Merge branch 'PaddlePaddle:develop' into ccl

Author: yongqiangma

backends/gcu/ci_test.sh

@@ -32,4 +32,4 @@ make -j $(nproc)
 
 python -m pip install --force-reinstall -U dist/paddle_custom_gcu*.whl
 
-ctest -j4
+ctest -j4 --output-on-failure

backends/gcu/custom_engine/ir_translator/operators/pool2d.cc

@@ -48,15 +48,15 @@ static GcuOpPtr TranslatePool2d(
   std::vector<int64_t> strides;
   for (size_t i = 0; i < strides_array.size(); ++i) {
     strides.push_back(
-        strides_array.at(i).dyn_cast<pir::Int32Attribute>().data());
+        strides_array.at(i).dyn_cast<pir::Int64Attribute>().data());
   }
 
   auto paddings_array =
       attributes.at("paddings").dyn_cast<pir::ArrayAttribute>();
   std::vector<int64_t> paddings;
   for (size_t i = 0; i < paddings_array.size(); ++i) {
     paddings.push_back(
-        paddings_array.at(i).dyn_cast<pir::Int32Attribute>().data());
+        paddings_array.at(i).dyn_cast<pir::Int64Attribute>().data());
   }
 
   bool ceil_mode =

backends/gcu/kernels/pool2d_kernel.cc

@@ -24,7 +24,7 @@ inline void UpdatePadding(std::vector<T>* paddings,
                           const std::string padding_algorithm,
                           const phi::DDim data_dims,
                           const std::vector<T>& strides,
-                          const std::vector<T>& kernel_size) {
+                          const std::vector<int>& kernel_size) {
   // set padding size == data_dims.size() * 2
   auto data_shape = phi::vectorize<T>(data_dims);
   if (static_cast<int>(paddings->size()) == data_dims.size()) {
@@ -72,8 +72,8 @@ template <typename T, typename Context>
 void Pool2dKernel(const Context& dev_ctx,
                   const phi::DenseTensor& in_x,
                   const phi::IntArray& kernel_size,
-                  const std::vector<int>& strides_t,
-                  const std::vector<int>& paddings_t,
+                  const std::vector<int64_t>& strides_t,
+                  const std::vector<int64_t>& paddings_t,
                   bool ceil_mode,
                   bool exclusive,
                   const std::string& data_format,
@@ -100,8 +100,8 @@ void Pool2dKernel(const Context& dev_ctx,
           << ", in_x dims:"
           << custom_kernel::VectorToStr<int64_t>(phi::vectorize(in_x_dims))
           << ", kernel_size:" << custom_kernel::VectorToStr<int>(ksize)
-          << ", strides:" << custom_kernel::VectorToStr<int>(strides)
-          << ", paddings:" << custom_kernel::VectorToStr<int>(paddings)
+          << ", strides:" << custom_kernel::VectorToStr<int64_t>(strides)
+          << ", paddings:" << custom_kernel::VectorToStr<int64_t>(paddings)
           << ", ceil_mode:" << ceil_mode << ", exclusive:" << exclusive
           << ", data_format:" << data_format
           << ", padding_algorithm:" << padding_algorithm
@@ -287,8 +287,8 @@ void Pool2dGradKernel(const Context& dev_ctx,
                       const phi::DenseTensor& out,
                       const phi::DenseTensor& out_grad,
                       const phi::IntArray& kernel_size,
-                      const std::vector<int>& strides_t,
-                      const std::vector<int>& paddings_t,
+                      const std::vector<int64_t>& strides_t,
+                      const std::vector<int64_t>& paddings_t,
                       bool ceil_mode,
                       bool exclusive,
                       const std::string& data_format,