diff --git a/PyTorch/contrib/cv/classification/GENet_for_Pytorch/README.md b/PyTorch/contrib/cv/classification/GENet_for_Pytorch/README.md
index 3a09b50b40d753d58f86cc14dc7a9ac75e2c6df5..53d619c91b063ef97659cc52cf784c28491cf2b3 100644
--- a/PyTorch/contrib/cv/classification/GENet_for_Pytorch/README.md
+++ b/PyTorch/contrib/cv/classification/GENet_for_Pytorch/README.md
@@ -51,9 +51,9 @@ All bash instructions output log files correctly.
 
 ## GENET training result
 
-| Acc@1    | FPS       | Device Type| Device Nums | Epochs   | AMP_Type |
-| :------: | :------:  | :------: | :------: | :------: |:------:
-|   94.73     | 1894.827   |   NPU | 1        | 300        | O2       |
-| 95.23   | 7858.025     |NPU  |8       | 300      | O2       |
-| 94.76    |  1350.074  |GPU  |1       | 300      | O2       |
-| 94.81    |  6536.289  |GPU  |8       | 300      | O2       |
\ No newline at end of file
+| Acc@1 |    FPS    | Device Type | Device Nums | Epochs | AMP_Type |
+|:-----:|:---------:|:-----------:|:-----------:|:------:|:--------:|
+| 94.73 | 2900.000  |     NPU     |      1      |  300   |    O2    |
+| 95.23 | 16912.437 |     NPU     |      8      |  300   |    O2    |
+| 94.76 | 1350.074  |     GPU     |      1      |  300   |    O2    |
+| 94.81 | 6536.289  |     GPU     |      8      |  300   |    O2    |
diff --git a/PyTorch/contrib/cv/classification/GENet_for_Pytorch/requirements.txt b/PyTorch/contrib/cv/classification/GENet_for_Pytorch/requirements.txt
index a6e282f9395b6d611a258d927e3a530a0376d8cf..57ad0ceb1253f5d1ade22e1b4f946c3437c06fed 100644
--- a/PyTorch/contrib/cv/classification/GENet_for_Pytorch/requirements.txt
+++ b/PyTorch/contrib/cv/classification/GENet_for_Pytorch/requirements.txt
@@ -1 +1,6 @@
-torchvision==0.5.0
\ No newline at end of file
+torchvision==0.9.1
+torch==1.8.1+ascend.rc2.20220601
+apex==0.1+ascend.20220601
+numpy==1.21.6
+decorator==5.1.1
+sympy==1.10.1
\ No newline at end of file
diff --git a/PyTorch/contrib/cv/classification/GENet_for_Pytorch/test/env_npu.sh b/PyTorch/contrib/cv/classification/GENet_for_Pytorch/test/env_npu.sh
index e8ca2d2c0da539d05fa6e57a5ac24a367dff1369..c34a5c85b4734e4568170bb9236224f5a1446dc1 100644
--- a/PyTorch/contrib/cv/classification/GENet_for_Pytorch/test/env_npu.sh
+++ b/PyTorch/contrib/cv/classification/GENet_for_Pytorch/test/env_npu.sh
@@ -34,23 +34,24 @@ ${install_path}/driver/tools/msnpureport -g error -d 5
 ${install_path}/driver/tools/msnpureport -g error -d 6
 ${install_path}/driver/tools/msnpureport -g error -d 7
 
-#将Host日志输出到串口,0-关闭/1-开启
+#��Host��־���������,0-�ر�/1-����
 export ASCEND_SLOG_PRINT_TO_STDOUT=0
-#设置默认日志级别,0-debug/1-info/2-warning/3-error
-export ASCEND_GLOBAL_LOG_LEVEL=3
-#设置Event日志开启标志,0-关闭/1-开启
+#����Ĭ����־����,0-debug/1-info/2-warning/3-error
+export ASCEND_GLOBAL_LOG_LEVEL==3
+#����Event��־������־,0-�ر�/1-����
 export ASCEND_GLOBAL_EVENT_ENABLE=0
-#设置是否开启taskque,0-关闭/1-开启
+#�����Ƿ���taskque,0-�ر�/1-����
 export TASK_QUEUE_ENABLE=1
-#设置是否开启PTCopy,0-关闭/1-开启
+#�����Ƿ���PTCopy,0-�ر�/1-����
 export PTCOPY_ENABLE=1
-#设置是否开启combined标志,0-关闭/1-开启
-export COMBINED_ENABLE=0
-#设置特殊场景是否需要重新编译,不需要修改
+#�����Ƿ���2��������combined��־,0-�ر�/1-����
+export COMBINED_ENABLE=1
+#�������ⳡ���Ƿ���Ҫ���±���,����Ҫ�޸�
 export DYNAMIC_OP="ADD#MUL"
-#HCCL白名单开关,1-关闭/0-开启
+# HCCL����������,1-�ر�/0-����
 export HCCL_WHITELIST_DISABLE=1
-export HCCL_IF_IP=$(hostname -I |awk '{print $1}')
+# HCCLĬ�ϳ�ʱʱ��120s���٣��޸�Ϊ1800s����PyTorchĬ������
+export HCCL_CONNECT_TIMEOUT=1800
 
