fix conv2d_grad_inpu infer strides.

mindspore/ccsrc/backend/kernel_compiler/cpu/cpu_kernel.h

@@ -34,6 +34,7 @@ const char KERNEL_SIZE[] = "kernel_size";
 const char STRIDE[] = "stride";
 const char STRIDES[] = "strides";
 const char DILATION[] = "dilation";
+const char FORMAT[] = "format";
 const char PAD[] = "pad";
 const char PAD_LIST[] = "pad_list";
 const char PAD_MODE[] = "pad_mode";

mindspore/ccsrc/backend/kernel_compiler/cpu/mkldnn/conv2d_grad_input_cpu_kernel.cc

@@ -15,6 +15,7 @@
  */
 #include "backend/kernel_compiler/cpu/mkldnn/conv2d_grad_input_cpu_kernel.h"
 #include <string>
+#include <map>
 #include <algorithm>
 #include "backend/kernel_compiler/cpu/mkldnn/mkl_kernel_engine.h"
 #include "runtime/device/cpu/cpu_device_address.h"
@@ -22,6 +23,7 @@
 
 namespace mindspore {
 namespace kernel {
+const std::map<std::string, size_t> kFormatIndexMap = {{"NCHW", 2}, {"HWCN", 0}, {"NHWC", 1}};
 void Conv2dGradInputCPUKernel::InitKernel(const CNodePtr &kernel_node) {
   MS_EXCEPTION_IF_NULL(kernel_node);
   std::vector<size_t> src_shape = AnfAlgo::GetOutputDeviceShape(kernel_node, 0);
@@ -47,7 +49,16 @@ void Conv2dGradInputCPUKernel::InitKernel(const CNodePtr &kernel_node) {
   std::vector<int> dilation_ori;
   auto stride_me = AnfAlgo::GetNodeAttr<std::vector<int64_t>>(kernel_node, STRIDE);
   auto dilation_me = AnfAlgo::GetNodeAttr<std::vector<int64_t>>(kernel_node, DILATION);
-  (void)std::transform(stride_me.begin(), stride_me.end(), std::back_inserter(stride_ori),
+  auto format_me = AnfAlgo::GetNodeAttr<std::string>(kernel_node, FORMAT);
+  auto iter = kFormatIndexMap.find(format_me);
+  if (iter == kFormatIndexMap.end()) {
+    MS_LOG(EXCEPTION) << "OriFormat is " << format_me << ", Please confirm that in {NCHW, HWCN, NHWC}.";
+  }
+  size_t h_index = iter->second;
+  if (stride_me.size() < h_index + 2) {
+    MS_LOG(EXCEPTION) << "Strides should greater than " << h_index + 1 << ", but got " << stride_me.size();
+  }
+  (void)std::transform(stride_me.begin() + h_index, stride_me.begin() + h_index + 2, std::back_inserter(stride_ori),
                        [](const int64_t &value) { return static_cast<int>(value); });
   (void)std::transform(dilation_me.begin(), dilation_me.end(), std::back_inserter(dilation_ori),
                        [](const int64_t &value) { return static_cast<int>(value); });

mindspore/ops/operations/nn_ops.py

@@ -2028,8 +2028,8 @@ class Conv2DBackpropInput(PrimitiveWithInfer):
             [dout_shape[0], dout_shape[2], dout_shape[3], dout_shape[1]]
         kernel_h = self.kernel_size[0]
         kernel_w = self.kernel_size[1]
-        stride_h = self.stride[0]
-        stride_w = self.stride[1]
+        stride_h = self.stride[2]
+        stride_w = self.stride[3]
         dilation_h = self.dilation[2]
         dilation_w = self.dilation[3]
         # default pad mode is valid