Merge branch 'develop' into add_clip_op

4eb44380 · jerrywgz · GitHub · e402c0ec · 30cc8b7a · 4eb44380
99 changed file
--- a/cmake/operators.cmake
+++ b/cmake/operators.cmake
@@ -52,8 +52,8 @@ function(op_library TARGET)
        endif()
        if(WITH_MKLDNN)
            string(REPLACE "_op" "_mkldnn_op" MKLDNN_FILE "${TARGET}")
-            if (EXISTS ${CMAKE_CURRENT_SOURCE_DIR}/${MKLDNN_FILE}.cc)
+            if (EXISTS ${CMAKE_CURRENT_SOURCE_DIR}/mkldnn/${MKLDNN_FILE}.cc)
-                list(APPEND mkldnn_cc_srcs ${MKLDNN_FILE}.cc)
+                list(APPEND mkldnn_cc_srcs mkldnn/${MKLDNN_FILE}.cc)
            endif()
        endif()
    else()

--- a/paddle/fluid/API.spec
+++ b/paddle/fluid/API.spec
@@ -122,7 +122,7 @@ paddle.fluid.layers.transpose ArgSpec(args=['x', 'perm', 'name'], varargs=None,
 paddle.fluid.layers.im2sequence ArgSpec(args=['input', 'filter_size', 'stride', 'padding', 'input_image_size', 'out_stride', 'name'], varargs=None, keywords=None, defaults=(1, 1, 0, None, 1, None))
 paddle.fluid.layers.nce ArgSpec(args=['input', 'label', 'num_total_classes', 'sample_weight', 'param_attr', 'bias_attr', 'num_neg_samples', 'name', 'sampler', 'custom_dist', 'seed', 'is_sparse'], varargs=None, keywords=None, defaults=(None, None, None, None, None, 'uniform', None, 0, False))
 paddle.fluid.layers.hsigmoid ArgSpec(args=['input', 'label', 'num_classes', 'param_attr', 'bias_attr', 'name', 'path_table', 'path_code', 'is_custom', 'is_sparse'], varargs=None, keywords=None, defaults=(None, None, None, None, None, False, False))
-paddle.fluid.layers.beam_search ArgSpec(args=['pre_ids', 'pre_scores', 'ids', 'scores', 'beam_size', 'end_id', 'level', 'is_accumulated', 'name'], varargs=None, keywords=None, defaults=(0, True, None))
+paddle.fluid.layers.beam_search ArgSpec(args=['pre_ids', 'pre_scores', 'ids', 'scores', 'beam_size', 'end_id', 'level', 'is_accumulated', 'name', 'return_parent_idx'], varargs=None, keywords=None, defaults=(0, True, None, False))
 paddle.fluid.layers.row_conv ArgSpec(args=['input', 'future_context_size', 'param_attr', 'act'], varargs=None, keywords=None, defaults=(None, None))
 paddle.fluid.layers.multiplex ArgSpec(args=['inputs', 'index'], varargs=None, keywords=None, defaults=None)
 paddle.fluid.layers.layer_norm ArgSpec(args=['input', 'scale', 'shift', 'begin_norm_axis', 'epsilon', 'param_attr', 'bias_attr', 'act', 'name'], varargs=None, keywords=None, defaults=(True, True, 1, 1e-05, None, None, None, None))
@@ -322,10 +322,10 @@ paddle.fluid.layers.generate_proposal_labels ArgSpec(args=['rpn_rois', 'gt_class
 paddle.fluid.layers.generate_proposals ArgSpec(args=['scores', 'bbox_deltas', 'im_info', 'anchors', 'variances', 'pre_nms_top_n', 'post_nms_top_n', 'nms_thresh', 'min_size', 'eta', 'name'], varargs=None, keywords=None, defaults=(6000, 1000, 0.5, 0.1, 1.0, None))
 paddle.fluid.layers.generate_mask_labels ArgSpec(args=['im_info', 'gt_classes', 'is_crowd', 'gt_segms', 'rois', 'labels_int32', 'num_classes', 'resolution'], varargs=None, keywords=None, defaults=None)
 paddle.fluid.layers.iou_similarity ArgSpec(args=['x', 'y', 'name'], varargs=None, keywords=None, defaults=(None,))
-paddle.fluid.layers.box_coder ArgSpec(args=['prior_box', 'prior_box_var', 'target_box', 'code_type', 'box_normalized', 'name'], varargs=None, keywords=None, defaults=('encode_center_size', True, None))