 ulimit -SHn 512000
 
@@ -74,4 +75,3 @@ print(result)"""
 echo ${path_lib}
 
 export LD_LIBRARY_PATH=/usr/local/python3.7.5/lib/:${path_lib}:$LD_LIBRARY_PATH
-export PYTHONWARNINGS='ignore:semaphore_tracker:UserWarning'
\ No newline at end of file
diff --git a/PyTorch/contrib/cv/classification/GENet_for_Pytorch/test/train_eval_8p.sh b/PyTorch/contrib/cv/classification/GENet_for_Pytorch/test/train_eval_8p.sh
index 936b28289d12371bbfedd974f02699e5dd9f6d91..046a39e76fca115d16eadef6c267e559337fe524 100644
--- a/PyTorch/contrib/cv/classification/GENet_for_Pytorch/test/train_eval_8p.sh
+++ b/PyTorch/contrib/cv/classification/GENet_for_Pytorch/test/train_eval_8p.sh
@@ -100,7 +100,7 @@ echo "------------------ Final result ------------------"
 # 输出训练精度,需要模型审视修改
 train_err=`grep -a '* Err@1'  ${test_path_dir}/output/${ASCEND_DEVICE_ID}/train_${ASCEND_DEVICE_ID}.log|awk 'END {print}'|awk -F "Err@1" '{print $NF}'|awk -F " " '{print $1}'`
 #打印，不需要修改
-echo "Final Train Accuracy : ${train_accuracy}"
+echo "Final Train Accuracy: `awk 'BEGIN{printf "%.2f\n", '100'-'${train_err}'}'`"
 echo "E2E Training Duration sec : $e2e_time"
 
 # 训练用例信息，不需要修改
@@ -118,6 +118,6 @@ echo "RankSize = ${RANK_SIZE}" >>  ${test_path_dir}/output/$ASCEND_DEVICE_ID/${C
 echo "BatchSize = ${BatchSize}" >>  ${test_path_dir}/output/$ASCEND_DEVICE_ID/${CaseName}.log
 echo "DeviceType = ${DeviceType}" >>  ${test_path_dir}/output/$ASCEND_DEVICE_ID/${CaseName}.log
 echo "CaseName = ${CaseName}" >>  ${test_path_dir}/output/$ASCEND_DEVICE_ID/${CaseName}.log
-echo "TrainErr = ${train_err}" >> ${test_path_dir}/output/$ASCEND_DEVICE_ID/${CaseName}.log
+echo "TrainAcc = `awk 'BEGIN{printf "%.2f\n", '100'-'${train_err}'}'`" >>  ${test_path_dir}/output/$ASCEND_DEVICE_ID/${CaseName}.log
 echo "ActualLoss = ${ActualLoss}" >>  ${test_path_dir}/output/$ASCEND_DEVICE_ID/${CaseName}.log
 echo "E2ETrainingTime = ${e2e_time}" >>  ${test_path_dir}/output/$ASCEND_DEVICE_ID/${CaseName}.log
diff --git a/PyTorch/contrib/cv/classification/GENet_for_Pytorch/test/train_finetune_1p.sh b/PyTorch/contrib/cv/classification/GENet_for_Pytorch/test/train_finetune_1p.sh
index 9a808cd46f6498d28c01a9fc8d0c1ff3df1cda4e..12f492afde9b183bf464f82296cab4764095a764 100644
--- a/PyTorch/contrib/cv/classification/GENet_for_Pytorch/test/train_finetune_1p.sh
+++ b/PyTorch/contrib/cv/classification/GENet_for_Pytorch/test/train_finetune_1p.sh
@@ -111,7 +111,7 @@ echo "Final Performance images/sec : $FPS"
 #输出训练精度,需要模型审视修改
 train_err=`grep -a '* Err@1'  ${test_path_dir}/output/${ASCEND_DEVICE_ID}/train_${ASCEND_DEVICE_ID}.log|awk 'END {print}'|awk -F "Err@1" '{print $NF}'|awk -F " " '{print $1}'`
 #打印，不需要修改
-echo "Final Train Accuracy : ${train_Err}"
+echo "Final Train Accuracy: `awk 'BEGIN{printf "%.2f\n", '100'-'${train_err}'}'`"
 echo "E2E Training Duration sec : $e2e_time"
 
