From f5190c4d26c1afff2929536d4f8aee63fe506126 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Spencer=E5=85=94=E5=AD=90?=
 <11326804+spencer-rabbit@user.noreply.gitee.com>
Date: Fri, 30 Dec 2022 06:22:28 +0000
Subject: [PATCH 1/4] =?UTF-8?q?add=20DeepCTR=5FSeries=5Ffor=5FTensorFlow/t?=
 =?UTF-8?q?est/train=5FRT2=5FID4032=5FDCNMix=5Ffull=5F1p.sh.=20=E5=A2=9E?=
 =?UTF-8?q?=E5=8A=A0train=5FRT2=5FID4032=5FDCNMix=5Ffull=5F1p.sh?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Signed-off-by: Spencer兔子 <11326804+spencer-rabbit@user.noreply.gitee.com>
---
 .../test/train_RT2_ID4032_DCNMix_full_1p.sh   | 152 ++++++++++++++++++
 1 file changed, 152 insertions(+)
 create mode 100644 TensorFlow/built-in/recommendation/DeepCTR_Series_for_TensorFlow/test/train_RT2_ID4032_DCNMix_full_1p.sh

diff --git a/TensorFlow/built-in/recommendation/DeepCTR_Series_for_TensorFlow/test/train_RT2_ID4032_DCNMix_full_1p.sh b/TensorFlow/built-in/recommendation/DeepCTR_Series_for_TensorFlow/test/train_RT2_ID4032_DCNMix_full_1p.sh
new file mode 100644
index 000000000..e108bb58f
--- /dev/null
+++ b/TensorFlow/built-in/recommendation/DeepCTR_Series_for_TensorFlow/test/train_RT2_ID4032_DCNMix_full_1p.sh
@@ -0,0 +1,152 @@
+#!/bin/bash
+
+#当前路径,不需要修改
+cur_path=`pwd`
+#集合通信参数,不需要修改
+export RANK_SIZE=1
+export JOB_ID=10087
+RANK_ID_START=0
+# 数据集路径,保持为空,不需要修改
+data_path=""
+#基础参数，需要模型审视修改
+#网络名称，同目录名称
+Network="DCNMix_ID4032_for_TensorFlow"
+#训练epoch
+train_epochs=4
+#训练batch_size
+batch_size=10240
+#训练step
+train_steps=16115
+#学习率
+learning_rate=
+
+#维测参数，precision_mode需要模型审视修改
+precision_mode="allow_fp32_to_fp16"
+#维持参数，以下不需要修改
+over_dump=False
+data_dump_flag=False
+data_dump_step="10"
+profiling=False
+
+# 帮助信息，不需要修改
+if [[ $1 == --help || $1 == -h ]];then
+    echo"usage:./train_performance_1P.sh <args>"
+    echo " "
+    echo "parameter explain:
+    --precision_mode         precision mode(allow_fp32_to_fp16/force_fp16/must_keep_origin_dtype/allow_mix_precision)
+    --over_dump		           if or not over detection, default is False
+    --data_dump_flag		     data dump flag, default is False
+    --data_dump_step		     data dump step, default is 10
+    --profiling		           if or not profiling for performance debug, default is False
+	--data_path		           source data of training
+    -h/--help		             show help message
+    "
+    exit 1
+fi
+
+#参数校验，不需要修改
+for para in $*
+do
+    if [[ $para == --precision_mode* ]];then
+        precision_mode=`echo ${para#*=}`
+    elif [[ $para == --over_dump* ]];then
+        over_dump=`echo ${para#*=}`
+        over_dump_path=${cur_path}/output/overflow_dump
+        mkdir -p ${over_dump_path}
+    elif [[ $para == --data_dump_flag* ]];then
+        data_dump_flag=`echo ${para#*=}`
+        data_dump_path=${cur_path}/output/data_dump
+        mkdir -p ${data_dump_path}
+    elif [[ $para == --data_dump_step* ]];then
+        data_dump_step=`echo ${para#*=}`
+    elif [[ $para == --profiling* ]];then
+        profiling=`echo ${para#*=}`
+        profiling_dump_path=${cur_path}/output/profiling
+        mkdir -p ${profiling_dump_path}
+    elif [[ $para == --data_path* ]];then
+        data_path=`echo ${para#*=}`
+    fi
+done
+
+#校验是否传入data_path,不需要修改
+if [[ $data_path == "" ]];then
+    echo "[Error] para \"data_path\" must be confing"
+    exit 1
+fi
+
+#训练开始时间，不需要修改
+start_time=$(date +%s)
+
+#进入训练脚本目录，需要模型审视修改
+cd $cur_path/../examples
+
+for((RANK_ID=$RANK_ID_START;RANK_ID<$((RANK_SIZE+RANK_ID_START));RANK_ID++));
+do
+    #设置环境变量，不需要修改
+    echo "Device ID: $ASCEND_DEVICE_ID"
