the generated files with ``./zipformer_mmi/jit_pretrained.py``:
.. code-block:: bash
./zipformer_mmi/jit_pretrained.py \
--nn-model-filename ./zipformer_mmi/exp/cpu_jit.pt \
--bpe-model ./data/lang_bpe_500/bpe.model \
--method 1best \
/path/to/foo.wav \
/path/to/bar.wav
Download pretrained models
--------------------------
If you don't want to train from scratch, you can download the pretrained models
by visiting the following links:
- `<https://huggingface.co/Zengwei/icefall-asr-librispeech-zipformer-mmi-2022-12-08>`_
See `<https://github.com/k2-fsa/icefall/blob/master/egs/librispeech/ASR/RESULTS.md>`_
for the details of the above pretrained models

1
docs/source/recipes/timit/index.rst → docs/source/recipes/Non-streaming-ASR/timit/index.rst
View File

@ -6,4 +6,3 @@ TIMIT
tdnn_ligru_ctc
tdnn_lstm_ctc

28
docs/source/recipes/timit/tdnn_ligru_ctc.rst → docs/source/recipes/Non-streaming-ASR/timit/tdnn_ligru_ctc.rst
View File

@ -148,10 +148,10 @@ Some commonly used options are:
$ ./tdnn_ligru_ctc/decode.py --epoch 25 --avg 17
uses the average of ``epoch-9.pt``, ``epoch-10.pt``, ``epoch-11.pt``,
``epoch-12.pt``, ``epoch-13.pt``, ``epoch-14.pt``, ``epoch-15.pt``,
``epoch-16.pt``, ``epoch-17.pt``, ``epoch-18.pt``, ``epoch-19.pt``,
``epoch-20.pt``, ``epoch-21.pt``, ``epoch-22.pt``, ``epoch-23.pt``,
uses the average of ``epoch-9.pt``, ``epoch-10.pt``, ``epoch-11.pt``,
``epoch-12.pt``, ``epoch-13.pt``, ``epoch-14.pt``, ``epoch-15.pt``,
``epoch-16.pt``, ``epoch-17.pt``, ``epoch-18.pt``, ``epoch-19.pt``,
``epoch-20.pt``, ``epoch-21.pt``, ``epoch-22.pt``, ``epoch-23.pt``,
``epoch-24.pt`` and ``epoch-25.pt``
for decoding.
@ -317,13 +317,13 @@ To decode with ``1best`` method, we can use:
.. code-block:: bash
./tdnn_ligru_ctc/pretrained.py
./tdnn_ligru_ctc/pretrained.py
--method 1best
--checkpoint ./tmp-ligru/icefall_asr_timit_tdnn_ligru_ctc/exp/pretrained_average_9_25.pt
--words-file ./tmp-ligru/icefall_asr_timit_tdnn_ligru_ctc/data/lang_phone/words.txt
--HLG ./tmp-ligru/icefall_asr_timit_tdnn_ligru_ctc/data/lang_phone/HLG.pt
./tmp-ligru/icefall_asr_timit_tdnn_ligru_ctc/test_waves/FDHC0_SI1559.WAV
./tmp-ligru/icefall_asr_timit_tdnn_ligru_ctc/test_waves/FELC0_SI756.WAV
--checkpoint ./tmp-ligru/icefall_asr_timit_tdnn_ligru_ctc/exp/pretrained_average_9_25.pt
--words-file ./tmp-ligru/icefall_asr_timit_tdnn_ligru_ctc/data/lang_phone/words.txt
--HLG ./tmp-ligru/icefall_asr_timit_tdnn_ligru_ctc/data/lang_phone/HLG.pt
