From 67424bedc5016db89f4940a4c17ade503ca178d0 Mon Sep 17 00:00:00 2001
From: Hui Zhang <zhtclz@foxmail.com>
Date: Thu, 27 Oct 2022 11:08:31 +0000
Subject: [PATCH] add quant model and script

---
 .../wenetspeech/local/recognizer_quant.sh     | 36 +++++++++++++++++++
 speechx/examples/u2pp_ol/wenetspeech/run.sh   | 13 ++++++-
 2 files changed, 48 insertions(+), 1 deletion(-)
 create mode 100755 speechx/examples/u2pp_ol/wenetspeech/local/recognizer_quant.sh

diff --git a/speechx/examples/u2pp_ol/wenetspeech/local/recognizer_quant.sh b/speechx/examples/u2pp_ol/wenetspeech/local/recognizer_quant.sh
new file mode 100755
index 000000000..1ce403a3c
--- /dev/null
+++ b/speechx/examples/u2pp_ol/wenetspeech/local/recognizer_quant.sh
@@ -0,0 +1,36 @@
+#!/bin/bash
+set -e
+
+data=data
+exp=exp
+nj=20
+
+. utils/parse_options.sh
+
+mkdir -p $exp
+ckpt_dir=./data/model
+model_dir=$ckpt_dir/asr1_chunk_conformer_u2pp_wenetspeech_static_quant_1.3.0.model/
+aishell_wav_scp=aishell_test.scp
+text=$data/test/text
+
+./local/split_data.sh $data $data/$aishell_wav_scp $aishell_wav_scp $nj
+
+utils/run.pl JOB=1:$nj $data/split${nj}/JOB/recognizer.quant.log \
+u2_recognizer_main \
+    --use_fbank=true \
+    --num_bins=80 \
+    --cmvn_file=$exp/cmvn.ark \
+    --model_path=$model_dir/export \
+    --vocab_path=$model_dir/unit.txt \
+    --nnet_decoder_chunk=16 \
+    --receptive_field_length=7 \
+    --subsampling_rate=4 \
+    --wav_rspecifier=scp:$data/split${nj}/JOB/${aishell_wav_scp} \
+    --result_wspecifier=ark,t:$data/split${nj}/JOB/recognizer.quant.rsl.ark
+
+
+cat $data/split${nj}/*/recognizer.quant.rsl.ark > $exp/aishell.recognizer.quant.rsl
+utils/compute-wer.py --char=1 --v=1 $text $exp/aishell.recognizer.quant.rsl > $exp/aishell.recognizer.quant.err
+echo "recognizer quant test have finished!!!"
+echo "please checkout in $exp/aishell.recognizer.quant.err"
+tail -n 7 $exp/aishell.recognizer.quant.err
diff --git a/speechx/examples/u2pp_ol/wenetspeech/run.sh b/speechx/examples/u2pp_ol/wenetspeech/run.sh
index 4659ca0f7..2bc855dec 100755
--- a/speechx/examples/u2pp_ol/wenetspeech/run.sh
+++ b/speechx/examples/u2pp_ol/wenetspeech/run.sh
@@ -28,7 +28,7 @@ model_dir=$ckpt_dir/asr1_chunk_conformer_u2pp_wenetspeech_static_1.3.0.model/
 
 
 if [ ${stage} -le 0 ] && [ ${stop_stage} -ge 0 ];then
-    #  download model
+    #  download u2pp model
     if [ ! -f $ckpt_dir/asr1_chunk_conformer_u2pp_wenetspeech_static_1.3.0.model.tar.gz ]; then
         mkdir -p $ckpt_dir
         pushd $ckpt_dir
@@ -39,6 +39,17 @@ if [ ${stage} -le 0 ] && [ ${stop_stage} -ge 0 ];then
         popd
     fi
 
+    # download u2pp quant model
+    if [ ! -f $ckpt_dir/asr1_chunk_conformer_u2pp_wenetspeech_static_quant_1.3.0.model.tar.gz ]; then
+        mkdir -p $ckpt_dir
+        pushd $ckpt_dir
+
+        wget -c https://paddlespeech.bj.bcebos.com/s2t/wenetspeech/asr1/static/asr1_chunk_conformer_u2pp_wenetspeech_static_quant_1.3.0.model.tar.gz
+        tar xzfv asr1_chunk_conformer_u2pp_wenetspeech_static_quant_1.3.0.model.tar.gz
+
+        popd
+    fi
+
     # test wav scp
     if [ ! -f data/wav.scp ]; then
         mkdir -p $data