+    export RANK_ID=$RANK_ID
+
+    #创建DeviceID输出目录，不需要修改
+    if [ -d ${cur_path}/output/${ASCEND_DEVICE_ID} ];then
+        rm -rf ${cur_path}/output/${ASCEND_DEVICE_ID}
+        mkdir -p ${cur_path}/output/$ASCEND_DEVICE_ID/ckpt
+    else
+        mkdir -p ${cur_path}/output/$ASCEND_DEVICE_ID/ckpt
+    fi
+    
+    #执行训练脚本，以下传参不需要修改，其他需要模型审视修改
+    #--data_dir, --model_dir, --precision_mode, --over_dump, --over_dump_path，--data_dump_flag，--data_dump_step，--data_dump_path，--profiling，--profiling_dump_path
+    nohup python3 run_dcnmix.py \
+        --data_path=${data_path} \
+        --train_batch_size=${batch_size} \
+        --eval_batch_size=${batch_size} \
+        --num_epochs=${train_epochs} \
+        --max_steps=${train_steps} \
+        --output_dir=${cur_path}/output/$ASCEND_DEVICE_ID/ckpt > ${cur_path}/output/${ASCEND_DEVICE_ID}/train_${ASCEND_DEVICE_ID}.log 2>&1 &
+done 
+wait
+
+#训练结束时间，不需要修改
+end_time=$(date +%s)
+e2e_time=$(( $end_time - $start_time ))
+
+
+#结果打印，不需要修改
+echo "------------------ Final result ------------------"
+# #输出性能FPS，需要模型审视修改
+fps=`grep "examples\/sec" $cur_path/output/${ASCEND_DEVICE_ID}/train_${ASCEND_DEVICE_ID}.log | awk '{print $2}'|tail -n +5 | awk 'NR>1{print p}{p=$0}'|awk '{sum+=$1} END {print sum/NR}'`
+FPS=`awk 'BEGIN{printf "%.2f\n", '${fps}'}'`
+# #打印，不需要修改
+echo "Final Performance item/sec : $FPS"
+
+# #输出训练精度,需要模型审视修改
+train_accuracy=`grep "AUC = " $cur_path/output/${ASCEND_DEVICE_ID}/train_${ASCEND_DEVICE_ID}.log | awk 'END {print $9}' | sed 's/,//g'`
+echo "Final Train Accuracy : ${train_accuracy}"
+echo "E2E Training Duration sec : $e2e_time"
+
+#性能看护结果汇总
+#训练用例信息，不需要修改
+BatchSize=${batch_size}
+DeviceType=`uname -m`
+CaseName=${Network}_bs${BatchSize}_${RANK_SIZE}'p'_'RT2'_'accu'
+
+##获取性能数据，不需要修改
+#吞吐量
+TrainingTime=`awk 'BEGIN{printf "%.6f\n",'${BatchSize}'/'${FPS}'}'`
+
+ActualFPS=${FPS}
+grep ":loss =" $cur_path/output/$ASCEND_DEVICE_ID/train_$ASCEND_DEVICE_ID.log| awk '{print $3}' | sed 's/,//g'| sed -n '1~2p' > $cur_path/output/$ASCEND_DEVICE_ID/train_${CaseName}_loss.txt
+#最后一个迭代loss值，不需要修改
+ActualLoss=`awk 'END {print}' $cur_path/output/$ASCEND_DEVICE_ID/train_${CaseName}_loss.txt`
+
+#关键信息打印到${CaseName}.log中，不需要修改
+echo "Network = ${Network}" > $cur_path/output/$ASCEND_DEVICE_ID/${CaseName}.log
+echo "RankSize = ${RANK_SIZE}" >> $cur_path/output/$ASCEND_DEVICE_ID/${CaseName}.log
+echo "BatchSize = ${BatchSize}" >> $cur_path/output/$ASCEND_DEVICE_ID/${CaseName}.log
+echo "DeviceType = ${DeviceType}" >> $cur_path/output/$ASCEND_DEVICE_ID/${CaseName}.log
+echo "CaseName = ${CaseName}" >> $cur_path/output/$ASCEND_DEVICE_ID/${CaseName}.log
+echo "ActualFPS = ${ActualFPS}" >> $cur_path/output/$ASCEND_DEVICE_ID/${CaseName}.log
+echo "TrainAccuracy = ${train_accuracy}" >> $cur_path/output/$ASCEND_DEVICE_ID/${CaseName}.log
+echo "TrainingTime = ${TrainingTime}" >> $cur_path/output/$ASCEND_DEVICE_ID/${CaseName}.log
+echo "ActualLoss = ${ActualLoss}" >> $cur_path/output/$ASCEND_DEVICE_ID/${CaseName}.log
+echo "E2ETrainingTime = ${e2e_time}" >> $cur_path/output/$ASCEND_DEVICE_ID/${CaseName}.log
\ No newline at end of file
-- 
Gitee


From b65552e896cb33638f3572eba86a092fcfc1f355 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Spencer=E5=85=94=E5=AD=90?=
 <11326804+spencer-rabbit@user.noreply.gitee.com>
Date: Fri, 30 Dec 2022 06:57:31 +0000
Subject: [PATCH 2/4] =?UTF-8?q?add=20Transformer=5FID0633=5Ffor=5FTensorFl?=
 =?UTF-8?q?ow2.X/test/train=5FRT2=5Ffull=5F1p=5F6144bs=5Fdynamic=5Fnoeval.?=
 =?UTF-8?q?sh.=20=E5=A2=9E=E5=8A=A0train=5FRT2=5Ffull=5F1p=5F6144bs=5Fdyna?=