 #性能看护结果汇总
@@ -140,6 +140,6 @@ echo "DeviceType = ${DeviceType}" >>  ${test_path_dir}/output/$ASCEND_DEVICE_ID/
 echo "CaseName = ${CaseName}" >>  ${test_path_dir}/output/$ASCEND_DEVICE_ID/${CaseName}.log
 echo "ActualFPS = ${ActualFPS}" >>  ${test_path_dir}/output/$ASCEND_DEVICE_ID/${CaseName}.log
 echo "TrainingTime = ${TrainingTime}" >>  ${test_path_dir}/output/$ASCEND_DEVICE_ID/${CaseName}.log
-echo "TrainErr = ${train_err}" >> ${test_path_dir}/output/$ASCEND_DEVICE_ID/${CaseName}.log
+echo "TrainAcc = `awk 'BEGIN{printf "%.2f\n", '100'-'${train_err}'}'`" >>  ${test_path_dir}/output/$ASCEND_DEVICE_ID/${CaseName}.log
 echo "ActualLoss = ${ActualLoss}" >>  ${test_path_dir}/output/$ASCEND_DEVICE_ID/${CaseName}.log
 echo "E2ETrainingTime = ${e2e_time}" >>  ${test_path_dir}/output/$ASCEND_DEVICE_ID/${CaseName}.log
diff --git a/PyTorch/contrib/cv/classification/GENet_for_Pytorch/test/train_full_1p.sh b/PyTorch/contrib/cv/classification/GENet_for_Pytorch/test/train_full_1p.sh
index 755e9c108fbb8205b06657189c1fe25ff48f65df..695dae8527f991f2937c669577dc073511bb0752 100644
--- a/PyTorch/contrib/cv/classification/GENet_for_Pytorch/test/train_full_1p.sh
+++ b/PyTorch/contrib/cv/classification/GENet_for_Pytorch/test/train_full_1p.sh
@@ -111,8 +111,9 @@ echo "Final Performance images/sec : $FPS"
 
 #输出训练精度,需要模型审视修改
 train_err=`grep -a '* Err@1'  ${test_path_dir}/output/${ASCEND_DEVICE_ID}/train_${ASCEND_DEVICE_ID}.log|awk 'END {print}'|awk -F "Err@1" '{print $NF}'|awk -F " " '{print $1}'`
+
 #打印，不需要修改
-echo "Final Train Error : ${train_err}"
+echo "Final Train Accuracy: `awk 'BEGIN{printf "%.2f\n", '100'-'${train_err}'}'`"
 echo "E2E Training Duration sec : $e2e_time"
 
 #性能看护结果汇总
@@ -141,6 +142,6 @@ echo "DeviceType = ${DeviceType}" >>  ${test_path_dir}/output/$ASCEND_DEVICE_ID/
 echo "CaseName = ${CaseName}" >>  ${test_path_dir}/output/$ASCEND_DEVICE_ID/${CaseName}.log
 echo "ActualFPS = ${ActualFPS}" >>  ${test_path_dir}/output/$ASCEND_DEVICE_ID/${CaseName}.log
 echo "TrainingTime = ${TrainingTime}" >>  ${test_path_dir}/output/$ASCEND_DEVICE_ID/${CaseName}.log
-echo "TrainErr = ${train_err}" >> ${test_path_dir}/output/$ASCEND_DEVICE_ID/${CaseName}.log
+echo "TrainAcc = `awk 'BEGIN{printf "%.2f\n", '100'-'${train_err}'}'`" >>  ${test_path_dir}/output/$ASCEND_DEVICE_ID/${CaseName}.log
 echo "ActualLoss = ${ActualLoss}" >>  ${test_path_dir}/output/$ASCEND_DEVICE_ID/${CaseName}.log
 echo "E2ETrainingTime = ${e2e_time}" >>  ${test_path_dir}/output/$ASCEND_DEVICE_ID/${CaseName}.log
diff --git a/PyTorch/contrib/cv/classification/GENet_for_Pytorch/test/train_full_8p.sh b/PyTorch/contrib/cv/classification/GENet_for_Pytorch/test/train_full_8p.sh
index 6c1dde371ed4bda73e0d4e3e680cd24bb200c8b1..ca1702dec1089eb8d84dbed018ecc8028d397ca3 100644
--- a/PyTorch/contrib/cv/classification/GENet_for_Pytorch/test/train_full_8p.sh
+++ b/PyTorch/contrib/cv/classification/GENet_for_Pytorch/test/train_full_8p.sh
@@ -99,7 +99,7 @@ echo "Final Performance images/sec : $FPS"
 #输出训练精度,需要模型审视修改
 train_err=`grep -a '* Err@1'  ${test_path_dir}/output/${ASCEND_DEVICE_ID}/train_${ASCEND_DEVICE_ID}.log|awk 'END {print}'|awk -F "Err@1" '{print $NF}'|awk -F " " '{print $1}'`
 #打印，不需要修改
-echo "Final Train Err : ${train_err}"
+echo "Final Train Accuracy: `awk 'BEGIN{printf "%.2f\n", '100'-'${train_err}'}'`"
 echo "E2E Training Duration sec : $e2e_time"
 
