update_scripts_benchmark

benchmark/run_all.sh

@@ -6,7 +6,8 @@
 # 2 拷贝该模型需要数据、预训练模型
 # 3 批量运行（如不方便批量，1，2需放到单个模型中）
  
-model_mode_list=(MobileNetV1 MobileNetV2 MobileNetV3_large_x1_0 EfficientNetB0 ShuffleNetV2_x1_0 DenseNet121 HRNet_W48_C SwinTransformer_tiny_patch4_window7_224 alt_gvt_base)
+model_mode_list=(MobileNetV1 MobileNetV2 MobileNetV3_large_x1_0 ShuffleNetV2_x1_0 HRNet_W48_C SwinTransformer_tiny_patch4_window7_224 alt_gvt_base)    # benchmark 监控模型列表
+#model_mode_list=(MobileNetV1 MobileNetV2 MobileNetV3_large_x1_0 EfficientNetB0 ShuffleNetV2_x1_0 DenseNet121 HRNet_W48_C SwinTransformer_tiny_patch4_window7_224 alt_gvt_base)   # 该脚本支持列表
 fp_item_list=(fp32)
 bs_list=(32 64 96 128)
 for model_mode in ${model_mode_list[@]}; do
@@ -14,11 +15,11 @@ for model_mode in ${model_mode_list[@]}; do
           for bs_item in ${bs_list[@]};do
 	    echo "index is speed, 1gpus, begin, ${model_name}"
 	    run_mode=sp
-	    CUDA_VISIBLE_DEVICES=0 bash benchmark/run_benchmark.sh ${run_mode} ${bs_item} ${fp_item} 10 ${model_mode}     #  (5min)
+	    CUDA_VISIBLE_DEVICES=0 bash benchmark/run_benchmark.sh ${run_mode} ${bs_item} ${fp_item} 1 ${model_mode} | tee ${log_path}/clas_${model_mode}_${run_mode}_bs${bs_item}_${fp_item}_1gpus 2>&1    #  (5min)
 	    sleep 10
             echo "index is speed, 8gpus, run_mode is multi_process, begin, ${model_name}"
             run_mode=mp
-            CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 bash benchmark/run_benchmark.sh ${run_mode} ${bs_item} ${fp_item} 10 ${model_mode} 
+            CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 bash benchmark/run_benchmark.sh ${run_mode} ${bs_item} ${fp_item} 1 ${model_mode}| tee ${log_path}/clas_${model_mode}_${run_mode}_bs${bs_item}_${fp_item}_8gpus8p 2>&1
             sleep 10
             done
       done

benchmark/run_benchmark.sh

@@ -6,10 +6,19 @@ function _set_params(){
     run_mode=${1:-"sp"}          # 单卡sp|多卡mp
     batch_size=${2:-"64"}
     fp_item=${3:-"fp32"}        # fp32|fp16
-    epochs=${4:-"10"}       # 可选，如果需要修改代码提前中断
+    epochs=${4:-"2"}       # 可选，如果需要修改代码提前中断
     model_name=${5:-"model_name"}
     run_log_path="${TRAIN_LOG_DIR:-$(pwd)}/benchmark"  # TRAIN_LOG_DIR 后续QA设置该参数
- 
+
+    index=1
+    mission_name="图像分类"           # 模型所属任务名称，具体可参考scripts/config.ini                                （必填）
+    direction_id=0                    # 任务所属方向，0：CV，1：NLP，2：Rec。                                          (必填)
+    skip_steps=8                      # 解析日志，有些模型前几个step耗时长，需要跳过                                  (必填)
+    keyword="ips:"              # 解析日志，筛选出数据所在行的关键字                                            (必填)
+    keyword_loss="loss:"       #选填
+    model_mode=-1                      # 解析日志，具体参考scripts/analysis.py.                                        (必填)
+    ips_unit="images/s"
+    base_batch_size=$batch_size
 #   以下不用修改   
     device=${CUDA_VISIBLE_DEVICES//,/ }
     arr=(${device})
@@ -51,6 +60,8 @@ function _train(){
         cp mylog/workerlog.0 ${log_file}
     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
+_run
+#_train