 =?UTF-8?q?mic=5Fnoeval.sh?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Signed-off-by: Spencer兔子 <11326804+spencer-rabbit@user.noreply.gitee.com>
---
 ...train_RT2_full_1p_6144bs_dynamic_noeval.sh | 196 ++++++++++++++++++
 1 file changed, 196 insertions(+)
 create mode 100644 TensorFlow2/built-in/nlp/Transformer_ID0633_for_TensorFlow2.X/test/train_RT2_full_1p_6144bs_dynamic_noeval.sh

diff --git a/TensorFlow2/built-in/nlp/Transformer_ID0633_for_TensorFlow2.X/test/train_RT2_full_1p_6144bs_dynamic_noeval.sh b/TensorFlow2/built-in/nlp/Transformer_ID0633_for_TensorFlow2.X/test/train_RT2_full_1p_6144bs_dynamic_noeval.sh
new file mode 100644
index 000000000..4d556a4ee
--- /dev/null
+++ b/TensorFlow2/built-in/nlp/Transformer_ID0633_for_TensorFlow2.X/test/train_RT2_full_1p_6144bs_dynamic_noeval.sh
@@ -0,0 +1,196 @@
+#!/bin/bash
+
+#当前路径,不需要修改
+cur_path=`pwd`
+
+#集合通信参数,不需要修改
+
+export RANK_SIZE=1
+export JOB_ID=10087
+export RANK_ID_START=0
+export PYTHONPATH=../transformer:$PYTHONPATH
+
+
+# 数据集路径,保持为空,不需要修改
+data_path=""
+
+#设置默认日志级别,不需要修改
+
+#基础参数，需要模型审视修改
+#网络名称，同目录名称
+Network="Transformer_ID3215_for_TensorFlow2.X"
+#训练batch_size
+batch_size=6144
+#训练step
+train_steps=250000
+
+#TF2.X独有，不需要修改
+#export NPU_ENABLE_PERF=true
+
+#维测参数，precision_mode需要模型审视修改
+precision_mode="allow_mix_precision"
+#维持参数，以下不需要修改
+over_dump=False
+data_dump_flag=False
+data_dump_step="10"
+profiling=False
+
+
+# 帮助信息，不需要修改
+if [[ $1 == --help || $1 == -h ]];then
+    echo"usage:./train_full_1p.sh <args>"
+    echo " "
+    echo "parameter explain:
+    --precision_mode         precision mode(allow_fp32_to_fp16/force_fp16/must_keep_origin_dtype/allow_mix_precision)
+    --over_dump		         if or not over detection, default is False
+    --data_dump_flag	     data dump flag, default is False
+    --data_dump_step		 data dump step, default is 10
+    --profiling		         if or not profiling for performance debug, default is False
+    --data_path		         source data of training
+    -h/--help		         show help message
+    "
+    exit 1
+fi
+
+#参数校验，不需要修改
+for para in $*
+do
+    if [[ $para == --precision_mode* ]];then
+        precision_mode=`echo ${para#*=}`
+    elif [[ $para == --over_dump* ]];then
+        over_dump=`echo ${para#*=}`
+        over_dump_path=${cur_path}/output/overflow_dump
+        mkdir -p ${over_dump_path}
+    elif [[ $para == --data_dump_flag* ]];then
+        data_dump_flag=`echo ${para#*=}`
+        data_dump_path=${cur_path}/output/data_dump
+        mkdir -p ${data_dump_path}
+    elif [[ $para == --data_dump_step* ]];then
+        data_dump_step=`echo ${para#*=}`
+    elif [[ $para == --profiling* ]];then
+        profiling=`echo ${para#*=}`
+        profiling_dump_path=${cur_path}/output/profiling
+        mkdir -p ${profiling_dump_path}
+    elif [[ $para == --data_path* ]];then
+        data_path=`echo ${para#*=}`
+    fi
+done
+
+#校验是否传入data_path,不需要修改
+if [[ $data_path == "" ]];then
+    echo "[Error] para \"data_path\" must be confing"
+    exit 1
+fi
+
+#训练开始时间，不需要修改
+start_time=$(date +%s)
+
+#进入训练脚本目录，需要模型审视修改
+
+for((RANK_ID=$RANK_ID_START;RANK_ID<$((RANK_SIZE+RANK_ID_START));RANK_ID++));
+do
+    #设置环境变量，不需要修改
+    echo "Device ID: $ASCEND_DEVICE_ID"
+    export RANK_ID=$RANK_ID
+    
+    
+    
+    #创建DeviceID输出目录，不需要修改
+    if [ -d ${cur_path}/output/${ASCEND_DEVICE_ID} ];then
+        rm -rf ${cur_path}/output/${ASCEND_DEVICE_ID}
+        mkdir -p ${cur_path}/output/$ASCEND_DEVICE_ID/ckpt_${learning_rate}
+    else
+        mkdir -p ${cur_path}/output/$ASCEND_DEVICE_ID/ckpt_${learning_rate}
