[ROCM] fix the backward maxpool (#32030)

paddle/fluid/operators/pool_cudnn_op.cu.cc

@@ -20,6 +20,8 @@ limitations under the License. */
 #include "paddle/fluid/platform/cudnn_helper.h"
 #endif
 #ifdef PADDLE_WITH_HIP
+#include "paddle/fluid/framework/data_type.h"
+#include "paddle/fluid/framework/operator.h"
 #include "paddle/fluid/platform/miopen_helper.h"
 #endif
 
@@ -264,6 +266,34 @@ class PoolCUDNNGradOpKernel : public framework::OpKernel<T> {
     std::string padding_algorithm = ctx.Attr<std::string>("padding_algorithm");
     const bool channel_last = (data_format == "NHWC" || data_format == "NDHWC");
 
+#ifdef PADDLE_WITH_HIP
+    if (pooling_type == "max") {
+      using OpKernelMap = paddle::framework::OperatorWithKernel::OpKernelMap;
+      using OpKernelFunc = paddle::framework::OperatorWithKernel::OpKernelFunc;
+      auto &all_op_kernels =
+          paddle::framework::OperatorWithKernel::AllOpKernels();
+      std::string op_type = "pool2d_grad";
+      auto kernels_iter = all_op_kernels.find(op_type);
+      PADDLE_ENFORCE_NE(
+          kernels_iter, all_op_kernels.end(),
+          platform::errors::Unavailable(
+              "There are no kernels which are registered in the %s operator.",
+              op_type));
+      OpKernelMap &kernels = kernels_iter->second;
+      paddle::framework::OpKernelType expected_kernel_key(
+          paddle::framework::ToDataType(typeid(T)), ctx.GetPlace());
+      auto kernel_iter = kernels.find(expected_kernel_key);
+      PADDLE_ENFORCE_NE(kernel_iter, kernels.end(),
+                        platform::errors::NotFound(
+                            "Operator (%s) does not have kernel for %s.",
+                            op_type, KernelTypeToString(expected_kernel_key)));
+      std::unique_ptr<OpKernelFunc> kernel_func_(
+          new OpKernelFunc(kernel_iter->second));
+      (*kernel_func_)(ctx);
+      return;
+    }
+#endif
+
     // update paddings
     auto in_x_dims = input->dims();
     framework::DDim data_dims;

python/paddle/fluid/dygraph/nn.py

@@ -174,6 +174,11 @@ class Conv2D(layers.Layer):
                  dtype='float32'):
         assert param_attr is not False, "param_attr should not be False here."
         super(Conv2D, self).__init__()
+
+        if (core.is_compiled_with_cuda() and paddle.fluid.get_flags(
+                "FLAGS_conv2d_disable_cudnn")["FLAGS_conv2d_disable_cudnn"]):
+            use_cudnn = False
+
         self._num_channels = num_channels
         self._groups = groups
         self._stride = utils.convert_to_list(stride, 2, 'stride')

python/paddle/fluid/tests/unittests/test_conv2d_op.py

@@ -1470,13 +1470,14 @@ class TestConv2DAPI_Error(unittest.TestCase):
     not (core.is_compiled_with_cuda() or core.is_compiled_with_rocm()),
     "core is not compiled with CUDA or ROCM")
 class TestConv2DEnviron(unittest.TestCase):
-    def run_conv2d_api(self):
+    def run1(self, place):
+        with fluid.program_guard(fluid.Program(), fluid.Program()):
             inputs = fluid.layers.data(
                 shape=[2, 3, 5, 5],
                 append_batch_size=False,
                 name="inputs",
                 dtype="float32")
-        fluid.layers.conv2d(
+            result = fluid.layers.conv2d(
                 input=inputs,
                 num_filters=4,
                 filter_size=[3, 3],
@@ -1485,20 +1486,43 @@ class TestConv2DEnviron(unittest.TestCase):
                 dilation=[1, 1],
                 groups=1,
                 data_format="NCHW")
-
-        x_var = paddle.uniform((2, 3, 5, 5), dtype="float32", min=-1., max=1.)
+            exe = fluid.Executor(place)
+            exe.run(fluid.default_startup_program())
+            fetches = exe.run(fluid.default_main_program(),
+                              feed={"inputs": self.input_np},
+                              fetch_list=[result])
+
+    def run2(self, place):
+        with fluid.dygraph.guard(place):
+            inputs = fluid.dygraph.to_variable(self.input_np)
             conv = paddle.nn.Conv2D(
                 in_channels=3,
                 out_channels=4,
                 kernel_size=(3, 3),
                 data_format="NCHW")
-        y_var = conv(x_var)
+            result = conv(inputs)
+
+    def run3(self, place):
+        with fluid.dygraph.guard(place):
+            inputs = fluid.dygraph.to_variable(self.input_np)
+            conv = paddle.fluid.dygraph.nn.Conv2D(
+                num_channels=3,
+                num_filters=4,
+                filter_size=(3, 3), )
+            result = conv(inputs)
+
+    def run_all(self, place):
+        self.run1(place)
+        self.run2(place)
+        self.run3(place)
 
     def test_environ(self):
+        self.input_np = np.random.random([2, 3, 5, 5]).astype("float32")
+        for place in [paddle.CPUPlace(), paddle.CUDAPlace(0)]:
             fluid.set_flags({'FLAGS_conv2d_disable_cudnn': False})
-        self.run_conv2d_api()
+            self.run_all(place)
             fluid.set_flags({'FLAGS_conv2d_disable_cudnn': True})
-        self.run_conv2d_api()
+            self.run_all(place)
 
 
 if __name__ == '__main__':