+paddle.fluid.layers.box_coder ArgSpec(args=['prior_box', 'prior_box_var', 'target_box', 'code_type', 'box_normalized', 'name', 'axis'], varargs=None, keywords=None, defaults=('encode_center_size', True, None, 0))
 paddle.fluid.layers.polygon_box_transform ArgSpec(args=['input', 'name'], varargs=None, keywords=None, defaults=(None,))
-paddle.fluid.layers.yolov3_loss ArgSpec(args=['x', 'gtbox', 'gtlabel', 'anchors', 'class_num', 'ignore_thresh', 'loss_weight_xy', 'loss_weight_wh', 'loss_weight_conf_target', 'loss_weight_conf_notarget', 'loss_weight_class', 'name'], varargs=None, keywords=None, defaults=(None, None, None, None, None, None))
 paddle.fluid.layers.box_clip ArgSpec(args=['input', 'im_info', 'name'], varargs=None, keywords=None, defaults=(None,))
+paddle.fluid.layers.yolov3_loss ArgSpec(args=['x', 'gtbox', 'gtlabel', 'anchors', 'anchor_mask', 'class_num', 'ignore_thresh', 'downsample_ratio', 'name'], varargs=None, keywords=None, defaults=(None,))
 paddle.fluid.layers.multiclass_nms ArgSpec(args=['bboxes', 'scores', 'score_threshold', 'nms_top_k', 'keep_top_k', 'nms_threshold', 'normalized', 'nms_eta', 'background_label', 'name'], varargs=None, keywords=None, defaults=(0.3, True, 1.0, 0, None))
 paddle.fluid.layers.accuracy ArgSpec(args=['input', 'label', 'k', 'correct', 'total'], varargs=None, keywords=None, defaults=(1, None, None))
 paddle.fluid.layers.auc ArgSpec(args=['input', 'label', 'curve', 'num_thresholds', 'topk', 'slide_steps'], varargs=None, keywords=None, defaults=('ROC', 4095, 1, 1))
@@ -362,6 +362,9 @@ paddle.fluid.contrib.QuantizeTranspiler.__init__ ArgSpec(args=['self', 'weight_b
 paddle.fluid.contrib.QuantizeTranspiler.convert_to_int8 ArgSpec(args=['self', 'program', 'place', 'scope'], varargs=None, keywords=None, defaults=(None,))
 paddle.fluid.contrib.QuantizeTranspiler.freeze_program ArgSpec(args=['self', 'program', 'place', 'fuse_bn', 'scope'], varargs=None, keywords=None, defaults=(False, None))
 paddle.fluid.contrib.QuantizeTranspiler.training_transpile ArgSpec(args=['self', 'program', 'startup_program'], varargs=None, keywords=None, defaults=(None, None))
+paddle.fluid.contrib.Calibrator.__init__ ArgSpec(args=['self'], varargs='args', keywords='kwargs', defaults=None)
+paddle.fluid.contrib.Calibrator.sample_data ArgSpec(args=['self'], varargs=None, keywords=None, defaults=None)
+paddle.fluid.contrib.Calibrator.save_int8_model ArgSpec(args=['self'], varargs=None, keywords=None, defaults=None)
 paddle.fluid.contrib.reader.ctr_reader.ctr_reader ArgSpec(args=['feed_dict', 'file_type', 'file_format', 'dense_slot_index', 'sparse_slot_index', 'capacity', 'thread_num', 'batch_size', 'file_list', 'slots', 'name'], varargs=None, keywords=None, defaults=(None,))
 paddle.fluid.contrib.build_compressor ArgSpec(args=['place', 'data_reader', 'data_feeder', 'scope', 'metrics', 'epoch', 'config'], varargs=None, keywords=None, defaults=(None, None, None, None, None, None, None))
 paddle.fluid.contrib.CompressPass.__init__ ArgSpec(args=['self', 'place', 'data_reader', 'data_feeder', 'scope', 'metrics', 'epoch', 'program_exe'], varargs=None, keywords=None, defaults=(None, None, None, None, None, None, None))

