[oneDNN] Fix adaptive Pooling (#58361)

zhanglirong1999 · web-flow · commit 11614e6dbfa3 · 2023-10-26T17:44:23.000+08:00
diff --git a/paddle/phi/backends/onednn/onednn_reuse.h b/paddle/phi/backends/onednn/onednn_reuse.h
@@ -1519,8 +1519,11 @@ class PoolingOneDNNHandler
     }
 
     if (adaptive) {
-      ComputeAdaptivePoolParameters(
-          src_tz, &copied_kernel_size, &copied_strides);
+      ComputeAdaptivePoolParameters(src_tz,
+                                    onednn_paddings[0],
+                                    onednn_paddings[1],
+                                    &copied_kernel_size,
+                                    &copied_strides);
     }
 
     bool is_test = dev_ctx.HasDnnAttr("is_test")
@@ -1612,8 +1615,11 @@ class PoolingOneDNNHandler
     }
 
     if (adaptive) {
-      ComputeAdaptivePoolParameters(
-          diff_src_tz, &copied_kernel_size, &copied_strides);
+      ComputeAdaptivePoolParameters(src_tz,
+                                    onednn_paddings[0],
+                                    onednn_paddings[1],
+                                    &copied_kernel_size,
+                                    &copied_strides);
     }
     memory::dims dilation = {0, 0};
 
@@ -1672,23 +1678,45 @@ class PoolingOneDNNHandler
     return mem_p;
   }
 
-  static void ComputeAdaptivePoolParameters(const std::vector<int64_t>& src_tz,
-                                            std::vector<int64_t>* kernel_size,
-                                            std::vector<int64_t>* strides) {
+  static void ComputeAdaptivePoolParameters(
+      const std::vector<int64_t>& src_tz,
+      const std::vector<int64_t>& padding_l,
+      const std::vector<int64_t>& padding_r,
+      std::vector<int64_t>* kernel_size,
+      std::vector<int64_t>* strides) {
     // https://github.com/oneapi-src/oneDNN/tree/bkocot/adaptive-pooling/rfcs/20200818-adaptive-pooling
     auto IH = static_cast<double>(src_tz[src_tz.size() - 2]);
     auto IW = static_cast<double>(src_tz[src_tz.size() - 1]);
     auto OH = static_cast<double>(kernel_size->at(0));
     auto OW = static_cast<double>(kernel_size->at(1));
 
-    strides->at(0) =
-        static_cast<int64_t>(floor((IH * 2.0) / OH) - floor(IH / OH));
-    strides->at(1) =
-        static_cast<int64_t>(floor((IW * 2.0) / OW) - floor(IW / OW));
-    kernel_size->at(0) =
-        static_cast<int64_t>(ceil((IH * 2.0) / OH) - floor(IH / OH));
-    kernel_size->at(1) =
-        static_cast<int64_t>(ceil((IW * 2.0) / OW) - floor(IW / OW));
+    /*
+    The previous calculation formula is given by OneDNN rfc, but in some odd
+    cases(mod(I/O)>=O/2) there will be problems with the calculation results.
+    Now change the formula to the general calculation formula of
+    AdaptivePool when in mod(I/O)>=O/2 case:
+    stride=floor(input_size/output_size)
+    kernel_size=input_size-(output_size-1)*stride
+    */
+    int mod_H = IH - floor(IH / OH) * OH;
+    int mod_W = IW - floor(IW / OW) * OW;
+    if (2 * mod_H < OH && 2 * mod_W < OW) {
+      strides->at(0) =
+          static_cast<int64_t>(floor((IH * 2.0) / OH) - floor(IH / OH));
+      strides->at(1) =
+          static_cast<int64_t>(floor((IW * 2.0) / OW) - floor(IW / OW));
+      kernel_size->at(0) =
+          static_cast<int64_t>(ceil((IH * 2.0) / OH) - floor(IH / OH));
+      kernel_size->at(1) =
+          static_cast<int64_t>(ceil((IW * 2.0) / OW) - floor(IW / OW));
+    } else {
+      strides->at(0) = static_cast<int64_t>(floor(IH / OH));
+      strides->at(1) = static_cast<int64_t>(floor(IW / OW));
+      kernel_size->at(0) = static_cast<int64_t>(
+          IH + padding_l[0] + padding_r[0] - floor((OH - 1) * strides->at(0)));
+      kernel_size->at(1) = static_cast<int64_t>(
+          IW + padding_l[1] + padding_r[1] - floor((OW - 1) * strides->at(1)));
+    }
   }
 
