[add]上传训练benchmark by z00560161

2020-10-19 20:22:23 +08:00
parent 22b83024f5
commit 82522e2f61
1225 changed files with 345421 additions and 0 deletions
@@ -0,0 +1,117 @@
+#!/bin/bash
+
+parentDir=$(dirname "$PWD")
+currentDir=$(cd "$(dirname "$0")"; pwd)
+
+execModel=AlexNet
+mode=host
+framework=tensorflow
+hardware=1p
+
+framework_group="tensorflow pytorch mindspore"
+hardware_group="1p 2p 4p 8p cluster ct"
+
+echo_help(){
+    echo ""
+    echo "    --execmodel, -e  (选填) 需要执行的模型名称, 默认 ResNet50"
+    echo "    --hardware, -hw  (选填) 选择 1p, 2p, 4p, 8p, cluster|ct(集群), 默认 1p"
+    echo "    --yamlpath, -y   (选填) yaml 文件的路径, 默认为 yaml 路径下的 {execmodel}.yaml"
+    echo "    --framework, -f  (选填) 模型训练框架, 默认 tensorflow"
+    echo "    -docker, -host   (选填) 选择 docker 或 host 下执行, 默认使用 host"
+    echo ""
+    echo "    --help, -h       显示帮助信息"
+    echo "    --list, -l       显示支持的框架与模型"
+    echo ""
+    echo "    示例1，docker 环境下启动 MobileNet 多卡（8p）训练："
+    echo "        ./benckmark.sh -e MobileNet -hw 8p -y ./yaml/MobileNet.yaml -docker"
+    echo "    示例2，host 环境下启动 MobileNet 单卡（1p）训练，yaml 使用默认文件："
+    echo "        ./benckmark.sh -e MobileNet"
+    echo ""
+    exit 0
+}
+
+error_log(){
+    echo -e "\nERROR:\n\n$1"
+    exit 1
+}
+
+exec_train(){
+
+    if [ x"${yamlPath}" == x"" ];then
+        yamlPath=$(find ${currentDir}/yaml/ -iregex ".*${execModel}.yaml$" 2>/dev/null)
+    fi
+
+    [ -f ${yamlPath} ] || error_log "No such file or directory: ${yamlPath}\n"
+
+    error_msg=""
+    echo $hardware_group | grep -wq "$hardware" || error_msg+="hardware: $hardware not in '$hardware_group'\n"
+    echo $framework_group | grep -wq "$framework" || error_msg+="framework: $framework not in '$framework_group'\n"
+    if [ x"$error_msg" != x"" ];then
+        error_log "$error_msg"
+    fi
+    chmod -R u+x ${currentDir}/*
+    exec_train_file=$(find ${currentDir} -iregex ".*${execModel}/${framework}/scripts/run.sh$" 2>/dev/null)
+    file_count=$(echo ${exec_train_file} | wc -w)
+    start_file=$currentDir/atlas_benchmark-master/utils/shell/start.sh
+    [ x"$hardware" == x"ct" ] && hardware="cluster"
+    if [ "${file_count}" -eq 1 ] && [ -a ${exec_train_file} ]; then
+        modelDir=$(cd $(dirname "$exec_train_file")/..;pwd)
+        echo "find script path success"
+        echo "run train script file path is "${exec_train_file}
+        bash "${start_file}" ${mode} ${hardware} "${yamlPath}" "${modelDir}" "${framework}"
+    else
+        error_log "The model($execModel) does not support the framework($framework) temporarily.\nplease use --list\n"
+        exit 1
+    fi
+}
+
+list_model (){
+    for i in $framework_group;do
+        echo -e "\n${i}:\n"
+        for d in  $(find ${currentDir} -iregex ".*/${i}/scripts/run.sh$" 2>/dev/null);do
+            echo $d|awk -F / '{print "    "$(NF-3)}'
+        done
+    done
+    echo ""
+}
+
+while [ -n "$1" ]
+do
+  case "$1" in
+    -e|--execmodel)
+        execModel=$2;
+        shift
+        ;;
+    -host|-docker)
+        tmp=$1
+        mode=${tmp:1}
+        ;;
+    -y|--yamlpath)
+        yamlPath=$2
+        shift
+        ;;
+    -f|--framework)
+        framework=$2
+        shift
+        ;;
+    -hw|--hardware)
+        hardware=$2
+        shift
+        ;;
+    -l|--list)
+        list_model;
+        exit
+        ;;
+    -h|--help)
+        echo_help;
+        exit
+        ;;
+    *)
+        echo "$1 is not an option, please use --help"
+        exit 1
+        ;;
+  esac
+  shift
+done
+
+exec_train