 #性能看护结果汇总
@@ -128,6 +128,6 @@ echo "DeviceType = ${DeviceType}" >>  ${test_path_dir}/output/$ASCEND_DEVICE_ID/
 echo "CaseName = ${CaseName}" >>  ${test_path_dir}/output/$ASCEND_DEVICE_ID/${CaseName}.log
 echo "ActualFPS = ${ActualFPS}" >>  ${test_path_dir}/output/$ASCEND_DEVICE_ID/${CaseName}.log
 echo "TrainingTime = ${TrainingTime}" >>  ${test_path_dir}/output/$ASCEND_DEVICE_ID/${CaseName}.log
-echo "TrainErr = ${train_err}" >> ${test_path_dir}/output/$ASCEND_DEVICE_ID/${CaseName}.log
+echo "TrainAcc = `awk 'BEGIN{printf "%.2f\n", '100'-'${train_err}'}'`" >>  ${test_path_dir}/output/$ASCEND_DEVICE_ID/${CaseName}.log
 echo "ActualLoss = ${ActualLoss}" >>  ${test_path_dir}/output/$ASCEND_DEVICE_ID/${CaseName}.log
 echo "E2ETrainingTime = ${e2e_time}" >>  ${test_path_dir}/output/$ASCEND_DEVICE_ID/${CaseName}.log
diff --git a/PyTorch/contrib/cv/classification/GENet_for_Pytorch/test/train_performance_1p.sh b/PyTorch/contrib/cv/classification/GENet_for_Pytorch/test/train_performance_1p.sh
index 9121a82e1002003c1f3ac2132be301d337cc3abc..4e38ab94651036acfed62191606d604cf2c74cc2 100644
--- a/PyTorch/contrib/cv/classification/GENet_for_Pytorch/test/train_performance_1p.sh
+++ b/PyTorch/contrib/cv/classification/GENet_for_Pytorch/test/train_performance_1p.sh
@@ -107,10 +107,11 @@ FPS=`grep -a 'FPS'  ${test_path_dir}/output/${ASCEND_DEVICE_ID}/train_${ASCEND_D
 #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_err=`grep -a '* Err@1'  ${test_path_dir}/output/${ASCEND_DEVICE_ID}/train_${ASCEND_DEVICE_ID}.log|awk 'END {print}'|awk -F "Err@1" '{print $NF}'|awk -F " " '{print $1}'`
 #打印，不需要修改
-echo "Final Train Accuracy : ${train_accuracy}"
+echo "Final Train Accuracy: `awk 'BEGIN{printf "%.2f\n", '100'-'${train_err}'}'`"
 echo "E2E Training Duration sec : $e2e_time"
 
 #性能看护结果汇总
diff --git a/PyTorch/contrib/cv/classification/GENet_for_Pytorch/test/train_performance_8p.sh b/PyTorch/contrib/cv/classification/GENet_for_Pytorch/test/train_performance_8p.sh
index 451d35b5ac7be810fb0bacb165b228a9efa20a54..1994c2d057c93303aeb1f26b63a3f229178542cf 100644
--- a/PyTorch/contrib/cv/classification/GENet_for_Pytorch/test/train_performance_8p.sh
+++ b/PyTorch/contrib/cv/classification/GENet_for_Pytorch/test/train_performance_8p.sh
@@ -104,7 +104,7 @@ echo "Final Performance images/sec : $FPS"
 #输出训练精度,需要模型审视修改
 train_err=`grep -a '* Err@1'  ${test_path_dir}/output/${ASCEND_DEVICE_ID}/train_${ASCEND_DEVICE_ID}.log|awk 'END {print}'|awk -F "Err@1" '{print $NF}'|awk -F " " '{print $1}'`
 #打印，不需要修改
-echo "Final Train Accuracy : ${train_accuracy}"
+echo "Final Train Accuracy: `awk 'BEGIN{printf "%.2f\n", '100'-'${train_err}'}'`"
 echo "E2E Training Duration sec : $e2e_time"
 