./tmp-ligru/icefall_asr_timit_tdnn_ligru_ctc/test_waves/FDHC0_SI1559.WAV
./tmp-ligru/icefall_asr_timit_tdnn_ligru_ctc/test_waves/FELC0_SI756.WAV
./tmp-ligru/icefall_asr_timit_tdnn_ligru_ctc/test_waves/FMGD0_SI1564.WAV
The output is:
@ -337,7 +337,7 @@ The output is:
2021-11-08 20:41:38,697 INFO [pretrained.py:210] Reading sound files: ['./tmp-ligru/icefall_asr_timit_tdnn_ligru_ctc/test_waves/FDHC0_SI1559.WAV', './tmp-ligru/icefall_asr_timit_tdnn_ligru_ctc/test_waves/FELC0_SI756.WAV', './tmp-ligru/icefall_asr_timit_tdnn_ligru_ctc/test_waves/FMGD0_SI1564.WAV']
2021-11-08 20:41:38,704 INFO [pretrained.py:216] Decoding started
2021-11-08 20:41:39,819 INFO [pretrained.py:246] Use HLG decoding
2021-11-08 20:41:39,829 INFO [pretrained.py:267]
2021-11-08 20:41:39,829 INFO [pretrained.py:267]
./tmp-ligru/icefall_asr_timit_tdnn_ligru_ctc/test_waves/FDHC0_SI1559.WAV:
sil dh ih sh uw ah l iy v iy z ih sil p r aa sil k s ih m ey dx ih sil d w uh dx ih w ih s f iy l ih ng w ih th ih n ih m s eh l f sil jh
@ -362,8 +362,8 @@ To decode with ``whole-lattice-rescoring`` methond, you can use
--HLG ./tmp-ligru/icefall_asr_timit_tdnn_ligru_ctc/data/lang_phone/HLG.pt \
--G ./tmp-ligru/icefall_asr_timit_tdnn_ligru_ctc/data/lm/G_4_gram.pt \
--ngram-lm-scale 0.1 \
./tmp-ligru/icefall_asr_timit_tdnn_ligru_ctc/test_waves/FDHC0_SI1559.WAV
./tmp-ligru/icefall_asr_timit_tdnn_ligru_ctc/test_waves/FELC0_SI756.WAV
./tmp-ligru/icefall_asr_timit_tdnn_ligru_ctc/test_waves/FDHC0_SI1559.WAV
./tmp-ligru/icefall_asr_timit_tdnn_ligru_ctc/test_waves/FELC0_SI756.WAV
./tmp-ligru/icefall_asr_timit_tdnn_ligru_ctc/test_waves/FMGD0_SI1564.WAV
The decoding output is:
@ -378,7 +378,7 @@ The decoding output is:
2021-11-08 20:37:54,715 INFO [pretrained.py:210] Reading sound files: ['./tmp-ligru/icefall_asr_timit_tdnn_ligru_ctc/test_waves/FDHC0_SI1559.WAV', './tmp-ligru/icefall_asr_timit_tdnn_ligru_ctc/test_waves/FELC0_SI756.WAV', './tmp-ligru/icefall_asr_timit_tdnn_ligru_ctc/test_waves/FMGD0_SI1564.WAV']
2021-11-08 20:37:54,720 INFO [pretrained.py:216] Decoding started
2021-11-08 20:37:55,808 INFO [pretrained.py:251] Use HLG decoding + LM rescoring
2021-11-08 20:37:56,348 INFO [pretrained.py:267]
2021-11-08 20:37:56,348 INFO [pretrained.py:267]
./tmp-ligru/icefall_asr_timit_tdnn_ligru_ctc/test_waves/FDHC0_SI1559.WAV:
sil dh ih sh uw ah l iy v iy z ah sil p r aa sil k s ih m ey dx ih sil d w uh dx iy w ih s f iy l iy ng w ih th ih n ih m s eh l f sil jh

