Merge pull request #977 from mmglove/fix_dsp_bk

update pwgan benchmark case

tests/benchmark/pwgan/run_all.sh

@@ -1,8 +1,10 @@
 #!/usr/bin/env bash
 
+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
+
 stage=0
 stop_stage=100
-
+sed -i '/set\ -xe/d' run_benchmark.sh
 # 提供可稳定复现性能的脚本，默认在标准docker环境内py37执行： paddlepaddle/paddle:latest-gpu-cuda10.1-cudnn7  paddle=2.1.2  py=37
 # 执行目录：需说明
 cd ../../../
@@ -33,20 +35,22 @@ if [ ${stage} -le 2 ] && [ ${stop_stage} -ge 2 ]; then
 fi
 # 3 批量运行（如不方便批量，1，2需放到单个模型中）
 if [ ${stage} -le 3 ] && [ ${stop_stage} -ge 3 ]; then
-      model_mode_list=(pwg)
+      model_mode_list=(pwgan)
       fp_item_list=(fp32)
       # 满 bs 是 26
       bs_item_list=(6 26)
       for model_mode in ${model_mode_list[@]}; do
             for fp_item in ${fp_item_list[@]}; do
             for bs_item in ${bs_item_list[@]}; do
-                  echo "index is speed, 1gpus, begin, ${model_name}"
+                  log_name=speech_${model_mode}_bs${bs_item}_${fp_item}   # 如:clas_MobileNetv1_mp_bs32_fp32_8
+                  echo "index is speed, 1gpus, begin, ${log_name}"
                   run_mode=sp
-                  CUDA_VISIBLE_DEVICES=0 bash tests/benchmark/pwgan/run_benchmark.sh ${run_mode} ${bs_item} ${fp_item} 100 ${model_mode}     #  (5min)
+                  CUDA_VISIBLE_DEVICES=0 bash tests/benchmark/pwgan/run_benchmark.sh ${run_mode} ${bs_item} ${fp_item} 100 ${model_mode}   | tee ${log_path}/${log_name}_speed_1gpus 2>&1   #  (5min)
                   sleep 60
-                  echo "index is speed, 8gpus, run_mode is multi_process, begin, ${model_name}"
+                  log_name=speech_${model_mode}_bs${bs_item}_${fp_item}   # 如:clas_MobileNetv1_mp_bs32_fp32_8
+                  echo "index is speed, 8gpus, run_mode is multi_process, begin, ${log_name}"
                   run_mode=mp
-                  CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 bash tests/benchmark/pwgan/run_benchmark.sh ${run_mode} ${bs_item} ${fp_item} 100 ${model_mode} 
+                  CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 bash tests/benchmark/pwgan/run_benchmark.sh ${run_mode} ${bs_item} ${fp_item} 100 ${model_mode} | tee ${log_path}/${log_name}_speed_8gpus8p 2>&1   #
                   sleep 60
                   done
             done

tests/benchmark/pwgan/run_benchmark.sh

@@ -7,14 +7,22 @@ function _set_params(){
     batch_size=${2:-"8"}
     fp_item=${3:-"fp32"}        # fp32|fp16
     max_iter=${4:-"500"}        # 可选，如果需要修改代码提前中断
-    model_name=${5:-"model_name"}
+    model_item=${5:-"model_item"}
     run_log_path=${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"
+    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)
+    keyword="avg_ips:"                 # 解析日志，筛选出数据所在行的关键字                                             (必填)
+    index="1"
+    model_name=${model_item}_bs${batch_size}_${fp_item}
 #   以下不用修改   
     device=${CUDA_VISIBLE_DEVICES//,/ }
     arr=(${device})
     num_gpu_devices=${#arr[*]}
-    log_file=${run_log_path}/${model_name}_${run_mode}_bs${batch_size}_${fp_item}_${num_gpu_devices}
+    log_file=${run_log_path}/${model_item}_${run_mode}_bs${batch_size}_${fp_item}_${num_gpu_devices}
 }
 function _train(){
     echo "Train on ${num_gpu_devices} GPUs"
@@ -52,5 +60,8 @@ function _train(){
     fi
 }
  
+source ${BENCHMARK_ROOT}/scripts/run_model.sh   # 在该脚本中会对符合benchmark规范的log使用analysis.py 脚本进行性能数据解析;该脚本在连调时可从benchmark repo中下载https://github.com/PaddlePaddle/benchmark/blob/master/scripts/run_model.sh;如果不联调只想要产出训练log可以注掉本行,提交时需打开
 _set_params $@
-_train
+# _train       # 如果只想产出训练log,不解析,可取消注释
+_run     # 该函数在run_model.sh中,执行时会调用_train; 如果不联调只想要产出训练log可以注掉本行,提交时需打开
+