diff --git a/mindspore/lite/src/runtime/kernel/arm/fp16/concat_fp16.cc b/mindspore/lite/src/runtime/kernel/arm/fp16/concat_fp16.cc
index 905d024f5e..d4137c09b6 100644
--- a/mindspore/lite/src/runtime/kernel/arm/fp16/concat_fp16.cc
+++ b/mindspore/lite/src/runtime/kernel/arm/fp16/concat_fp16.cc
@@ -41,15 +41,7 @@ int ConcatFp16CPUKernel::Init() {
   return ReSize();
 }
 
-int ConcatFp16CPUKernel::ReSize() {
-  FreeTmpBuffer();
-  auto ret = MallocTmpBuffer();
-  if (ret != RET_OK) {
-    FreeTmpBuffer();
-    return ret;
-  }
-  return ConcatBaseCPUKernel::ReSize();
-}
+int ConcatFp16CPUKernel::ReSize() { return ConcatBaseCPUKernel::ReSize(); }
 
 int ConcatFp16CPUKernel::MallocTmpBuffer() {
   for (const auto &in_tensor : in_tensors_) {
@@ -105,6 +97,13 @@ int ConcatFp16CPUKernel::Run() {
     MS_LOG(ERROR) << "Prepare fail!ret: " << prepare_ret;
     return prepare_ret;
   }
+
+  auto ret = MallocTmpBuffer();
+  if (ret != RET_OK) {
+    FreeTmpBuffer();
+    return ret;
+  }
+
   auto input_num = in_tensors_.size();
   std::vector<int *> inputs_output_shape(input_num + 1, nullptr);
 
diff --git a/mindspore/lite/src/runtime/kernel/arm/fp16/reduce_fp16.cc b/mindspore/lite/src/runtime/kernel/arm/fp16/reduce_fp16.cc
index 8ba4cb05f1..c17baab5a2 100644
--- a/mindspore/lite/src/runtime/kernel/arm/fp16/reduce_fp16.cc
+++ b/mindspore/lite/src/runtime/kernel/arm/fp16/reduce_fp16.cc
@@ -58,17 +58,7 @@ int ReduceFp16CPUKernel::Init() {
 }
 
 int ReduceFp16CPUKernel::ReSize() {
-  FreeTmpBuffer();
-  auto ret = ReduceBaseCPUKernel::ReSize();
-  if (ret != RET_OK) {
-    return ret;
-  }
-  ret = MallocTmpBuffer();
-  if (ret != RET_OK) {
-    FreeTmpBuffer();
-    return ret;
-  }
-  return RET_OK;
+  return ReduceBaseCPUKernel::ReSize();
 }
 
 int ReduceFp16CPUKernel::CallReduceUnit(int task_id) {
@@ -94,6 +84,12 @@ int ReduceFp16CPUKernel::Run() {
     return prepare_ret;
   }
 
+  auto ret = MallocTmpBuffer();
+  if (ret != RET_OK) {
+    FreeTmpBuffer();
+    return ret;
+  }
+
   tmp_shape_ = in_tensors_.at(0)->shape();
   auto in_tensor = in_tensors_.at(0);
   if (in_tensor->data_type() == kNumberTypeFloat32 || in_tensor->data_type() == kNumberTypeFloat) {
diff --git a/mindspore/lite/src/runtime/kernel/arm/fp16/transpose_fp16.cc b/mindspore/lite/src/runtime/kernel/arm/fp16/transpose_fp16.cc
index 3c0433a509..efd45f5213 100644
--- a/mindspore/lite/src/runtime/kernel/arm/fp16/transpose_fp16.cc
+++ b/mindspore/lite/src/runtime/kernel/arm/fp16/transpose_fp16.cc
@@ -59,12 +59,6 @@ int TransposeFp16CPUKernel::ReSize() {
     param->out_strides_[i] = out_shape[i + 1] * param->out_strides_[i + 1];
   }
 
-  FreeFp16Buffer();
-  auto ret = MallocFp16Buffer();
-  if (ret != RET_OK) {
-    FreeFp16Buffer();
-    return ret;
-  }
   return RET_OK;
 }
 
@@ -149,10 +143,16 @@ int TransposeFp16CPUKernel::Run() {
   auto &out_tensor = out_tensors_.front();
   if (in_tensor == nullptr || out_tensor == nullptr) {
     MS_LOG(ERROR) << "null pointer referencing.";
-    FreeFp16Buffer();
     return RET_ERROR;
   }
 
+  // malloc when Run
+  ret = MallocFp16Buffer();
+  if (ret != RET_OK) {
+    FreeFp16Buffer();
+    return ret;
+  }
+
   if (in_tensor->data_type() == kNumberTypeFloat || in_tensor->data_type() == kNumberTypeFloat32) {
     in_data_ = reinterpret_cast<float *>(in_tensor->Data());
     Float32ToFloat16(in_data_, fp16_in_data_, in_tensor->ElementsNum());