diff --git a/PyTorch/dev/cv/image_classification/SRCNN_ID1770_for_PyTorch/test/train_performance_8p.sh b/PyTorch/dev/cv/image_classification/SRCNN_ID1770_for_PyTorch/test/train_performance_8p.sh
new file mode 100644
index 0000000000000000000000000000000000000000..0e9f1aadebac8f1c59168fe8c6820fd41cd7a0e9
--- /dev/null
+++ b/PyTorch/dev/cv/image_classification/SRCNN_ID1770_for_PyTorch/test/train_performance_8p.sh
@@ -0,0 +1,145 @@
+#!/bin/bash
+
+cur_path=`pwd`/../
+path=`pwd`
+#失败用例打屏
+export ASCEND_GLOBAL_LOG_LEVEL=3
+
+export HDF5_DISABLE_VERSION_CHECK=1
+
+export PATH=/usr/local/hdf5/bin:$PATH
+export LD_LIBRARY_PATH=/usr/local/hdf5/lib:$LD_LIBRARY_PATH
+export LIBRARY_PATH=/usr/local/hdf5/lib:$LIBRARY_PATH
+export CPATH=/usr/local/hdf5/include:$CPATH
+
+#基础参数，需要模型审视修改
+#Batch Size
+batch_size=160
+#网络名称，同目录名称
+Network="SRCNN_ID1770_for_PyTorch"
+#Device数量，单卡默认为1
+RankSize=8
+RANK_SIZE=8
+#训练epoch，可选
+train_epochs=10
+#训练step
+train_steps=
+#学习率
+learning_rate=1e-3
+
+num_workers=192
+
+#参数配置
+data_path=""
+PREC="--apex --apex_opt_level O2"
+
+if [[ $1 == --help || $1 == --h ]];then
+	echo "usage:./train_performance_1p.sh "
+	exit 1
+fi
+
+for para in $*
+do
+	if [[ $para == --precision_mode* ]];then
+        apex_opt_level=`echo ${para#*=}`
+		if [[ $apex_opt_level != "O1" ]] && [[ $apex_opt_level != "O2" ]] && [[ $apex_opt_level != "O3" ]]; then
+			echo "[ERROR] para \"precision_mode\" must be config O1 or O2 or O3"
+			exit 1
+		fi
+        PREC="--apex --apex-opt-level "$apex_opt_level
+	elif [[ $para == --data_path* ]];then
+		data_path=`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
+
+#sed -i "s|THUCNews|$data_path|g" ${cur_path}/run.py
+#sed -i "s|self.num_epochs = 10|self.num_epochs = 1|g" ${cur_path}/models/TextRNN.p
+export NPU_CALCULATE_DEVICE=$ASCEND_DEVICE_ID
+
+start=$(date +%s)
+export MASTER_ADDR=localhost
+export MASTER_PORT=29688
+export HCCL_WHITELIST_DISABLE=1
+
+NPUS=($(seq 0 7))
+export NPU_WORLD_SIZE=${#NPUS[@]}
+rank=0
+for i in ${NPUS[@]}
+do
+    mkdir -p  $cur_path/output/${i}/
+    export NPU_CALCULATE_DEVICE=${i}
+    export RANK=${rank}
+    export ASCEND_DEVICE_ID=${i}
+    echo run process ${rank}
+    python3 train_8p.py $PREC --ddp --train-file "$data_path/SRCNN/91-image_x2.h5" \
+                       --eval-file "$data_path/SRCNN/Set5_x2.h5" \
+                       --outputs-dir "outputs" \
+                       --scale 3 \
+                       --lr $learning_rate \
+                       --batch-size $batch_size \
+                       --num-epochs $train_epochs  \
+                       --num-workers $num_workers \
+                       --seed 123 > $cur_path/test/output/$ASCEND_DEVICE_ID/train_${i}.log 2>&1 &
+    let rank++
+done
+wait
+end=$(date +%s)
+e2e_time=$(( $end - $start ))
+sed -i "s|\r|\n|g" $cur_path/test/output/$ASCEND_DEVICE_ID/train_$ASCEND_DEVICE_ID.log
+#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 "FPS"  $path/output/${ASCEND_DEVICE_ID}/train_${ASCEND_DEVICE_ID}.log|awk -F ", loss" '{print$1}' | awk '{print$NF}' |awk '{sum+=$1} END {print"",sum*8/NR}'`