+    fi
+    
+    #绑核，不需要绑核的模型删除，需要绑核的模型根据实际修改
+    cpucount=`lscpu | grep "CPU(s):" | head -n 1 | awk '{print $2}'`
+    cpustep=`expr $cpucount / 8`
+    echo "taskset c steps:" $cpustep
+    let a=RANK_ID*$cpustep
+    let b=RANK_ID+1
+    let c=b*$cpustep-1
+    
+    #执行训练脚本，以下传参不需要修改，其他需要模型审视修改
+    #--data_dir, --model_dir, --precision_mode, --over_dump, --over_dump_path，--data_dump_flag，--data_dump_step，--data_dump_path，--profiling，--profiling_dump_path，--autotune
+    nohup taskset -c $a-$c python3 ../transformer/official/nlp/transformer/transformer_main.py \
+  	--data_dir=${data_path} \
+ 	--model_dir=${cur_path}/output/$ASCEND_DEVICE_ID/ckpt \
+	--vocab_file=${data_path}/vocab.ende.32768 \
+	--param_set=big \
+	--train_steps=${train_steps} \
+	--batch_size=${batch_size} \
+	--steps_between_evals=10000 \
+	--max_length=64 \
+	--mode=train \
+	--decode_batch_size=32 \
+	--decode_max_length=97 \
+	--padded_decode=False \
+	--num_gpus=1 \
+	--dtype=fp16 \
+	--distribution_strategy='one_device' \
+	--enable_time_history=true \
+  --log_steps=1000 \
+  --loss_scale='dynamic' \
+	--precision_mode=${precision_mode} \
+        --over_dump=${over_dump} \
+        --over_dump_path=${over_dump_path} \
+        --data_dump_flag=${data_dump_flag} \
+        --data_dump_step=${data_dump_step} \
+        --data_dump_path=${data_dump_path} \
+	--profiling=${profiling} \
+	--profiling_dump_path=${profiling_dump_path} > ${cur_path}/output/${ASCEND_DEVICE_ID}/train_${ASCEND_DEVICE_ID}.log 2>&1 &
+done 
+wait
+
+#训练结束时间，不需要修改
+end_time=$(date +%s)
+e2e_time=$(( $end_time - $start_time ))
+
+#############冒烟看护#########################
+BatchSize=${batch_size}
+#设备类型
+DeviceType=`uname -m`
+#用例名称
+CaseName=${Network}_bs${BatchSize}_${RANK_SIZE}'p'_'RT2'_'acc'
+
+echo "------------------ Final result ------------------"
+#输出性能FPS，需要模型审视修改
+single_batch_step_sec=`grep TimeHistory  $cur_path/output/${ASCEND_DEVICE_ID}/train_${ASCEND_DEVICE_ID}.log|awk 'END {print $8}'`
+FPS=`awk 'BEGIN{printf "%.2f\n",'${single_batch_step_sec}'}'`
+#打印，不需要修改
+echo "Final Performance images/sec : $FPS"
+
+#grep "Train history" $cur_path/output/$ASCEND_DEVICE_ID/train_$ASCEND_DEVICE_ID.log|awk '{print$8}'|sed 's/,//g'|sed 's/\[//g'|sed 's/\]//g' |sed 's/\}//g'>> $cur_path/output/$ASCEND_DEVICE_ID/train_${CaseName}_loss.txt
+ActualLoss=`grep 10000/10000  $cur_path/output/${ASCEND_DEVICE_ID}/train_${ASCEND_DEVICE_ID}.log|awk 'END {print $NF}'`
+#输出训练精度,需要模型审视修改
+#train_accuracy=`grep "Bleu" $cur_path/output/$ASCEND_DEVICE_ID/train_$ASCEND_DEVICE_ID.log | awk -F ": " '{print $2}' | tail -n 1`
+train_accuracy=`grep 10000/10000  $cur_path/output/${ASCEND_DEVICE_ID}/train_${ASCEND_DEVICE_ID}.log|awk 'END {print $NF}'`
+#打印，不需要修改
+echo "Final Train Accuracy : ${train_accuracy}"
+echo "E2E Training Duration sec : $e2e_time"
+
+##获取性能数据
+#吞吐量，不需要修改
+ActualFPS=${FPS}
+#单迭代训练时长，不需要修改
+TrainingTime=`awk 'BEGIN{printf "%.2f\n",'${BatchSize}'*'${RANK_SIZE}'*1000/'${FPS}'}'`
+
+##获取Loss
+#从train_$ASCEND_DEVICE_ID.log提取Loss到train_${CaseName}_loss.txt中
+grep 10000/10000 $cur_path/output/$ASCEND_DEVICE_ID/train_$ASCEND_DEVICE_ID.log|awk '{print $NF}' > $cur_path/output/$ASCEND_DEVICE_ID/train_${CaseName}_loss.txt
+
+#最后一个迭代loss值
+#ActualLoss=`awk 'END {print}' $cur_path/output/$ASCEND_DEVICE_ID/train_${CaseName}_loss.txt`
+
+#关键信息打印到${CaseName}.log中
+echo "Network = ${Network}" > $cur_path/output/$ASCEND_DEVICE_ID/${CaseName}.log
+echo "RankSize = ${RANK_SIZE}" >> $cur_path/output/$ASCEND_DEVICE_ID/${CaseName}.log
+echo "BatchSize = ${BatchSize}" >> $cur_path/output/$ASCEND_DEVICE_ID/${CaseName}.log
