diff --git a/paddle/fluid/operators/mkldnn/pool_mkldnn_op.cc b/paddle/fluid/operators/mkldnn/pool_mkldnn_op.cc
index 83e9cfd90a8515c8fd15842c114e6b2c59f45d18..9c0893456abf8bbf56fc52c7444034d12cbe197b 100644
--- a/paddle/fluid/operators/mkldnn/pool_mkldnn_op.cc
+++ b/paddle/fluid/operators/mkldnn/pool_mkldnn_op.cc
@@ -74,7 +74,7 @@ class PoolMKLDNNOpKernel : public paddle::framework::OpKernel<T> {
         src_tz, dst_tz, ksize, strides, paddings, pooling_type,
         ctx.Attr<bool>("ceil_mode"), input->format(),
         paddle::framework::ToMKLDNNDataType(input->type()), is_test, dev_ctx,
-        ctx.GetPlace(), ctx.op().Output("Out"));
+        ctx.GetPlace(), ctx.op().Output("Out"), ctx.Attr<bool>("exclusive"));
 
     auto src_memory = handler.AcquireSrcMemory(input);
     auto dst_memory = handler.AcquireDstMemory(output);
@@ -158,7 +158,7 @@ class PoolMKLDNNGradOpKernel : public paddle::framework::OpKernel<T> {
         diff_dst_tz, diff_src_tz, ksize, strides, paddings, pooling_type,
         ctx.Attr<bool>("ceil_mode"), in_x->format(), out_grad->format(),
         paddle::framework::ToMKLDNNDataType(out_grad->type()), dev_ctx,
-        ctx.GetPlace(), ctx.op().Input("Out"));
+        ctx.GetPlace(), ctx.op().Input("Out"), ctx.Attr<bool>("exclusive"));
 
     auto diff_dst_memory = handler.AcquireDiffDstMemory(out_grad);
     auto diff_src_memory = handler.AcquireDiffSrcMemory(in_x_grad);
diff --git a/paddle/fluid/platform/mkldnn_reuse.h b/paddle/fluid/platform/mkldnn_reuse.h
index 7396b90ea3d0728a6c63069e8cb3089cc3c47f98..50c629e53ef280975346fac9a3912c0b6ea6c924 100644
--- a/paddle/fluid/platform/mkldnn_reuse.h
+++ b/paddle/fluid/platform/mkldnn_reuse.h
@@ -529,7 +529,7 @@ class PoolingMKLDNNHandler : public MKLDNNHandlerT<T, mkldnn::pooling_forward,
       bool ceil_mode, const MKLDNNMemoryFormat fmt,
       mkldnn::memory::data_type dt, bool is_test,
       const platform::MKLDNNDeviceContext& dev_ctx, platform::Place cpu_place,
-      const std::string& unique_name)
+      const std::string& unique_name, bool exclude_padding)
       : platform::MKLDNNHandlerT<T, mkldnn::pooling_forward,
                                  mkldnn::pooling_backward>(
             dev_ctx, dev_ctx.GetEngine(), cpu_place,
@@ -553,8 +553,11 @@ class PoolingMKLDNNHandler : public MKLDNNHandlerT<T, mkldnn::pooling_forward,
     this->AcquireForwardPrimitiveDescriptor(
         is_test ? mkldnn::prop_kind::forward_inference
                 : mkldnn::prop_kind::forward_training,
-        pooling_type == "max" ? mkldnn::algorithm::pooling_max
-                              : mkldnn::algorithm::pooling_avg,
+        pooling_type == "max"
+            ? mkldnn::algorithm::pooling_max
+            : (exclude_padding
+                   ? mkldnn::algorithm::pooling_avg_exclude_padding
+                   : mkldnn::algorithm::pooling_avg_include_padding),
         src_md, dst_md, strides, ksize, padding_left_top, padding_right_bottom,
         mkldnn::padding_kind::zero);
   }
@@ -567,7 +570,7 @@ class PoolingMKLDNNHandler : public MKLDNNHandlerT<T, mkldnn::pooling_forward,
       const MKLDNNMemoryFormat fmt, const MKLDNNMemoryFormat diff_dst_fmt,
       mkldnn::memory::data_type dt,
       const platform::MKLDNNDeviceContext& dev_ctx, platform::Place cpu_place,
-      const std::string& unique_name)
+      const std::string& unique_name, bool exclude_padding)
       : platform::MKLDNNHandlerT<T, mkldnn::pooling_forward,
                                  mkldnn::pooling_backward>(
             dev_ctx, dev_ctx.GetEngine(), cpu_place,
@@ -580,8 +583,11 @@ class PoolingMKLDNNHandler : public MKLDNNHandlerT<T, mkldnn::pooling_forward,
                              MKLDNNMemoryFormat::any);
 
