diff --git a/mindspore/lite/src/runtime/kernel/arm/fp16/group_convolution_fp16.cc b/mindspore/lite/src/runtime/kernel/arm/fp16/group_convolution_fp16.cc
index 9babb83e9a..3fded17066 100644
--- a/mindspore/lite/src/runtime/kernel/arm/fp16/group_convolution_fp16.cc
+++ b/mindspore/lite/src/runtime/kernel/arm/fp16/group_convolution_fp16.cc
@@ -52,19 +52,22 @@ int GroupConvolutionFP16CPUKernel::ReSize() {
 }
 
 void GroupConvolutionFP16CPUKernel::FreeSubKernel() {
-  for (auto sub_conv : group_convs_) {
+  for (auto &sub_conv : group_convs_) {
     // free sub conv input tensors / output tensors manually
     auto sub_in_tensors = sub_conv->in_tensors();
     auto sub_in_tensor_num = sub_in_tensors.size();
     for (size_t i = 0; i < sub_in_tensor_num; ++i) {
       delete sub_in_tensors[i];
+      sub_in_tensors[i] = nullptr;
     }
     auto sub_out_tensors = sub_conv->out_tensors();
     auto sub_out_tensor_num = sub_out_tensors.size();
     for (size_t i = 0; i < sub_out_tensor_num; ++i) {
       delete sub_out_tensors[i];
+      sub_out_tensors[i] = nullptr;
     }
     delete sub_conv;
+    sub_conv = nullptr;
   }
 }
 
diff --git a/mindspore/lite/src/runtime/kernel/arm/fp32/group_convolution_fp32.cc b/mindspore/lite/src/runtime/kernel/arm/fp32/group_convolution_fp32.cc
index 8d526de8a6..2624efb960 100644
--- a/mindspore/lite/src/runtime/kernel/arm/fp32/group_convolution_fp32.cc
+++ b/mindspore/lite/src/runtime/kernel/arm/fp32/group_convolution_fp32.cc
@@ -57,19 +57,22 @@ int GroupConvolutionCPUKernel::ReSize() {
 }
 
 void GroupConvolutionCPUKernel::FreeSubKernel() {
-  for (auto sub_conv : group_convs_) {
+  for (auto &sub_conv : group_convs_) {
     // free sub conv input tensors / output tensors manually
     auto sub_in_tensors = sub_conv->in_tensors();
     auto sub_in_tensor_num = sub_in_tensors.size();
     for (size_t i = 0; i < sub_in_tensor_num; ++i) {
       delete sub_in_tensors[i];
+      sub_in_tensors[i] = nullptr;
     }
     auto sub_out_tensors = sub_conv->out_tensors();
     auto sub_out_tensor_num = sub_out_tensors.size();
     for (size_t i = 0; i < sub_out_tensor_num; ++i) {
       delete sub_out_tensors[i];
+      sub_out_tensors[i] = nullptr;
     }
     delete sub_conv;
+    sub_conv = nullptr;
   }
 }
 
diff --git a/mindspore/lite/src/runtime/kernel/arm/int8/group_convolution_int8.h b/mindspore/lite/src/runtime/kernel/arm/int8/group_convolution_int8.h
index 2ef3444c88..1330d71794 100644
--- a/mindspore/lite/src/runtime/kernel/arm/int8/group_convolution_int8.h
+++ b/mindspore/lite/src/runtime/kernel/arm/int8/group_convolution_int8.h
@@ -33,7 +33,6 @@ class GroupConvolutionInt8CPUKernel : public GroupConvolutionCPUKernel {
       : GroupConvolutionCPUKernel(parameter, inputs, outputs, ctx, primitive, group_convs, group_num) {
   }  // opParameter(in channel, out channel) in this kernel has been split to groups, if
   // you want to get real params, multiply in channel / out channel with group num
-  ~GroupConvolutionInt8CPUKernel() override { GroupConvolutionCPUKernel::FreeSubKernel(); }
 
   int Run() override;
   void SeparateInput(int group_id) override;
diff --git a/mindspore/lite/test/models_tflite_awaretraining.cfg b/mindspore/lite/test/models_tflite_awaretraining.cfg
index 1cd319105f..3f3842761c 100644
--- a/mindspore/lite/test/models_tflite_awaretraining.cfg
+++ b/mindspore/lite/test/models_tflite_awaretraining.cfg
@@ -38,3 +38,4 @@ ssd_mobilenet_v1_1_default_1.tflite
 object_detection_mobile_object_localizer_v1_1_default_1.tflite
 gts_detect_0730_quant_frozen.tflite
 gts_model_quant_frozen.tflite
+inception_v2_224_quant.tflite