--- a/paddle/fluid/framework/ir/CMakeLists.txt
+++ b/paddle/fluid/framework/ir/CMakeLists.txt
@@ -10,8 +10,22 @@ function(pass_library TARGET DEST)
    set(options "")
    set(oneValueArgs "")
    set(multiValueArgs SRCS DEPS)
+    set(targetPrefix "")
+    # Get optional argument
+    set(extraMacroArgs ${ARGN})
+    list(LENGTH extraMacroArgs numExtraMacroArgs)
+    if(numExtraMacroArgs GREATER 0)
+        list(GET extraMacroArgs 0 targetPrefix)
+    endif()
    cmake_parse_arguments(op_library "${options}" "${oneValueArgs}" "${multiValueArgs}" ${ARGN})
-    cc_library(${TARGET} SRCS ${TARGET}.cc DEPS graph_pattern_detector pass fuse_pass_base ${op_library_DEPS})
+    if(targetPrefix)
+        cc_library(${TARGET} SRCS ${targetPrefix}/${TARGET}.cc DEPS graph_pattern_detector pass fuse_pass_base ${op_library_DEPS})
+    else()
+        cc_library(${TARGET} SRCS ${TARGET}.cc DEPS graph_pattern_detector pass fuse_pass_base ${op_library_DEPS})
+    endif()
    # add more DEST here, such as train, dist and collect USE_PASS into a file automatically.
    if (${DEST} STREQUAL "base" OR ${DEST} STREQUAL "inference")
        message(STATUS "add pass ${TARGET} ${DEST}")
@@ -62,11 +76,11 @@ foreach (index RANGE 3 6)
 endforeach()
 if(WITH_MKLDNN)
-    pass_library(mkldnn_placement_pass base)
+    pass_library(mkldnn_placement_pass base mkldnn)
-    pass_library(depthwise_conv_mkldnn_pass base)
+    pass_library(depthwise_conv_mkldnn_pass base mkldnn)
-    pass_library(conv_bias_mkldnn_fuse_pass inference)
+    pass_library(conv_bias_mkldnn_fuse_pass inference mkldnn)
-    pass_library(conv_relu_mkldnn_fuse_pass inference)
+    pass_library(conv_relu_mkldnn_fuse_pass inference mkldnn)
-    pass_library(conv_elementwise_add_mkldnn_fuse_pass inference)
+    pass_library(conv_elementwise_add_mkldnn_fuse_pass inference mkldnn)
 endif()
 cc_library(fuse_elewise_add_act_pass SRCS fuse_elewise_add_act_pass.cc DEPS pass graph_pattern_detector )
@@ -86,7 +100,7 @@ cc_test(test_fc_fuse_pass SRCS fc_fuse_pass_tester.cc DEPS fc_fuse_pass framewor
 cc_test(test_seqpool_concat_fuse_pass SRCS seqpool_concat_fuse_pass_tester.cc DEPS seqpool_concat_fuse_pass framework_proto)
 cc_test(test_is_test_pass SRCS is_test_pass_tester.cc DEPS is_test_pass)
 if (WITH_MKLDNN)
-    cc_test(test_depthwise_conv_mkldnn_pass SRCS depthwise_conv_mkldnn_pass_tester.cc DEPS depthwise_conv_mkldnn_pass)
+    cc_test(test_depthwise_conv_mkldnn_pass SRCS mkldnn/depthwise_conv_mkldnn_pass_tester.cc DEPS depthwise_conv_mkldnn_pass)
-    cc_test(test_conv_relu_mkldnn_fuse_pass SRCS conv_relu_mkldnn_fuse_pass_tester.cc DEPS conv_relu_mkldnn_fuse_pass)
+    cc_test(test_conv_relu_mkldnn_fuse_pass SRCS mkldnn/conv_relu_mkldnn_fuse_pass_tester.cc DEPS conv_relu_mkldnn_fuse_pass)
-    cc_test(test_conv_elementwise_add_mkldnn_fuse_pass SRCS conv_elementwise_add_mkldnn_fuse_pass_tester.cc DEPS conv_elementwise_add_mkldnn_fuse_pass)
+    cc_test(test_conv_elementwise_add_mkldnn_fuse_pass SRCS mkldnn/conv_elementwise_add_mkldnn_fuse_pass_tester.cc DEPS conv_elementwise_add_mkldnn_fuse_pass)
 endif ()
--- a/paddle/fluid/framework/ir/conv_bias_mkldnn_fuse_pass.cc
+++ b/paddle/fluid/framework/ir/conv_bias_mkldnn_fuse_pass.cc
@@ -12,7 +12,7 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
-#include "paddle/fluid/framework/ir/conv_bias_mkldnn_fuse_pass.h"
+#include "paddle/fluid/framework/ir/mkldnn/conv_bias_mkldnn_fuse_pass.h"
 #include <functional>
 #include <string>
 #include <vector>