     this->AcquireBackwardPrimitiveDescriptor(
-        pooling_type == "max" ? mkldnn::algorithm::pooling_max
-                              : mkldnn::algorithm::pooling_avg,
+        pooling_type == "max"
+            ? mkldnn::algorithm::pooling_max
+            : (exclude_padding
+                   ? mkldnn::algorithm::pooling_avg_exclude_padding
+                   : mkldnn::algorithm::pooling_avg_include_padding),
         diff_src_md, diff_dst_md, strides, ksize, paddings, paddings,
         mkldnn::padding_kind::zero);
   }
diff --git a/python/paddle/fluid/tests/unittests/mkldnn/test_pool2d_int8_mkldnn_op.py b/python/paddle/fluid/tests/unittests/mkldnn/test_pool2d_int8_mkldnn_op.py
index fca906fecc5fe8d25b9251c886398f8df778043f..b65cdcf3d3489dbd5c6c4c9a22c3245a662904f8 100644
--- a/python/paddle/fluid/tests/unittests/mkldnn/test_pool2d_int8_mkldnn_op.py
+++ b/python/paddle/fluid/tests/unittests/mkldnn/test_pool2d_int8_mkldnn_op.py
@@ -34,6 +34,13 @@ class TestPool2dMKLDNNInt8_Op(TestPool2D_Op):
         TestPool2D_Op.setUp(self)
         assert self.dtype in [np.int8, np.uint8
                               ], 'Dtype should be int8 or uint8'
+        input = np.random.randint(0, 100, self.shape).astype(self.dtype)
+        output = (self.pool2D_forward_naive(
+            input, self.ksize, self.strides, self.paddings, self.global_pool,
+            self.ceil_mode, self.exclusive, self.adaptive,
+            self.dtype)).astype(self.dtype)
+        self.inputs = {'X': OpTest.np_dtype_to_fluid_dtype(input)}
+        self.outputs = {'Out': output}
 
     def test_check_output(self):
         self.check_output_with_place(core.CPUPlace(), atol=1e-5)
@@ -52,6 +59,9 @@ class TestCase1Avg(TestPool2dMKLDNNInt8_Op):
     def init_global_pool(self):
         self.global_pool = False
 
+    def init_exclusive(self):
+        self.exclusive = True
+
 
 class TestCase2Avg(TestPool2dMKLDNNInt8_Op):
     def init_test_case(self):
@@ -63,6 +73,9 @@ class TestCase2Avg(TestPool2dMKLDNNInt8_Op):
     def init_global_pool(self):
         self.global_pool = False
 
+    def init_exclusive(self):
+        self.exclusive = False
+
 
 class TestCase0Max(TestPool2dMKLDNNInt8_Op):
     def init_pool_type(self):
diff --git a/python/paddle/fluid/tests/unittests/test_pool2d_op.py b/python/paddle/fluid/tests/unittests/test_pool2d_op.py
index 84f6526b8052d77a32130487e1bc80c6439db7b7..9d501b709425e9f993503ea3eb5ed7488973ca7f 100644
--- a/python/paddle/fluid/tests/unittests/test_pool2d_op.py
+++ b/python/paddle/fluid/tests/unittests/test_pool2d_op.py
@@ -37,7 +37,8 @@ def max_pool2D_forward_naive(x,
                              global_pool=0,
                              ceil_mode=False,
                              exclusive=True,
-                             adaptive=False):
+                             adaptive=False,
+                             data_type=np.float32):
     N, C, H, W = x.shape
     if global_pool == 1:
         ksize = [H, W]
@@ -76,7 +77,8 @@ def avg_pool2D_forward_naive(x,
                              global_pool=0,
                              ceil_mode=False,
                              exclusive=True,
-                             adaptive=False):
+                             adaptive=False,
+                             data_type=np.float32):
     N, C, H, W = x.shape
     if global_pool == 1:
         ksize = [H, W]
@@ -106,7 +108,13 @@ def avg_pool2D_forward_naive(x,
 
             field_size = ((r_end - r_start) * (c_end - c_start)) \
                         if (exclusive or adaptive) else (ksize[0] * ksize[1])
-            out[:, :, i, j] = np.sum(x_masked, axis=(2, 3)) / field_size
+            if data_type == np.int8 or data_type == np.uint8:
+                out[:, :, i, j] = (np.rint(
+                    np.sum(x_masked, axis=(2, 3)) /
+                    field_size)).astype(data_type)
+            else:
+                out[:, :, i, j] = (np.sum(x_masked, axis=(2, 3)) /
+                                   field_size).astype(data_type)
     return out
 
 
@@ -126,9 +134,10 @@ class TestPool2D_Op(OpTest):
         if self.global_pool:
             self.paddings = [0 for _ in range(len(self.paddings))]
         input = np.random.random(self.shape).astype(self.dtype)
-        output = self.pool2D_forward_naive(
+        output = (self.pool2D_forward_naive(
             input, self.ksize, self.strides, self.paddings, self.global_pool,
-            self.ceil_mode, self.exclusive, self.adaptive).astype(self.dtype)
+            self.ceil_mode, self.exclusive, self.adaptive,
+            self.dtype)).astype(self.dtype)
         self.inputs = {'X': OpTest.np_dtype_to_fluid_dtype(input)}
 
         self.attrs = {