  private:
diff --git a/test/cpp/fluid/mkldnn/CMakeLists.txt b/test/cpp/fluid/mkldnn/CMakeLists.txt
@@ -83,3 +83,18 @@ else()
   cc_test_old(test_mkldnn_op_nhwc SRCS test_mkldnn_op_nhwc.cc DEPS
               ${paddle_lib} python)
 endif()
+
+cc_test(
+  test_mkldnn_pool_adaptive_op
+  SRCS test_mkldnn_pool_adaptive_op.cc
+  DEPS fleet_executor
+       conditional_block_op
+       standalone_executor
+       executor
+       op_registry
+       generated_static_op
+       generated_op
+       phi
+       scope
+       device_context
+       enforce)
diff --git a/test/cpp/fluid/mkldnn/test_mkldnn_pool_adaptive_op.cc b/test/cpp/fluid/mkldnn/test_mkldnn_pool_adaptive_op.cc
@@ -0,0 +1,91 @@
+/* Copyright (c) 2023 PaddlePaddle Authors. All Rights Reserved.
+
+   Licensed under the Apache License, Version 2.0 (the "License");
+   you may not use this file except in compliance with the License.
+   You may obtain a copy of the License at
+
+   https://www.apache.org/licenses/LICENSE-2.0
+
+   Unless required by applicable law or agreed to in writing, software
+   distributed under the License is distributed on an "AS IS" BASIS,
+   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+   See the License for the specific language governing permissions and
+   limitations under the License. */
+#include <gtest/gtest.h>
+
+#include <fstream>
+
+#include "paddle/fluid/framework/lod_tensor.h"
+#include "paddle/fluid/framework/naive_executor.h"
+#include "paddle/fluid/framework/op_registry.h"
+#include "paddle/fluid/framework/operator.h"
+#include "paddle/fluid/framework/scope.h"
+#include "paddle/phi/common/place.h"
+#include "paddle/phi/core/enforce.h"
+#include "paddle/phi/core/kernel_registry.h"
+
+namespace paddle {
+namespace inference {
+namespace tensorrt {
+
+template <typename DataType>
+void AddVarToScope(const std::string var_name,
+                   paddle::framework::Scope* scope,
+                   const paddle::framework::DDim& dims) {
+  std::random_device seed;
+  std::default_random_engine engine(seed());
+  std::uniform_real_distribution<float> dist(0, 100);
+
+  phi::DenseTensor tmp_tensor;
+  auto* tmp_data =
+      tmp_tensor.mutable_data<DataType>(dims, paddle::platform::CPUPlace());
+  auto* tensor = scope->Var(var_name)->GetMutable<phi::DenseTensor>();
+  tensor->mutable_data<DataType>(dims, paddle::platform::CPUPlace());
+  for (auto i = 0; i < tensor->numel(); ++i) {
+    tmp_data[i] = static_cast<DataType>(dist(engine));
+  }
+  paddle::framework::TensorCopySync(
+      tmp_tensor, paddle::platform::CPUPlace(), tensor);
+}
+void test_pool2d(bool adaptive, bool ceil_mode, std::string pool_type = "max") {
+  framework::Scope scope;
+  paddle::platform::CPUPlace cpu_place;
+
+  // Prepare Op description
+  framework::OpDesc desc;
+  desc.SetType("pool2d");
+  desc.SetInput("X", {"pool2d-X"});
+  desc.SetOutput("Out", {"pool2d-Out"});
+  AddVarToScope<float>("pool2d-X", &scope, {1, 3, 9, 12});
+  AddVarToScope<float>("pool2d-Out", &scope, {1, 3, 2, 2});
+  std::vector<int> ksize({2, 2});
+  std::vector<int> strides({1, 1});
+  std::vector<int> paddings({0, 0});
+  std::string pooling_t = pool_type;
+
+  desc.SetAttr("pooling_type", pooling_t);
+  desc.SetAttr("ksize", ksize);
+  desc.SetAttr("strides", strides);
+  desc.SetAttr("paddings", paddings);
+  desc.SetAttr("adaptive", adaptive);
+  desc.SetAttr("ceil_mode", ceil_mode);
+  desc.SetAttr("use_mkldnn", true);
+
+  auto op = paddle::framework::OpRegistry::CreateOp(desc);
+
+  op->Run(scope, cpu_place);
+}
+
+TEST(Pool2dOpConverter, normal) { test_pool2d(false, false); }
+TEST(Pool2dOpConverter, adaptive) { test_pool2d(true, false); }
+
+TEST(Pool2dOpConverter, max_ceil_test) { test_pool2d(false, true); }
+TEST(Pool2dOpConverter, avg_ceil_test) { test_pool2d(true, true, "avg"); }
+
+}  // namespace tensorrt
+}  // namespace inference
+}  // namespace paddle
+
+USE_OP_ITSELF(pool2d);
+PD_DECLARE_KERNEL(pool2d, OneDNN, ONEDNN);
+PD_DECLARE_KERNEL(pool2d, CPU, ALL_LAYOUT);
