Begin to support CTC models (#119)

Please see https://k2-fsa.github.io/sherpa/onnx/pretrained_models/offline-ctc/nemo/index.html for a list of pre-trained CTC models from NeMo.

.github/scripts/test-offline-ctc.sh

@@ -0,0 +1,47 @@
+#!/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]}) $*"
+}
+
+echo "EXE is $EXE"
+echo "PATH: $PATH"
+
+which $EXE
+
+log "------------------------------------------------------------"
+log "Run Citrinet (stt_en_citrinet_512, English)"
+log "------------------------------------------------------------"
+
+repo_url=http://huggingface.co/csukuangfj/sherpa-onnx-nemo-ctc-en-citrinet-512
+log "Start testing ${repo_url}"
+repo=$(basename $repo_url)
+log "Download pretrained model and test-data from $repo_url"
+
+GIT_LFS_SKIP_SMUDGE=1 git clone $repo_url
+pushd $repo
+git lfs pull --include "*.onnx"
+ls -lh *.onnx
+popd
+
+time $EXE \
+  --tokens=$repo/tokens.txt \
+  --nemo-ctc-model=$repo/model.onnx \
+  --num-threads=2 \
+  $repo/test_wavs/0.wav \
+  $repo/test_wavs/1.wav \
+  $repo/test_wavs/8k.wav
+
+time $EXE \
+  --tokens=$repo/tokens.txt \
+  --nemo-ctc-model=$repo/model.int8.onnx \
+  --num-threads=2 \
+  $repo/test_wavs/0.wav \
+  $repo/test_wavs/1.wav \
+  $repo/test_wavs/8k.wav
+
+rm -rf $repo

.github/scripts/test-python.sh

@@ -95,6 +95,8 @@ python3 ./python-api-examples/offline-decode-files.py \
 
 python3 sherpa-onnx/python/tests/test_offline_recognizer.py --verbose
 
+rm -rf $repo
+
 log "Test non-streaming paraformer models"
 
 pushd $dir
@@ -128,3 +130,39 @@ python3 ./python-api-examples/offline-decode-files.py \
   $repo/test_wavs/8k.wav
 
 python3 sherpa-onnx/python/tests/test_offline_recognizer.py --verbose
+
+rm -rf $repo
+
+log "Test non-streaming NeMo CTC models"
+
+pushd $dir
+repo_url=http://huggingface.co/csukuangfj/sherpa-onnx-nemo-ctc-en-citrinet-512
+
+log "Start testing ${repo_url}"
+repo=$dir/$(basename $repo_url)
+log "Download pretrained model and test-data from $repo_url"
+
+GIT_LFS_SKIP_SMUDGE=1 git clone $repo_url
+cd $repo
+git lfs pull --include "*.onnx"
+popd
+
+ls -lh $repo
+
+python3 ./python-api-examples/offline-decode-files.py \
+  --tokens=$repo/tokens.txt \
+  --nemo-ctc=$repo/model.onnx \
+  $repo/test_wavs/0.wav \
+  $repo/test_wavs/1.wav \
+  $repo/test_wavs/8k.wav
+
+python3 ./python-api-examples/offline-decode-files.py \
+  --tokens=$repo/tokens.txt \
+  --nemo-ctc=$repo/model.int8.onnx \
+  $repo/test_wavs/0.wav \
+  $repo/test_wavs/1.wav \
+  $repo/test_wavs/8k.wav
+
+python3 sherpa-onnx/python/tests/test_offline_recognizer.py --verbose
+
+rm -rf $repo