fix space_to_batch_int8 padding bug

mindspore/lite/nnacl/int8/space_to_batch_int8.c

@@ -16,8 +16,7 @@
 #include "nnacl/int8/space_to_batch_int8.h"
 #include "nnacl/arithmetic_common.h"
 
-void DoSpaceToBatchNHWCInt8(const int8_t *input, int8_t *output, int *block_sizes, int *in_shape,
-                            int *out_shape) {
+void DoSpaceToBatchNHWCInt8(const int8_t *input, int8_t *output, int *block_sizes, int *in_shape, int *out_shape) {
   int out_dim0 = out_shape[0];
   int out_dim1 = out_shape[1];
   int out_dim2 = out_shape[2];
@@ -46,7 +45,8 @@ void DoSpaceToBatchNHWCInt8(const int8_t *input, int8_t *output, int *block_size
   }
 }
 
-void DoSpaceToBatchPaddingNHWCInt8(const int8_t *input, int8_t *output, int *in_shape, int *padding, int *out_shape) {
+void DoSpaceToBatchPaddingNHWCInt8(const int8_t *input, int8_t *output, int *in_shape, int *padding, int *out_shape,
+                                   int32_t zp) {
   int in_h = in_shape[1];
   int in_w = in_shape[2];
   int in_c = in_shape[3];
@@ -64,13 +64,13 @@ void DoSpaceToBatchPaddingNHWCInt8(const int8_t *input, int8_t *output, int *in_
   for (int i = 0; i < in_shape[0]; ++i) {
     size_t in_offset0 = i * in_strides[0];
     for (int pad_h_top = 0; pad_h_top < padding[0]; ++pad_h_top) {
-        memset(output + out_offset, 0, ped_h_size);
-        out_offset += ped_h_num;
+      memset(output + out_offset, zp, ped_h_size);
+      out_offset += ped_h_num;
     }
     for (int j = 0; j < in_h; ++j) {
       size_t in_offset1 = in_offset0 + j * in_strides[1];
       for (int pad_w_left = 0; pad_w_left < padding[2]; ++pad_w_left) {
-        memset(output + out_offset, 0, ped_w_size);
+        memset(output + out_offset, zp, ped_w_size);
         out_offset += out_c;
       }
       for (int k = 0; k < in_w; ++k) {
@@ -79,12 +79,12 @@ void DoSpaceToBatchPaddingNHWCInt8(const int8_t *input, int8_t *output, int *in_
         out_offset += in_c;
       }
       for (int pad_w_right = 0; pad_w_right < padding[3]; ++pad_w_right) {
-        memset(output + out_offset, 0, ped_w_size);
+        memset(output + out_offset, zp, ped_w_size);
         out_offset += out_c;
       }
     }
     for (int pad_h_bottom = 0; pad_h_bottom < padding[1]; ++pad_h_bottom) {
-      memset(output + out_offset, 0, ped_h_size);
+      memset(output + out_offset, zp, ped_h_size);
       out_offset += ped_h_num;
     }
   }

mindspore/lite/nnacl/int8/space_to_batch_int8.h

@@ -22,7 +22,8 @@
 extern "C" {
 #endif
 void DoSpaceToBatchNHWCInt8(const int8_t *input, int8_t *output, int *block_sizes, int *in_shape, int *out_shape);
-void DoSpaceToBatchPaddingNHWCInt8(const int8_t *input, int8_t *output, int *in_shape, int *padding, int *out_shape);
+void DoSpaceToBatchPaddingNHWCInt8(const int8_t *input, int8_t *output, int *in_shape, int *padding, int *out_shape,
+                                   int32_t zp);
 #ifdef __cplusplus
 }
 #endif

mindspore/lite/src/runtime/kernel/arm/int8/space_to_batch_int8.cc

@@ -36,6 +36,11 @@ int SpaceToBatchInt8CPUKernel::Run() {
   auto input_ptr = reinterpret_cast<const int8_t *>(input_tensor->MutableData());
   auto output_ptr = reinterpret_cast<int8_t *>(output_tensor->MutableData());
   SpaceToBatchParameter *param = reinterpret_cast<SpaceToBatchParameter *>(this->op_parameter_);
+  if (output_tensor->GetQuantParams().empty()) {
+    MS_LOG(ERROR) << "SpaceToBatchInt8 need quantization parameters which is not found.";
+    return RET_ERROR;
+  }
+  auto quant_arg = output_tensor->GetQuantParams().front();
 
   if (param->need_paddings_) {
     padded_input_ = context_->allocator->Malloc(param->padded_input_element_num * sizeof(int8_t));
@@ -45,7 +50,7 @@ int SpaceToBatchInt8CPUKernel::Run() {
     }
     auto padded_input = reinterpret_cast<int8_t *>(padded_input_);
     DoSpaceToBatchPaddingNHWCInt8(input_ptr, padded_input, param->input_shape_, param->paddings_,
-                                  param->padded_in_shape_);
+                                  param->padded_in_shape_, quant_arg.zeroPoint);
     DoSpaceToBatchNHWCInt8(padded_input, output_ptr, param->block_sizes_, param->padded_in_shape_,
                            param->output_shape_);
     FreeTmpBuffer();