 #性能看护结果汇总
diff --git a/PyTorch/contrib/cv/classification/GENet_for_Pytorch/train.py b/PyTorch/contrib/cv/classification/GENet_for_Pytorch/train.py
index 783fe250a92b7b2197c4dd6b8c27a725b1a4acce..2d459e44e02eae1ba109f17a341f342d1e722363 100644
--- a/PyTorch/contrib/cv/classification/GENet_for_Pytorch/train.py
+++ b/PyTorch/contrib/cv/classification/GENet_for_Pytorch/train.py
@@ -12,6 +12,12 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 # ============================================================================
+
+import sys
+import torch
+if torch.__version__ >= '1.8.1':
+    import torch_npu
+
 import argparse
 import json
 import os
@@ -25,12 +31,13 @@ import torchvision
 import torchvision.transforms as transforms
 import torch.multiprocessing as mp
 import random
-
+from torch import nn
 from models import *
+from models.wideresnet import WideResNet
 from utils import *
-
 import apex
 from apex import amp
+
 import warnings
 
 #Basic
@@ -267,7 +274,7 @@ def main_worker(gpu, ngpus_per_node, args):
     #amp 
     if args.amp:
         print('Using amp!')
-        model, optimizer = amp.initialize(model, optimizer, opt_level=args.opt_level, loss_scale=args.loss_scale)
+        model, optimizer = amp.initialize(model, optimizer, opt_level=args.opt_level, loss_scale=args.loss_scale, combine_grad=True)
     
     #DDP  
     if args.distributed:
@@ -325,7 +332,7 @@ def main_worker(gpu, ngpus_per_node, args):
         trainloader = torch.utils.data.DataLoader(trainset, batch_size=args.batch_size,
                                           shuffle=(train_sampler is None),
                                           num_workers=args.workers,
-                                          pin_memory=False,
+                                          pin_memory=True,
                                           sampler=train_sampler if not args.eval else None,
                                           drop_last=True)
     else:
@@ -425,34 +432,34 @@ def train(trainloader, model, criterion, optimizer, epoch, args, ngpus_per_node)
 
         # measure data loading time
         data_time.update(time.time() - end)
-        with torch.autograd.profiler.profile(use_cuda=True) as prof:  
-            if args.device == 'npu': 
-                loc = 'npu:{}'.format(args.gpu)
-                input = input.to(loc, non_blocking=True).to(torch.float)
-                target = target.to(torch.int32).to(loc, non_blocking=True)
-            else:
-                input = input.cuda(args.gpu, non_blocking=True)
-                target = target.cuda(args.gpu, non_blocking=True)     
+
+        if args.device == 'npu':
+            loc = 'npu:{}'.format(args.gpu)
+            input = input.to(loc, non_blocking=True).to(torch.float)
+            target = target.to(torch.int32).to(loc, non_blocking=True)
+        else:
+            input = input.cuda(args.gpu, non_blocking=True)
+            target = target.cuda(args.gpu, non_blocking=True)
                            
-            output = model(input)
-            loss = criterion(output, target)
+        output = model(input)
+        loss = criterion(output, target)
 
             # measure accuracy and record loss
-            err1, err5 = get_error(output.detach(), target, topk=(1, 5))
-            losses.update(loss.item(), input.size(0))
-            top1.update(err1.item(), input.size(0))
-            top5.update(err5.item(), input.size(0))
+        err1, err5 = get_error(output.detach(), target, topk=(1, 5))
+        losses.update(loss.item(), input.size(0))
+        top1.update(err1.item(), input.size(0))
+        top5.update(err5.item(), input.size(0))
             
-            if args.amp:
-                with amp.scale_loss(loss, optimizer) as scaled_loss:
-                    optimizer.zero_grad()
-                    scaled_loss.backward()
-            else:
+        if args.amp:
+            with amp.scale_loss(loss, optimizer) as scaled_loss:
                 optimizer.zero_grad()
-                loss.backward()
-            optimizer.step()
-            if args.device == 'npu':
-                torch.npu.synchronize()   
+                scaled_loss.backward()
+        else:
+            optimizer.zero_grad()
+            loss.backward()
+        optimizer.step()
+        if args.device == 'npu':
+            torch.npu.synchronize()
 
         # measure elapsed time
         if i == 9: