Merge pull request #1572 from zh794390558/egs

[speechx] refactor speechx egs

speechx/.gitignore

+tools/valgrind*

speechx/README.md

+# SpeechX -- Speech Inference All in One
+
+> Test under `Ubuntu 16.04.7 LTS`.
+
+## Build
+
+```
+./build.sh
+```l
+
+## Valgrind
+
+> If using docker please check `--privileged` is set when `docker run`.
+
+1. Fatal error at startup: a function redirection which is mandatory for this platform-tool combination cannot be set up
+```
+apt-get install libc6-dbg
+```
+
+```
+pushd tools
+./setup_valgrind.sh
+popd
+```

speechx/examples/.gitignore

+*.ark
+paddle_asr_model/

speechx/examples/path.sh → speechx/examples/decoder/path.sh

@@ -2,7 +2,6 @@
 
 SPEECHX_ROOT=$PWD/../..
 SPEECHX_EXAMPLES=$SPEECHX_ROOT/build/examples
-SPEECHX_BIN=$SPEECHX_EXAMPLES/nnet:$SPEECHX_EXAMPLES/decoder:$SPEECHX_EXAMPLES/feat
 
 SPEECHX_TOOLS=$SPEECHX_ROOT/tools
 TOOLS_BIN=$SPEECHX_TOOLS/valgrind/install/bin
@@ -11,4 +10,5 @@ TOOLS_BIN=$SPEECHX_TOOLS/valgrind/install/bin
 
 export LC_AL=C
 
+SPEECHX_BIN=$SPEECHX_EXAMPLES/decoder
 export PATH=$PATH:$SPEECHX_BIN:$TOOLS_BIN

speechx/examples/decoder/run.sh

@@ -2,18 +2,19 @@
 set +x
 set -e
 
+. path.sh
+
 # 1. compile
-if [ ! -d ../../build/examples ]; then
-    cd ../..
+if [ ! -d ${SPEECHX_EXAMPLES} ]; then
+    pushd ${SPEECHX_ROOT} 
     bash build.sh
-    cd -
+    popd
 fi
 
-. ../path.sh
 
 # 2. download model
 if [ ! -d ../paddle_asr_model ]; then
-    wget https://paddlespeech.bj.bcebos.com/s2t/paddle_asr_online/paddle_asr_model.tar.gz
+    wget -c https://paddlespeech.bj.bcebos.com/s2t/paddle_asr_online/paddle_asr_model.tar.gz
     tar xzfv paddle_asr_model.tar.gz
     mv ./paddle_asr_model ../
     # produce wav scp
@@ -25,8 +26,15 @@ feat_wspecifier=./feats.ark
 cmvn=./cmvn.ark
 
 # 3. run feat
-linear_spectrogram_main --wav_rspecifier=scp:$model_dir/wav.scp --feature_wspecifier=ark:$feat_wspecifier --cmvn_write_path=$cmvn
+linear_spectrogram_main \
+    --wav_rspecifier=scp:$model_dir/wav.scp \
+    --feature_wspecifier=ark,t:$feat_wspecifier \
+    --cmvn_write_path=$cmvn
 
 # 4. run decoder
-offline_decoder_main --feature_respecifier=ark:$feat_wspecifier --model_path=$model_dir/avg_1.jit.pdmodel --param_path=$model_dir/avg_1.jit.pdparams --dict_file=$model_dir/vocab.txt --lm_path=$model_dir/avg_1.jit.klm
-
+offline_decoder_main \
+    --feature_respecifier=ark:$feat_wspecifier \
+    --model_path=$model_dir/avg_1.jit.pdmodel \
+    --param_path=$model_dir/avg_1.jit.pdparams \
+    --dict_file=$model_dir/vocab.txt \
+    --lm_path=$model_dir/avg_1.jit.klm
\ No newline at end of file

speechx/examples/decoder/valgrind.sh

@@ -5,16 +5,22 @@
 set +x
 set -e
 
-if [ ! -d ../../tools/valgrind/install ]; then
+. ./path.sh
+
+if [ ! -d ${SPEECHX_TOOLS}/valgrind/install ]; then
   echo "please install valgrind in the speechx tools dir.\n" 
   exit 1
 fi
 
-. ../path.sh
-
 model_dir=../paddle_asr_model
 feat_wspecifier=./feats.ark
 cmvn=./cmvn.ark
 
-valgrind --tool=memcheck --track-origins=yes --leak-check=full --show-leak-kinds=all offline_decoder_main --feature_respecifier=ark:$feat_wspecifier --model_path=$model_dir/avg_1.jit.pdmodel --param_path=$model_dir/avg_1.jit.pdparams --dict_file=$model_dir/vocab.txt --lm_path=$model_dir/avg_1.jit.klm
+valgrind --tool=memcheck --track-origins=yes --leak-check=full --show-leak-kinds=all \
+  offline_decoder_main \
+  --feature_respecifier=ark:$feat_wspecifier \
+  --model_path=$model_dir/avg_1.jit.pdmodel \
+  --param_path=$model_dir/avg_1.jit.pdparams \
+  --dict_file=$model_dir/vocab.txt \
+  --lm_path=$model_dir/avg_1.jit.klm
 

speechx/examples/feat/path.sh

+# This contains the locations of binarys build required for running the examples.
+
+SPEECHX_ROOT=$PWD/../..
+SPEECHX_EXAMPLES=$SPEECHX_ROOT/build/examples
+
+SPEECHX_TOOLS=$SPEECHX_ROOT/tools
+TOOLS_BIN=$SPEECHX_TOOLS/valgrind/install/bin
+
+[ -d $SPEECHX_EXAMPLES ] || { echo "Error: 'build/examples' directory not found. please ensure that the project build successfully"; }
+
+export LC_AL=C
+
+SPEECHX_BIN=$SPEECHX_EXAMPLES/feat
+export PATH=$PATH:$SPEECHX_BIN:$TOOLS_BIN

speechx/examples/feat/run.sh

@@ -2,15 +2,15 @@
 set +x
 set -e
 
+. ./path.sh
+
 # 1. compile
-if [ ! -d ../../build/examples ]; then
-    cd ../..
+if [ ! -d ${SPEECHX_EXAMPLES} ]; then
+    pushd ${SPEECHX_ROOT} 
     bash build.sh
-    cd -
+    popd
 fi
 
