@ -1,4 +1,6 @@
cd ../../../
pip install -e . # 安装pdspeech
cd -
#Enter the example dir
pushd ../../../examples/aishell/s1
@ -3,7 +3,6 @@
# 执行目录:需说明
CUR_DIR=${PWD} # PaddleSpeech/tests/benchmark/conformer
log_path=${LOG_PATH_INDEX_DIR:-$(pwd)} # benchmark系统指定该参数,不需要跑profile时,log_path指向存speed的目录
cd ${CUR_DIR}
sed -i '/set\ -xe/d' run_benchmark.sh
@ -17,7 +17,7 @@ function _set_params(){
run_log_path=${TRAIN_LOG_DIR:-$(pwd)} # TRAIN_LOG_DIR 后续QA设置该参数
# 添加日志解析需要的参数
base_batch_size=${batch_size}
mission_name="语音合成"
mission_name="语音识别"
direction_id="1"
ips_unit="sequences/sec"
skip_steps=10 # 解析日志,有些模型前几个step耗时长,需要跳过 (必填)