[x86] fix ai-rank model (#7678) (#7690)

Author: mjp9527

lite/backends/x86/math/avx/conv_utils.cc

@@ -868,27 +868,26 @@ void im2col_s1<float>(const float* data_im,
   memset(data_col, 0, mem_size);
 #pragma omp parallel for
   for (int c = 0; c < channels; c++) {
-    unsigned int data_im_z = c * in_channel_size;
-    unsigned int data_col_z1 = c * output_plane_size;
+    unsigned int data_im_z = static_cast<unsigned int>(c * in_channel_size);
+    int data_col_z1 = c * output_plane_size;
     for (int ky = 0, h_offset = 0; ky < kernel_h;
          ky++, h_offset += dilation_h) {
-      unsigned int data_col_z2 = ky * out_channel_size * kernel_w;
+      int data_col_z2 = ky * out_channel_size * kernel_w;
       for (int kx = 0, w_offset = 0; kx < kernel_w;
            kx++, w_offset += dilation_w) {
-        unsigned int data_col_z3 = kx * out_channel_size;
-        unsigned int data_col_z = data_col_z1 + data_col_z2 + data_col_z3;
-        unsigned int oh_begin = std::max(((pad_top - h_offset)), 0);
-        unsigned int oh_end =
-            std::min(((height + pad_bottom - h_offset)), output_h);
+        int data_col_z3 = kx * out_channel_size;
+        unsigned int data_col_z =
+            static_cast<unsigned int>(data_col_z1 + data_col_z2 + data_col_z3);
+        int oh_begin = std::max(((pad_top - h_offset)), 0);  // always >= 0
+        int oh_end = std::min(((height + pad_bottom - h_offset)), output_h);
         oh_end = std::max(oh_begin, oh_end);
-        unsigned int ow_begin = std::max(((pad_left - w_offset)), 0);
-        unsigned int ow_end =
-            std::min(((width + pad_right - w_offset)), output_w);
+        int ow_begin = std::max(((pad_left - w_offset)), 0);
+        int ow_end = std::min(((width + pad_right - w_offset)), output_w);
         ow_end = std::max(ow_begin, ow_end);
-        unsigned int ih = oh_begin - pad_top + h_offset;
+        int ih = oh_begin - pad_top + h_offset;
         for (int oh = oh_begin; oh < oh_end; ++oh, ++ih) {
-          unsigned int iw = ow_begin - pad_left + w_offset;
-          unsigned int ow = ow_begin;
+          int iw = ow_begin - pad_left + w_offset;
+          int ow = ow_begin;
           unsigned int data_im_offset = data_im_z + ih * width;
           unsigned int data_col_offset = data_col_z + oh * output_w;
           const float* data_im_ptr = data_im + data_im_offset;
@@ -941,27 +940,28 @@ void im2col_s2<float>(const float* data_im,
   memset(data_col, 0, mem_size);
 #pragma omp parallel for
   for (int c = 0; c < channels; c++) {
-    unsigned int data_im_z = c * in_channel_size;
-    unsigned int data_col_z1 = c * output_plane_size;
+    unsigned int data_im_z = static_cast<unsigned int>(c * in_channel_size);
+    int data_col_z1 = c * output_plane_size;
     for (int ky = 0, h_offset = 0; ky < kernel_h;
          ky++, h_offset += dilation_h) {
-      unsigned int data_col_z2 = ky * output_h * output_w * kernel_w;
+      int data_col_z2 = ky * output_h * output_w * kernel_w;
       for (int kx = 0, w_offset = 0; kx < kernel_w;
            kx++, w_offset += dilation_w) {
-        unsigned int data_col_z3 = kx * output_h * output_w;
-        unsigned int data_col_z = data_col_z1 + data_col_z2 + data_col_z3;
-        unsigned int oh_begin = std::max(((pad_top - h_offset + 1) / 2), 0);
-        unsigned int oh_end =
+        int data_col_z3 = kx * output_h * output_w;
+        unsigned int data_col_z =
+            static_cast<unsigned int>(data_col_z1 + data_col_z2 + data_col_z3);
+        int oh_begin = std::max(((pad_top - h_offset + 1) / 2), 0);
+        int oh_end =
             std::min(((height + pad_bottom - h_offset + 1) / 2), output_h);
         oh_end = std::max(oh_begin, oh_end);
-        unsigned int ow_begin = std::max(((pad_left - w_offset + 1) / 2), 0);
-        unsigned int ow_end =
+        int ow_begin = std::max(((pad_left - w_offset + 1) / 2), 0);
+        int ow_end =
             std::min(((width + pad_right - w_offset + 1) / 2), output_w);
         ow_end = std::max(ow_begin, ow_end);
-        unsigned int ih = oh_begin * 2 - pad_top + h_offset;
+        int ih = oh_begin * 2 - pad_top + h_offset;
         for (int oh = oh_begin; oh < oh_end; ++oh, ih += 2) {
-          unsigned int iw = ow_begin * 2 - pad_left + w_offset;
-          unsigned int ow = ow_begin;
+          int iw = ow_begin * 2 - pad_left + w_offset;
+          int ow = ow_begin;
           unsigned int data_im_offset = data_im_z + ih * width;
           unsigned int data_col_offset = data_col_z + oh * output_w;
           const float* data_im_ptr = data_im + data_im_offset;

lite/kernels/host/multiclass_nms_compute.h

@@ -385,12 +385,12 @@ class MulticlassNmsCompute : public KernelLite<TType, PType> {
 
     if (return_rois_num) {
       auto* nms_rois_num = param.nms_rois_num;
-      nms_rois_num->template mutable_data<int>();
+      nms_rois_num->Resize({n});
       int* num_data = nms_rois_num->template mutable_data<int>();
+
       for (int i = 1; i <= n; i++) {
         num_data[i - 1] = batch_starts[i] - batch_starts[i - 1];
       }
-      nms_rois_num->Resize({n});
     }
 
     LoD lod;

lite/operators/elementwise_ops.cc

@@ -73,7 +73,7 @@ bool ElementwiseOp::InferShapeImpl() const {
     }
     for (size_t i = 0; i < max_dim; i++) {
       if (x_dims_array[i] == -1 || y_dims_array[i] == -1) {
-        out_dims_array[i] = -1;
+        out_dims_array[i] = 1;
       } else {
         out_dims_array[i] = (std::max)(x_dims_array[i], y_dims_array[i]);
       }