-. ../path.sh
-
 # 2. download model
 if [ ! -d ../paddle_asr_model ]; then
     wget https://paddlespeech.bj.bcebos.com/s2t/paddle_asr_online/paddle_asr_model.tar.gz
@@ -25,4 +25,7 @@ feat_wspecifier=./feats.ark
 cmvn=./cmvn.ark
 
 # 3. run feat
-linear_spectrogram_main --wav_rspecifier=scp:$model_dir/wav.scp --feature_wspecifier=ark,t:$feat_wspecifier --cmvn_write_path=$cmvn
+linear_spectrogram_main \
+    --wav_rspecifier=scp:$model_dir/wav.scp \
+    --feature_wspecifier=ark,t:$feat_wspecifier \
+    --cmvn_write_path=$cmvn

speechx/examples/feat/valgrind.sh

@@ -5,16 +5,20 @@
 set +x
 set -e
 
-if [ ! -d ../../tools/valgrind/install ]; then
+. ./path.sh
+
+if [ ! -d ${SPEECHX_TOOLS}/valgrind/install ]; then
   echo "please install valgrind in the speechx tools dir.\n" 
   exit 1
 fi
 
-. ../path.sh
-
 model_dir=../paddle_asr_model
 feat_wspecifier=./feats.ark
 cmvn=./cmvn.ark
 
-valgrind --tool=memcheck --track-origins=yes --leak-check=full --show-leak-kinds=all linear_spectrogram_main --wav_rspecifier=scp:$model_dir/wav.scp --feature_wspecifier=ark,t:$feat_wspecifier --cmvn_write_path=$cmvn
+valgrind --tool=memcheck --track-origins=yes --leak-check=full --show-leak-kinds=all \
+  linear_spectrogram_main \
+  --wav_rspecifier=scp:$model_dir/wav.scp \
+  --feature_wspecifier=ark,t:$feat_wspecifier \
+  --cmvn_write_path=$cmvn
 

speechx/examples/nnet/path.sh

+# This contains the locations of binarys build required for running the examples.
+
+SPEECHX_ROOT=$PWD/../..
+SPEECHX_EXAMPLES=$SPEECHX_ROOT/build/examples
+
+SPEECHX_TOOLS=$SPEECHX_ROOT/tools
+TOOLS_BIN=$SPEECHX_TOOLS/valgrind/install/bin
+
+[ -d $SPEECHX_EXAMPLES ] || { echo "Error: 'build/examples' directory not found. please ensure that the project build successfully"; }
+
+export LC_AL=C
+
+SPEECHX_BIN=$SPEECHX_EXAMPLES/nnet
+export PATH=$PATH:$SPEECHX_BIN:$TOOLS_BIN

speechx/examples/nnet/run.sh

@@ -2,15 +2,15 @@
 set +x
 set -e
 
+. path.sh
+
 # 1. compile
-if [ ! -d ../../build/examples ]; then
-    cd ../..
+if [ ! -d ${SPEECHX_EXAMPLES} ]; then
+    pushd ${SPEECHX_ROOT} 
     bash build.sh
-    cd -
+    popd
 fi
 
-. ../path.sh
-
 # 2. download model
 if [ ! -d ../paddle_asr_model ]; then
     wget https://paddlespeech.bj.bcebos.com/s2t/paddle_asr_online/paddle_asr_model.tar.gz
@@ -23,5 +23,7 @@ fi
 model_dir=../paddle_asr_model
 
 # 4. run decoder
-pp-model-test --model_path=$model_dir/avg_1.jit.pdmodel --param_path=$model_dir/avg_1.jit.pdparams
+pp-model-test \
+    --model_path=$model_dir/avg_1.jit.pdmodel \
+    --param_path=$model_dir/avg_1.jit.pdparams
 

speechx/examples/nnet/valgrind.sh

@@ -5,14 +5,16 @@
 set +x
 set -e
 
-if [ ! -d ../../tools/valgrind/install ]; then
+. ./path.sh
+
+if [ ! -d ${SPEECHX_TOOLS}/valgrind/install ]; then
   echo "please install valgrind in the speechx tools dir.\n" 
   exit 1
 fi
 
-. ../path.sh
-
 model_dir=../paddle_asr_model
 
-valgrind --tool=memcheck --track-origins=yes --leak-check=full --show-leak-kinds=all pp-model-test --model_path=$model_dir/avg_1.jit.pdmodel --param_path=$model_dir/avg_1.jit.pdparams
-
+valgrind --tool=memcheck --track-origins=yes --leak-check=full --show-leak-kinds=all \
+  pp-model-test \
+  --model_path=$model_dir/avg_1.jit.pdmodel \
+  --param_path=$model_dir/avg_1.jit.pdparams
\ No newline at end of file

speechx/tools/setup_valgrind.sh

@@ -11,7 +11,7 @@ url=https://sourceware.org/pub/valgrind/valgrind-3.18.1.tar.bz2
 if [ -f $tarball ]; then
   echo "use the $tarball have downloaded."
 else
-  wget -t3 --no-check-certificate $url
+  wget -c -t3 --no-check-certificate $url
 fi
 
 tar xjfv $tarball