+#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 "Acc" $path/output/${ASCEND_DEVICE_ID}/train_${ASCEND_DEVICE_ID}.log|awk -F ',' '{print $1}'|tail -1|awk '{print $NF}'`
+#打印，不需要修改
+#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}_${RANK_SIZE}'p'_'perf'
+
+##获取性能数据，不需要修改
+ActualFPS=${FPS}
+#ActualAccuracy=${train_accuracy}
+#单迭代训练时长
+TrainingTime=`echo "${BatchSize} ${FPS}"|awk '{printf("%.4f\n", $1*1000/$2)}'`
+
+#从train_$ASCEND_DEVICE_ID.log提取Loss到train_${CaseName}_loss.txt中，需要根据模型审视
+# grep -r -B 3 'eval psnr' $path/output/$ASCEND_DEVICE_ID/train_$ASCEND_DEVICE_ID.log | grep epoch | grep psnr -B 1 | grep 21904/21904 | awk -F "loss=" '{print$2}' | sed 's/]//g' > $path/output/$ASCEND_DEVICE_ID/train_${CaseName}_loss.txt
+# grep 21904/21904 $path/output/$ASCEND_DEVICE_ID/train_$ASCEND_DEVICE_ID.log | awk -F "loss=" 'END {print$2}' | sed 's/]//g' >> $path/output/$ASCEND_DEVICE_ID/train_${CaseName}_loss.txt
+grep "FPS"  $path/output/${ASCEND_DEVICE_ID}/train_${ASCEND_DEVICE_ID}.log| awk '{print$NF}' > $path/output/$ASCEND_DEVICE_ID/train_${CaseName}_loss.txt
+
+#最后一个迭代loss值，不需要修改
+ActualLoss=`awk 'END {print}' $path/output/$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}" > $path/output/$ASCEND_DEVICE_ID/${CaseName}.log
+echo "RankSize = ${RankSize}" >> $path/output/$ASCEND_DEVICE_ID/${CaseName}.log
+echo "BatchSize = ${BatchSize}" >> $path/output/$ASCEND_DEVICE_ID/${CaseName}.log
+echo "DeviceType = ${DeviceType}" >> $path/output/$ASCEND_DEVICE_ID/${CaseName}.log
+echo "CaseName = ${CaseName}" >> $path/output/$ASCEND_DEVICE_ID/${CaseName}.log
+echo "ActualFPS = ${ActualFPS}" >> $path/output/$ASCEND_DEVICE_ID/${CaseName}.log
+echo "TrainingTime = ${TrainingTime}" >> $path/output/$ASCEND_DEVICE_ID/${CaseName}.log
+echo "ActualLoss = ${ActualLoss}" >> $path/output/$ASCEND_DEVICE_ID/${CaseName}.log
+echo "E2ETrainingTime = ${e2e_time}" >> $path/output/$ASCEND_DEVICE_ID/${CaseName}.log
diff --git a/PyTorch/dev/cv/image_classification/SRCNN_ID1770_for_PyTorch/train_8p.py b/PyTorch/dev/cv/image_classification/SRCNN_ID1770_for_PyTorch/train_8p.py
new file mode 100644
index 0000000000000000000000000000000000000000..30709a0c50bc5fcbabffea442038a0fddacc33ca
--- /dev/null
+++ b/PyTorch/dev/cv/image_classification/SRCNN_ID1770_for_PyTorch/train_8p.py
@@ -0,0 +1,249 @@
+#
+# BSD 3-Clause License
+#
+# Copyright (c) 2017 xxxx
+# All rights reserved.
+# Copyright 2021 Huawei Technologies Co., Ltd
+#
+# Redistribution and use in source and binary forms, with or without
+# modification, are permitted provided that the following conditions are met:
+#
+# * Redistributions of source code must retain the above copyright notice, this
+#   list of conditions and the following disclaimer.
+#
+# * Redistributions in binary form must reproduce the above copyright notice,
+#   this list of conditions and the following disclaimer in the documentation
+#   and/or other materials provided with the distribution.
+#
+# * Neither the name of the copyright holder nor the names of its
+#   contributors may be used to endorse or promote products derived from
+#   this software without specific prior written permission.