+echo "DeviceType = ${DeviceType}" >> $cur_path/output/$ASCEND_DEVICE_ID/${CaseName}.log
+echo "CaseName = ${CaseName}" >> $cur_path/output/$ASCEND_DEVICE_ID/${CaseName}.log
+echo "ActualFPS = ${ActualFPS}" >> $cur_path/output/$ASCEND_DEVICE_ID/${CaseName}.log
+echo "TrainingTime = ${TrainingTime}" >> $cur_path/output/$ASCEND_DEVICE_ID/${CaseName}.log
+echo "TrainAccuracy = ${train_accuracy}" >> $cur_path/output/$ASCEND_DEVICE_ID/${CaseName}.log
+echo "ActualLoss = ${ActualLoss}" >> $cur_path/output/$ASCEND_DEVICE_ID/${CaseName}.log
+echo "E2ETrainingTime = ${e2e_time}" >> $cur_path/output/$ASCEND_DEVICE_ID/${CaseName}.log
\ No newline at end of file
-- 
Gitee


From 924b43d942e912d40a9f6db80c3373d3a3ff8b43 Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Spencer=E5=85=94=E5=AD=90?=
 <11326804+spencer-rabbit@user.noreply.gitee.com>
Date: Fri, 30 Dec 2022 07:11:03 +0000
Subject: [PATCH 3/4] =?UTF-8?q?add=20TensorFlow/built-in/recommendation/Wi?=
 =?UTF-8?q?deDeep=5FID2940=5Ffor=5FTensorFlow/test/train=5FRT2=5Ffull=5F1p?=
 =?UTF-8?q?.sh.=20=E5=A2=9E=E5=8A=A0train=5FRT2=5Ffull=5F1p.sh?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Signed-off-by: Spencer兔子 <11326804+spencer-rabbit@user.noreply.gitee.com>
---
 .../test/train_RT2_full_1p.sh                 | 170 ++++++++++++++++++
 1 file changed, 170 insertions(+)
 create mode 100644 TensorFlow/built-in/recommendation/WideDeep_ID2940_for_TensorFlow/test/train_RT2_full_1p.sh

diff --git a/TensorFlow/built-in/recommendation/WideDeep_ID2940_for_TensorFlow/test/train_RT2_full_1p.sh b/TensorFlow/built-in/recommendation/WideDeep_ID2940_for_TensorFlow/test/train_RT2_full_1p.sh
new file mode 100644
index 000000000..b5e4709e8
--- /dev/null
+++ b/TensorFlow/built-in/recommendation/WideDeep_ID2940_for_TensorFlow/test/train_RT2_full_1p.sh
@@ -0,0 +1,170 @@
+#!/bin/bash
+
+#当前路径,不需要修改
+cur_path=`pwd`
+#export ASCEND_SLOG_PRINT_TO_STDOUT=1
+
+#集合通信参数,不需要修改
+
+export RANK_SIZE=1
+export JOB_ID=10087
+RANK_ID_START=0
+
+# 数据集路径,保持为空,不需要修改
+data_path=""
+
+
+#基础参数，需要模型审视修改
+#Batch Size
+batch_size=131072
+#网络名称，同目录名称
+Network="WideDeep_TF_ID2940_for_TensorFlow"
+#Device数量，单卡默认为1
+RankSize=1
+#训练epoch，可选
+train_epochs=120
+
+#参数配置
+data_path=""
+
+#维测参数，precision_mode需要模型审视修改
+precision_mode="allow_mix_precision"
+#维持参数，以下不需要修改
+over_dump=False
+data_dump_flag=False
+data_dump_step="10"
+profiling=False
+autotune=False
+
+# 帮助信息，不需要修改
+if [[ $1 == --help || $1 == -h ]];then
+    echo"usage:./train_performance_1p.sh <args>"
+    echo " "
+    echo "parameter explain:
+    --precision_mode         precision mode(allow_fp32_to_fp16/force_fp16/must_keep_origin_dtype/allow_mix_precision)
+    --over_dump		         if or not over detection, default is False
+    --data_dump_flag	     data dump flag, default is False
+    --data_dump_step		 data dump step, default is 10
+    --profiling		         if or not profiling for performance debug, default is False
+    --data_path		         source data of training
+    -h/--help		         show help message
+    "
+    exit 1
+fi
+
+#参数校验，不需要修改
+for para in $*
+do
+    if [[ $para == --precision_mode* ]];then
+        precision_mode=`echo ${para#*=}`
+elif [[ $para == --over_dump* ]];then
+        over_dump=`echo ${para#*=}`
+        over_dump_path=${cur_path}/output/overflow_dump
+        mkdir -p ${over_dump_path}
+    elif [[ $para == --data_dump_flag* ]];then
+        data_dump_flag=`echo ${para#*=}`
+        data_dump_path=${cur_path}/output/data_dump
+        mkdir -p ${data_dump_path}
