mirrors/icefall
1
0
Fork 0
mirror of https://github.com/k2-fsa/icefall.git synced 2025-09-06 23:54:17 +00:00
Code Issues Packages Projects Releases Wiki Activity

Add test for zipformer ctc HLG decoding

Browse Source
This commit is contained in:
Fangjun Kuang 2023-10-02 12:23:55 +08:00
parent ceff45316c
commit 880a8867cb
2 changed files with 38 additions and 10 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

40
.github/scripts/run-pre-trained-conformer-ctc.sh → .github/scripts/run-pre-trained-ctc.sh vendored
View File

@ -10,7 +10,37 @@ log() {
pushd egs/librispeech/ASR pushd egs/librispeech/ASR
# repo_url=https://github.com/csukuangfj/icefall-asr-conformer-ctc-bpe-500 repo_url=https://huggingface.co/csukuangfj/sherpa-onnx-zipformer-ctc-en-2023-10-02
log "Downloading pre-trained model from $repo_url"
git lfs install
git clone $repo_url
repo=$(basename $repo_url)
log "Display test files"
tree $repo/
ls -lh $repo/test_wavs/*.wav
log "CTC greedy search"
./zipformer/onnx_pretrained_ctc.py \
--nn-model $repo/model.onnx \
--tokens $repo/tokens.txt \
$repo/test_wavs/0.wav \
$repo/test_wavs/1.wav \
$repo/test_wavs/2.wav
log "CTC HLG decoding"
./zipformer/onnx_pretrained_ctc_hlg.py \
--nn-model $repo/model.onnx \
--words $repo/words.txt \
--HLG $repo/HLG.fst \
$repo/test_wavs/0.wav \
$repo/test_wavs/1.wav \
$repo/test_wavs/2.wav
rm -rf $repo
repo_url=https://huggingface.co/csukuangfj/icefall-asr-librispeech-conformer-ctc-jit-bpe-500-2021-11-09 repo_url=https://huggingface.co/csukuangfj/icefall-asr-librispeech-conformer-ctc-jit-bpe-500-2021-11-09
log "Downloading pre-trained model from $repo_url" log "Downloading pre-trained model from $repo_url"
GIT_LFS_SKIP_SMUDGE=1 git clone $repo_url GIT_LFS_SKIP_SMUDGE=1 git clone $repo_url
@ -128,7 +158,9 @@ repo=$(basename $repo_url)
pushd $repo pushd $repo
git lfs pull --include "exp/pretrained.pt" git lfs pull --include "exp/pretrained.pt"
git lfs pull --include "data/lm/G_3_gram_char.fst.txt" git lfs pull --include "data/lang_char/H.fst"
git lfs pull --include "data/lang_char/HL.fst"
git lfs pull --include "data/lang_char/HLG.fst"
popd popd
@ -153,10 +185,6 @@ popd
ls -lh $repo/exp ls -lh $repo/exp
log "Generating H.fst, HL.fst"
./local/prepare_lang_fst.py --lang-dir $repo/data/lang_char --ngram-G $repo/data/lm/G_3_gram_char.fst.txt
ls -lh $repo/data/lang_char ls -lh $repo/data/lang_char
log "Decoding with H on CPU with OpenFst" log "Decoding with H on CPU with OpenFst"

8
.github/workflows/run-pretrained-conformer-ctc.yml → .github/workflows/run-pretrained-ctc.yml vendored
View File

@ -14,7 +14,7 @@
# See the License for the specific language governing permissions and # See the License for the specific language governing permissions and
# limitations under the License. # limitations under the License.
name: run-pre-trained-conformer-ctc name: run-pre-trained-ctc
on: on:
push: push:
@ -31,12 +31,12 @@ on:
default: 'y' default: 'y'
concurrency: concurrency:
group: run_pre_trained_conformer_ctc-${{ github.ref }} group: run_pre_trained_ctc-${{ github.ref }}
cancel-in-progress: true cancel-in-progress: true
jobs: jobs:
run_pre_trained_conformer_ctc: run_pre_trained_conformer_ctc:
if: github.event.label.name == 'ready' || github.event_name == 'push' || github.event.inputs.test-run == 'y' if: github.event.label.name == 'ready' || github.event_name == 'push' || github.event.inputs.test-run == 'y' || github.event.label.name == 'ctc'
runs-on: ${{ matrix.os }} runs-on: ${{ matrix.os }}
strategy: strategy:
matrix: matrix:
@ -84,4 +84,4 @@ jobs:
export PYTHONPATH=$PWD:$PYTHONPATH export PYTHONPATH=$PWD:$PYTHONPATH
export PYTHONPATH=~/tmp/kaldifeat/kaldifeat/python:$PYTHONPATH export PYTHONPATH=~/tmp/kaldifeat/kaldifeat/python:$PYTHONPATH
export PYTHONPATH=~/tmp/kaldifeat/build/lib:$PYTHONPATH export PYTHONPATH=~/tmp/kaldifeat/build/lib:$PYTHONPATH
.github/scripts/run-pre-trained-conformer-ctc.sh .github/scripts/run-pre-trained-ctc.sh