+#
+# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+# AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+# IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
+# DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE
+# FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
+# DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+# SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER
+# CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
+# OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+# ============================================================================
+#
+import argparse
+import os
+import copy
+
+import torch
+from torch import nn
+import torch.optim as optim
+import torch.backends.cudnn as cudnn
+from torch.utils.data.dataloader import DataLoader
+from tqdm import tqdm
+import time
+from models import SRCNN
+from datasets import TrainDataset, EvalDataset
+from utils import AverageMeter, calc_psnr
+import torch.npu
+import os
+
+# print("1111111111")
+# import torch
+# torch.npu.global_step_inc()
+# print("2222222222")
+
+# 使能混合精度
+try:
+    from apex import amp
+except ImportError:
+    amp = None
+import apex
+# 使能混合精度
+
+NPU_CALCULATE_DEVICE = 0
+if os.getenv('NPU_CALCULATE_DEVICE') and str.isdigit(os.getenv('NPU_CALCULATE_DEVICE')):
+    NPU_CALCULATE_DEVICE = int(os.getenv('NPU_CALCULATE_DEVICE'))
+if torch.npu.current_device() != NPU_CALCULATE_DEVICE:
+    torch.npu.set_device(f'npu:{NPU_CALCULATE_DEVICE}')
+
+
+if __name__ == '__main__':
+
+    # 开启模糊编译
+    # print("ttttttttttttttttttt")
+    # import torch
+    # torch.npu.global_step_inc()
+    # print("zzzzzzzzzzzzzzzz")
+    # 开启模糊编译
+
+    parser = argparse.ArgumentParser()
+    parser.add_argument('--train-file', type=str, required=True)
+    parser.add_argument('--eval-file', type=str, required=True)
+    parser.add_argument('--outputs-dir', type=str, required=True)
+    parser.add_argument('--scale', type=int, default=3)
+    parser.add_argument('--lr', type=float, default=1e-4)
+    parser.add_argument('--batch-size', type=int, default=16)
+    parser.add_argument('--num-epochs', type=int, default=400)
+    parser.add_argument('--num-workers', type=int, default=128)
+    parser.add_argument('--seed', type=int, default=123)
+    parser.add_argument('--ddp', action='store_true', help='default 1p')
+
+    # 使能混合精度
+    parser.add_argument('--apex', action='store_true',
+                        help='Use apex for mixed precision training')
+    parser.add_argument('--apex_opt_level', default='O1', type=str,
+                        help='For apex mixed precision training'
+                             'O0 for FP32 training, O1 for mixed precision training.'
+                             'For further detail, see https://github.com/NVIDIA/apex/tree/master/examples/imagenet')
+    parser.add_argument('--loss_scale_value', default=1024., type=float,
+                        help='loss scale using in amp, default -1 means dynamic')
+    # 使能混合精度
+    args = parser.parse_args()
+
+    if args.ddp:
+        NPU_CALCULATE_DEVICE = 0
+        if os.getenv('NPU_CALCULATE_DEVICE') and str.isdigit(os.getenv('NPU_CALCULATE_DEVICE')):
+            NPU_CALCULATE_DEVICE = int(os.getenv('NPU_CALCULATE_DEVICE'))
+        if torch.npu.current_device() != NPU_CALCULATE_DEVICE:
+            torch.npu.set_device(f'npu:{NPU_CALCULATE_DEVICE}')
+        NPU_WORLD_SIZE = int(os.getenv('NPU_WORLD_SIZE'))
+        RANK = int(os.getenv('RANK'))
+        torch.distributed.init_process_group('hccl', rank=RANK, world_size=NPU_WORLD_SIZE)
+
+    args.outputs_dir = os.path.join(args.outputs_dir, 'x{}'.format(args.scale))
+
+    if not os.path.exists(args.outputs_dir):
+        os.makedirs(args.outputs_dir)
+
+    cudnn.benchmark = True
+    device = torch.device(f'npu:{NPU_CALCULATE_DEVICE}')
+
+    torch.manual_seed(args.seed)
+
+    model = SRCNN().to(f'npu:{NPU_CALCULATE_DEVICE}')
+    criterion = nn.MSELoss()
+    #使用PT原生优化器接口
+    # optimizer = optim.Adam([
+    #     {'params': model.conv1.parameters()},
+    #     {'params': model.conv2.parameters()},
+    #     {'params': model.conv3.parameters(), 'lr': args.lr * 0.1}
+    # ], lr=args.lr)
+
+    #替换亲和性接口，但是不合并paras
+    # optimizer = apex.optimizers.NpuFusedAdam([
+    #     {'params': model.conv1.parameters()},
+    #     {'params': model.conv2.parameters()},
+    #     {'params': model.conv3.parameters(), 'lr': args.lr * 0.1}
+    # ], lr=args.lr)
+
+    #替换亲和性接口，同时将合并第1和第2组paras