+    elif [[ $para == --data_dump_step* ]];then
+        data_dump_step=`echo ${para#*=}`
+    elif [[ $para == --profiling* ]];then
+        profiling=`echo ${para#*=}`
+        profiling_dump_path=${cur_path}/output/profiling
+        mkdir -p ${profiling_dump_path}
+    elif [[ $para == --data_path* ]];then
+        data_path=`echo ${para#*=}`
+    fi
+done
+
+#校验是否传入data_path,不需要修改
+if [[ $data_path == "" ]];then
+    echo "[Error] para \"data_path\" must be confing"
+    exit 1
+fi
+
+##############执行训练##########
+if [ -d $cur_path/output ];then
+   rm -rf $cur_path/output/*
+   mkdir -p $cur_path/output/$ASCEND_DEVICE_ID
+else
+   mkdir -p $cur_path/output/$ASCEND_DEVICE_ID/ckpt
+fi
+
+wait
+
+if [ -d $cur_path/output ];then
+   rm -rf $cur_path/output/*
+   mkdir -p $cur_path/output/$ASCEND_DEVICE_ID
+else
+   mkdir -p $cur_path/output/$ASCEND_DEVICE_ID/ckpt
+fi
+
+if [ -d $cur_path/../config/1p_$ASCEND_DEVICE.json ];then
+    export RANK_TABLE_FILE=$cur_path/../config/1p_$ASCEND_DEVICE.json
+    export RANK_ID=$ASCEND_DEVICE_ID
+else
+    export RANK_TABLE_FILE=$cur_path/../config/1p_0.json
+    export RANK_ID=0
+fi
+
+cd $cur_path/../
+start=$(date +%s)
+nohup python3  -m trainer.task --gpu \
+ --model_type=wide \
+ --train_data_pattern=$data_path/outbrain/tfrecords/train/part* \
+ --eval_data_pattern=$data_path/outbrain/tfrecords/eval/part* \
+ --model_dir=$cur_path/output/$ASCEND_DEVICE_ID/ckpt \
+ --transformed_metadata_path=$data_path/outbrain/tfrecords \
+ --num_epochs=$train_epochs > $cur_path/output/$ASCEND_DEVICE_ID/train_$ASCEND_DEVICE_ID.log 2>&1 &
+wait
+end=$(date +%s)
+e2etime=$(( $end - $start ))
+
+
+#结果打印，不需要修改
+echo "------------------ Final result ------------------"
+#输出性能FPS，需要模型审视修改
+
+Time=`grep "INFO:tensorflow:global_step/sec: "  $cur_path/output/${ASCEND_DEVICE_ID}/train_${ASCEND_DEVICE_ID}.log| awk -F' ' '{print $2}' | tail -n 2 | head -n +1`
+FPS=`awk 'BEGIN{printf "%.2f\n", '${batch_size}'*'${Time}'}'`
+
+
+#打印，不需要修改
+echo "Final Performance images/sec : $FPS"
+
+#输出训练精度,需要模型审视修改
+train_accuracy=`grep 'INFO:tensorflow:Saving dict for global step' $cur_path/output/${ASCEND_DEVICE_ID}/train_${ASCEND_DEVICE_ID}.log| tail -n 1 |awk -F' = ' '{print $3}' |awk -F',' '{print $1}'`
+#打印，不需要修改
+echo "Final Train Accuracy : ${train_accuracy}"
+echo "E2E Training Duration sec : $e2e_time"
+
+#稳定性精度看护结果汇总
+#训练用例信息，不需要修改
+BatchSize=${batch_size}
+DeviceType=`uname -m`
+CaseName=${Network}_bs${BatchSize}_${RANK_SIZE}'p'_'RT2'_'acc'
+
+##获取性能数据
+#吞吐量，不需要修改
+ActualFPS=${FPS}
+#单迭代训练时长，不需要修改
+TrainingTime=`awk -v x=320 -v y="$FPS" 'BECIN{printf "%3.f\n",y/x}'`
+
+#从train_$ASCEND_DEVICE_ID.log提取Loss到train_${CaseName}_loss.txt中，需要根据模型审视
+loss=`grep 'INFO:tensorflow:loss' $cur_path/output/${ASCEND_DEVICE_ID}/train_${ASCEND_DEVICE_ID}.log | tr -d '\b\r' | grep -Eo "INFO:tensorflow:loss = [0-9]*\.[0-9]*" | awk -F' = ' '{print $2}'`
+echo "${loss}"> $cur_path/output/$ASCEND_DEVICE_ID/train_${CaseName}_loss.txt
+
+#最后一个迭代loss值，不需要修改
+ActualLoss=`cat $cur_path/output/$ASCEND_DEVICE_ID/train_${CaseName}_loss.txt | tail -n 1`
+
+#关键信息打印到${CaseName}.log中，不需要修改
+echo "Network = ${Network}" > $cur_path/output/$ASCEND_DEVICE_ID/${CaseName}.log
+echo "RankSize = ${RANK_SIZE}" >> $cur_path/output/$ASCEND_DEVICE_ID/${CaseName}.log
+echo "BatchSize = ${BatchSize}" >> $cur_path/output/$ASCEND_DEVICE_ID/${CaseName}.log
+echo "DeviceType = ${DeviceType}" >> $cur_path/output/$ASCEND_DEVICE_ID/${CaseName}.log
+echo "CaseName = ${CaseName}" >> $cur_path/output/$ASCEND_DEVICE_ID/${CaseName}.log