24
docs/source/recipes/timit/tdnn_lstm_ctc.rst → docs/source/recipes/Non-streaming-ASR/timit/tdnn_lstm_ctc.rst
View File

@ -148,8 +148,8 @@ Some commonly used options are:
$ ./tdnn_lstm_ctc/decode.py --epoch 25 --avg 10
uses the average of ``epoch-16.pt``, ``epoch-17.pt``, ``epoch-18.pt``,
``epoch-19.pt``, ``epoch-20.pt``, ``epoch-21.pt``, ``epoch-22.pt``,
uses the average of ``epoch-16.pt``, ``epoch-17.pt``, ``epoch-18.pt``,
``epoch-19.pt``, ``epoch-20.pt``, ``epoch-21.pt``, ``epoch-22.pt``,
``epoch-23.pt``, ``epoch-24.pt`` and ``epoch-25.pt``
for decoding.
@ -315,13 +315,13 @@ To decode with ``1best`` method, we can use:
.. code-block:: bash
./tdnn_lstm_ctc/pretrained.py
./tdnn_lstm_ctc/pretrained.py
--method 1best
--checkpoint ./tmp-lstm/icefall_asr_timit_tdnn_lstm_ctc/exp/pretrained_average_16_25.pt
--words-file ./tmp-lstm/icefall_asr_timit_tdnn_lstm_ctc/data/lang_phone/words.txt
--HLG ./tmp-lstm/icefall_asr_timit_tdnn_lstm_ctc/data/lang_phone/HLG.pt
./tmp-lstm/icefall_asr_timit_tdnn_lstm_ctc/test_waves/FDHC0_SI1559.WAV
./tmp-lstm/icefall_asr_timit_tdnn_lstm_ctc/test_waves/FELC0_SI756.WAV
--checkpoint ./tmp-lstm/icefall_asr_timit_tdnn_lstm_ctc/exp/pretrained_average_16_25.pt
--words-file ./tmp-lstm/icefall_asr_timit_tdnn_lstm_ctc/data/lang_phone/words.txt
--HLG ./tmp-lstm/icefall_asr_timit_tdnn_lstm_ctc/data/lang_phone/HLG.pt
./tmp-lstm/icefall_asr_timit_tdnn_lstm_ctc/test_waves/FDHC0_SI1559.WAV
./tmp-lstm/icefall_asr_timit_tdnn_lstm_ctc/test_waves/FELC0_SI756.WAV
./tmp-lstm/icefall_asr_timit_tdnn_lstm_ctc/test_waves/FMGD0_SI1564.WAV
The output is:
@ -335,7 +335,7 @@ The output is:
2021-11-08 21:02:53,827 INFO [pretrained.py:210] Reading sound files: ['./tmp-lstm/icefall_asr_timit_tdnn_lstm_ctc/test_waves/FDHC0_SI1559.WAV', './tmp-lstm/icefall_asr_timit_tdnn_lstm_ctc/test_waves/FELC0_SI756.WAV', './tmp-lstm/icefall_asr_timit_tdnn_lstm_ctc/test_waves/FMGD0_SI1564.WAV']
2021-11-08 21:02:53,831 INFO [pretrained.py:216] Decoding started
2021-11-08 21:02:54,380 INFO [pretrained.py:246] Use HLG decoding
2021-11-08 21:02:54,387 INFO [pretrained.py:267]
2021-11-08 21:02:54,387 INFO [pretrained.py:267]
./tmp-lstm/icefall_asr_timit_tdnn_lstm_ctc/test_waves/FDHC0_SI1559.WAV:
sil dh ih sh uw ah l iy v iy z ih sil p r aa sil k s ih m ey dx ih sil d w uh dx iy w ih s f iy l iy w ih th ih n ih m s eh l f sil jh
@ -360,8 +360,8 @@ To decode with ``whole-lattice-rescoring`` methond, you can use
--HLG ./tmp-lstm/icefall_asr_timit_tdnn_lstm_ctc/data/lang_phone/HLG.pt \
--G ./tmp-lstm/icefall_asr_timit_tdnn_lstm_ctc/data/lm/G_4_gram.pt \
--ngram-lm-scale 0.08 \
./tmp-lstm/icefall_asr_timit_tdnn_lstm_ctc/test_waves/FDHC0_SI1559.WAV
./tmp-lstm/icefall_asr_timit_tdnn_lstm_ctc/test_waves/FELC0_SI756.WAV
./tmp-lstm/icefall_asr_timit_tdnn_lstm_ctc/test_waves/FDHC0_SI1559.WAV
./tmp-lstm/icefall_asr_timit_tdnn_lstm_ctc/test_waves/FELC0_SI756.WAV
./tmp-lstm/icefall_asr_timit_tdnn_lstm_ctc/test_waves/FMGD0_SI1564.WAV
The decoding output is:
@ -376,7 +376,7 @@ The decoding output is:
2021-11-08 20:05:26,978 INFO [pretrained.py:210] Reading sound files: ['./tmp-lstm/icefall_asr_timit_tdnn_lstm_ctc/test_waves/FDHC0_SI1559.WAV', './tmp-lstm/icefall_asr_timit_tdnn_lstm_ctc/test_waves/FELC0_SI756.WAV', './tmp-lstm/icefall_asr_timit_tdnn_lstm_ctc/test_waves/FMGD0_SI1564.WAV']
2021-11-08 20:05:26,981 INFO [pretrained.py:216] Decoding started
2021-11-08 20:05:27,519 INFO [pretrained.py:251] Use HLG decoding + LM rescoring
2021-11-08 20:05:27,878 INFO [pretrained.py:267]
2021-11-08 20:05:27,878 INFO [pretrained.py:267]
./tmp-lstm/icefall_asr_timit_tdnn_lstm_ctc/test_waves/FDHC0_SI1559.WAV:
sil dh ih sh uw l iy v iy z ih sil p r aa sil k s ah m ey dx ih sil w uh dx iy w ih s f iy l ih ng w ih th ih n ih m s eh l f sil jh

0
docs/source/recipes/yesno/images/tdnn-tensorboard-log.png → docs/source/recipes/Non-streaming-ASR/yesno/images/tdnn-tensorboard-log.png
View File

Side by Side Swipe Overlay

Before

Width:  |  Height:  |  Size: 121 KiB

After

Width:  |  Height:  |  Size: 121 KiB

0
docs/source/recipes/yesno/index.rst → docs/source/recipes/Non-streaming-ASR/yesno/index.rst
View File

0
docs/source/recipes/yesno/tdnn.rst → docs/source/recipes/Non-streaming-ASR/yesno/tdnn.rst
View File

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