+    optimizer = apex.optimizers.NpuFusedAdam([
+        {'params': list(model.conv1.parameters()) + list(model.conv2.parameters())},
+        {'params': model.conv3.parameters(), 'lr': args.lr * 0.1}
+    ], lr=args.lr)
+
+
+
+
+    if args.apex:
+        # print("args.apex=============================", args.apex)
+        model, optimizer = amp.initialize(model, optimizer,
+                                           opt_level=args.apex_opt_level,
+                                           loss_scale=args.loss_scale_value,
+                                           combine_grad=True)
+    #使能混合精度
+    if args.ddp:
+        model = model.to(f'npu:{NPU_CALCULATE_DEVICE}')
+        if not isinstance(model, torch.nn.parallel.DistributedDataParallel):
+            model = torch.nn.parallel.DistributedDataParallel(model, device_ids=[NPU_CALCULATE_DEVICE],
+                                                              broadcast_buffers=False)
+
+    train_dataset = TrainDataset(args.train_file)
+    eval_dataset = EvalDataset(args.eval_file)
+    if args.ddp:
+        train_dataloader_sampler = torch.utils.data.distributed.DistributedSampler(train_dataset)
+        train_dataloader_batch_size = args.batch_size
+        train_dataloader = DataLoader(dataset=train_dataset,
+                                      batch_size=train_dataloader_batch_size,
+                                      shuffle=False,
+                                      num_workers=args.num_workers,
+                                      pin_memory=True,
+                                      drop_last=True, sampler=train_dataloader_sampler)
+        eval_dataloader = DataLoader(dataset=eval_dataset, batch_size=1)
+    else:
+        train_dataloader = DataLoader(dataset=train_dataset,
+                                      batch_size=args.batch_size,
+                                      shuffle=True,
+                                      num_workers=args.num_workers,
+                                      pin_memory=True,
+                                      drop_last=True)
+        eval_dataloader = DataLoader(dataset=eval_dataset, batch_size=1)
+
+    best_weights = copy.deepcopy(model.state_dict())
+    best_epoch = 0
+    best_psnr = 0.0
+
+    for epoch in range(args.num_epochs):
+        if args.ddp:
+            train_dataloader.sampler.set_epoch(epoch)
+        model.train()
+        epoch_losses = AverageMeter()
+
+        # with tqdm(total=(len(train_dataset) - len(train_dataset) % args.batch_size)) as t:
+        #     t.set_description('epoch: {}/{}'.format(epoch, args.num_epochs - 1))
+
+        # iter = 0
+	start_time = time.time()
+        # print("train_dataloader length============================", len(train_dataloader))
+        
+        
+        for i, (inputs, labels) in enumerate(train_dataloader):
+            
+	    inputs = inputs.to(f'npu:{NPU_CALCULATE_DEVICE}', non_blocking=True)
+	    labels = labels.to(f'npu:{NPU_CALCULATE_DEVICE}', non_blocking=True)
+                    
+	    preds = model(inputs)
+	    loss = criterion(preds, labels)
+	    epoch_losses.update(loss.item(), len(inputs))
+	    optimizer.zero_grad()
+
+	    # 使能混合精度
+	    # loss.backward()
+	    if args.apex:
+	    	with amp.scale_loss(loss, optimizer) as scaled_loss:
+	            scaled_loss.backward()
+	    else:
+	        loss.backward()
+	    # 使能混合精度
+	    optimizer.step()
+	    FPS = args.batch_size/(time.time()-start_time)
+	    start_time = time.time()
+	    print("Epoch: {}, iter: {}, FPS: {:.4f}, loss: {:.6f}".format(epoch, i, FPS, epoch_losses.avg))
+
+
+        torch.save(model.state_dict(), os.path.join(args.outputs_dir, 'epoch_{}.pth'.format(epoch)))
+
+        model.eval()
+        epoch_psnr = AverageMeter()
+
+        for data in eval_dataloader:
+            inputs, labels = data
+
+            inputs = inputs.to(f'npu:{NPU_CALCULATE_DEVICE}')
+            labels = labels.to(f'npu:{NPU_CALCULATE_DEVICE}')
+
+            with torch.no_grad():
+                preds = model(inputs).clamp(0.0, 1.0)
+
+            epoch_psnr.update(calc_psnr(preds, labels), len(inputs))
+
+        print('eval psnr: {:.2f}'.format(epoch_psnr.avg))
+
+        if epoch_psnr.avg > best_psnr:
+            best_epoch = epoch
+            best_psnr = epoch_psnr.avg
+            best_weights = copy.deepcopy(model.state_dict())
+
+    print('best epoch: {}, psnr: {:.2f}'.format(best_epoch, best_psnr))
+    torch.save(best_weights, os.path.join(args.outputs_dir, 'best.pth'))
\ No newline at end of file