+echo "ActualFPS = ${ActualFPS}" >> $cur_path/output/$ASCEND_DEVICE_ID/${CaseName}.log
+echo "TrainingTime = ${TrainingTime}" >> $cur_path/output/$ASCEND_DEVICE_ID/${CaseName}.log
+echo "TrainAccuracy = ${train_accuracy}" >> $cur_path/output/$ASCEND_DEVICE_ID/${CaseName}.log
+echo "ActualLoss = ${ActualLoss}" >> $cur_path/output/$ASCEND_DEVICE_ID/${CaseName}.log
+echo "E2ETrainingTime = ${e2e_time}" >> $cur_path/output/$ASCEND_DEVICE_ID/${CaseName}.log
\ No newline at end of file
-- 
Gitee


From 1813a77b1f563a7b4bd53c8e5a8931c1e638cb2e Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Spencer=E5=85=94=E5=AD=90?=
 <11326804+spencer-rabbit@user.noreply.gitee.com>
Date: Fri, 30 Dec 2022 07:21:27 +0000
Subject: [PATCH 4/4] =?UTF-8?q?add=20Swin-Transformer=5FID2374=5Ffor=5FTen?=
 =?UTF-8?q?sorFlow2.X/test/train=5FRT2=5Ffull=5F1p.sh.=20=E5=A2=9E?=
 =?UTF-8?q?=E5=8A=A0train=5FRT2=5Ffull=5F1p.sh?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

Signed-off-by: Spencer兔子 <11326804+spencer-rabbit@user.noreply.gitee.com>
---
 .../test/train_RT2_full_1p.sh                 | 152 ++++++++++++++++++
 1 file changed, 152 insertions(+)
 create mode 100644 TensorFlow2/built-in/cv/detection/Swin-Transformer_ID2374_for_TensorFlow2.X/test/train_RT2_full_1p.sh

diff --git a/TensorFlow2/built-in/cv/detection/Swin-Transformer_ID2374_for_TensorFlow2.X/test/train_RT2_full_1p.sh b/TensorFlow2/built-in/cv/detection/Swin-Transformer_ID2374_for_TensorFlow2.X/test/train_RT2_full_1p.sh
new file mode 100644
index 000000000..366ab7f63
--- /dev/null
+++ b/TensorFlow2/built-in/cv/detection/Swin-Transformer_ID2374_for_TensorFlow2.X/test/train_RT2_full_1p.sh
@@ -0,0 +1,152 @@
+#!/bin/bash
+
+cur_path=`pwd`/../
+#失败用例打屏
+#export ASCEND_SLOG_PRINT_TO_STDOUT=1
+
+#基础参数，需要模型审视修改
+#Batch Size
+batch_size=32
+#网络名称，同目录名称
+Network="Swin-Transformer_ID2374_for_TensorFlow2.X"
+#Device数量，单卡默认为1
+RankSize=1
+#训练epoch，可选
+train_epochs=
+#训练step
+train_steps=
+#学习率
+learning_rate=0.01
+
+#参数配置
+data_path=""
+
+############维测参数##############
+precision_mode="allow_mix_precision"
+#维持参数，以下不需要修改
+over_dump=False
+if [[ $over_dump == True ]];then
+    over_dump_path=$cur_path/test/overflow_dump #此处cur_path为代码根目录
+    mkdir -p ${over_dump_path}
+fi
+data_dump_flag=False
+data_dump_step="10"
+profiling=False
+use_mixlist=False
+mixlist_file="./configs/ops_info.json"
+fusion_off_flag=False
+fusion_off_file="./configs/fusion_switch.cfg"
+############维测参数##############
+
+if [[ $1 == --help || $1 == --h ]];then
+   echo "usage:./train_full_1p.sh "
+   exit 1
+fi
+
+for para in $*
+do
+    if [[ $para == --data_path* ]];then
+        data_path=`echo ${para#*=}`
+    elif [[ $para == --precision_mode* ]];then
+        precision_mode=`echo ${para#*=}`
+    elif [[ $para == --over_dump* ]];then
+        over_dump=`echo ${para#*=}`
+        over_dump_path=${cur_path}/output/overflow_dump
+        mkdir -p ${over_dump_path}
+    elif [[ $para == --data_dump_flag* ]];then
+        data_dump_flag=`echo ${para#*=}`
+        data_dump_path=${cur_path}/output/data_dump
+        mkdir -p ${data_dump_path}
+    elif [[ $para == --data_dump_step* ]];then
+        data_dump_step=`echo ${para#*=}`
+    elif [[ $para == --profiling* ]];then
+        profiling=`echo ${para#*=}`
+        profiling_dump_path=${cur_path}/output/profiling
+        mkdir -p ${profiling_dump_path}
+    elif [[ $para == --data_path* ]];then
+        data_path=`echo ${para#*=}`
+    elif [[ $para == --use_mixlist* ]];then
+        use_mixlist=`echo ${para#*=}`
+    elif [[ $para == --mixlist_file* ]];then
+        mixlist_file=`echo ${para#*=}`
+    elif [[ $para == --fusion_off_flag* ]];then
+        fusion_off_flag=`echo ${para#*=}`
+    elif [[ $para == --fusion_off_file* ]];then
+        fusion_off_file=`echo ${para#*=}`
+    fi
+done
+
+if [[ $data_path  == "" ]];then
+   echo "[Error] para \"data_path\" must be config"
+   exit 1
+fi
+##############执行训练##########
+cd $cur_path
+if [ -d $cur_path/test/output ];then
+   rm -rf $cur_path/test/output/*
+   mkdir -p $cur_path/test/output/$ASCEND_DEVICE_ID
+else
+   mkdir -p $cur_path/test/output/$ASCEND_DEVICE_ID
+fi
+wait
+
+#拷贝并修改数据集txt文件
+cp -r ${data_path}/cifar-100-python /root/.keras/datasets/
+
+start=$(date +%s)
+nohup python3 swin_transformers.py \
+        --precision_mode=${precision_mode} \
+        --over_dump=${over_dump} \
+        --over_dump_path=${over_dump_path} \
+        --data_dump_flag=${data_dump_flag} \
+        --data_dump_step=${data_dump_step} \
+        --data_dump_path=${data_dump_path} \
+        --profiling=${profiling} \
+        --use_mixlist=${use_mixlist} \
+        --fusion_off_flag=${fusion_off_flag} \
+        --mixlist_file=${mixlist_file} \
+        --fusion_off_file=${fusion_off_file} \
+        --profiling_dump_path=${profiling_dump_path}} > $cur_path/test/output/$ASCEND_DEVICE_ID/train_$ASCEND_DEVICE_ID.log 2>&1 &
+wait
+end=$(date +%s)
+e2etime=$(( $end - $start ))
+
+#echo "Final Performance ms/step : $average_perf"
+echo "Final Training Duration sec : $e2etime"
+
+#结果打印，不需要修改
+echo "------------------ Final result ------------------"
+#输出性能FPS，需要模型审视修改
+TrainingTime=`grep "352/352" $cur_path/test/output/$ASCEND_DEVICE_ID/train_$ASCEND_DEVICE_ID.log|awk -F '352/352' '{print $2}'|grep -v 'ETA'|grep 'loss:'|awk 'END {print $4}'|cut -d 'm' -f -1`
+
+
+#性能看护结果汇总
+#训练用例信息，不需要修改
+BatchSize=${batch_size}
+DeviceType=`uname -m`
+CaseName=${Network}_bs${BatchSize}_${RANK_SIZE}'p'_'RT2'_'acc'
+
+##获取性能数据，不需要修改
+#吞吐量
+ActualFPS=`awk 'BEGIN{printf "%.2f\n", '1000'*'${batch_size}'/'${TrainingTime}'}'`
+
+#获取模型精度
+train_accuracy=`grep "352/352" $cur_path/test/output/$ASCEND_DEVICE_ID/train_$ASCEND_DEVICE_ID.log|awk -F '352/352' '{print $2}'|grep 'loss:'|awk 'END {print $10}'`
+
+#从train_$ASCEND_DEVICE_ID.log提取Loss到train_${CaseName}_loss.txt中，需要根据模型审视
+grep "352/352" $cur_path/test/output/$ASCEND_DEVICE_ID/train_$ASCEND_DEVICE_ID.log|awk -F '352/352' '{print $2}'|grep -v 'ETA'|grep 'loss:'|awk '{print $7}' > $cur_path/test/output/$ASCEND_DEVICE_ID/train_${CaseName}_loss.txt
+
+#最后一个迭代loss值，不需要修改
+ActualLoss=`awk 'END {print}' $cur_path/test/output/$ASCEND_DEVICE_ID/train_${CaseName}_loss.txt`
+
+#关键信息打印到${CaseName}.log中，不需要修改
+echo "Network = ${Network}" > $cur_path/test/output/$ASCEND_DEVICE_ID/${CaseName}.log
+echo "RankSize = ${RANK_SIZE}" >> $cur_path/test/output/$ASCEND_DEVICE_ID/${CaseName}.log
+echo "BatchSize = ${BatchSize}" >> $cur_path/test/output/$ASCEND_DEVICE_ID/${CaseName}.log
+echo "DeviceType = ${DeviceType}" >> $cur_path/test/output/$ASCEND_DEVICE_ID/${CaseName}.log
+echo "CaseName = ${CaseName}" >> $cur_path/test/output/$ASCEND_DEVICE_ID/${CaseName}.log
+echo "ActualFPS = ${ActualFPS}" >> $cur_path/test/output/$ASCEND_DEVICE_ID/${CaseName}.log
+echo "TrainingTime = ${TrainingTime}" >> $cur_path/test/output/$ASCEND_DEVICE_ID/${CaseName}.log
+echo "ActualLoss = ${ActualLoss}" >> $cur_path/test/output/$ASCEND_DEVICE_ID/${CaseName}.log
+echo "TrainAccuracy = ${train_accuracy}" >> $cur_path/test/output/$ASCEND_DEVICE_ID/${CaseName}.log
+echo "E2ETrainingTime = ${e2etime}" >> $cur_path/test/output/$ASCEND_DEVICE_ID/${CaseName}.log
\ No newline at end of file
-- 
Gitee