--- a/paddle/fluid/framework/ir/conv_bias_mkldnn_fuse_pass.h
+++ b/paddle/fluid/framework/ir/conv_bias_mkldnn_fuse_pass.h
--- a/paddle/fluid/framework/ir/conv_elementwise_add_mkldnn_fuse_pass.cc
+++ b/paddle/fluid/framework/ir/conv_elementwise_add_mkldnn_fuse_pass.cc
@@ -12,7 +12,7 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
-#include "paddle/fluid/framework/ir/conv_elementwise_add_mkldnn_fuse_pass.h"
+#include "paddle/fluid/framework/ir/mkldnn/conv_elementwise_add_mkldnn_fuse_pass.h"
 #include <functional>
 #include <list>
 #include <map>

--- a/paddle/fluid/framework/ir/conv_elementwise_add_mkldnn_fuse_pass.h
+++ b/paddle/fluid/framework/ir/conv_elementwise_add_mkldnn_fuse_pass.h
--- a/paddle/fluid/framework/ir/conv_elementwise_add_mkldnn_fuse_pass_tester.cc
+++ b/paddle/fluid/framework/ir/conv_elementwise_add_mkldnn_fuse_pass_tester.cc
@@ -15,8 +15,8 @@
 #include <gtest/gtest.h>
 #include <string>
-#include "paddle/fluid/framework/ir/conv_elementwise_add_mkldnn_fuse_pass.h"
 #include "paddle/fluid/framework/ir/graph_traits.h"
+#include "paddle/fluid/framework/ir/mkldnn/conv_elementwise_add_mkldnn_fuse_pass.h"
 namespace paddle {
 namespace framework {

--- a/paddle/fluid/framework/ir/conv_relu_mkldnn_fuse_pass.cc
+++ b/paddle/fluid/framework/ir/conv_relu_mkldnn_fuse_pass.cc
@@ -12,7 +12,7 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
-#include "paddle/fluid/framework/ir/conv_relu_mkldnn_fuse_pass.h"
+#include "paddle/fluid/framework/ir/mkldnn/conv_relu_mkldnn_fuse_pass.h"
 #include <string>
 #include <vector>
 #include "paddle/fluid/platform/enforce.h"

--- a/paddle/fluid/framework/ir/conv_relu_mkldnn_fuse_pass.h
+++ b/paddle/fluid/framework/ir/conv_relu_mkldnn_fuse_pass.h
--- a/paddle/fluid/framework/ir/conv_relu_mkldnn_fuse_pass_tester.cc
+++ b/paddle/fluid/framework/ir/conv_relu_mkldnn_fuse_pass_tester.cc
@@ -12,7 +12,7 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
-#include "paddle/fluid/framework/ir/conv_relu_mkldnn_fuse_pass.h"
+#include "paddle/fluid/framework/ir/mkldnn/conv_relu_mkldnn_fuse_pass.h"
 #include <gtest/gtest.h>
 #include "paddle/fluid/framework/op_proto_maker.h"

--- a/paddle/fluid/framework/ir/depthwise_conv_mkldnn_pass.cc
+++ b/paddle/fluid/framework/ir/depthwise_conv_mkldnn_pass.cc
@@ -12,7 +12,7 @@ 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 "paddle/fluid/framework/ir/depthwise_conv_mkldnn_pass.h"
+#include "paddle/fluid/framework/ir/mkldnn/depthwise_conv_mkldnn_pass.h"
 #include "paddle/fluid/framework/ir/graph_pattern_detector.h"
 namespace paddle {

--- a/paddle/fluid/framework/ir/depthwise_conv_mkldnn_pass.h
+++ b/paddle/fluid/framework/ir/depthwise_conv_mkldnn_pass.h
--- a/paddle/fluid/framework/ir/depthwise_conv_mkldnn_pass_tester.cc
+++ b/paddle/fluid/framework/ir/depthwise_conv_mkldnn_pass_tester.cc
@@ -12,7 +12,7 @@
 // See the License for the specific language governing permissions and
 // limitations under the License.
-#include "paddle/fluid/framework/ir/depthwise_conv_mkldnn_pass.h"
+#include "paddle/fluid/framework/ir/mkldnn/depthwise_conv_mkldnn_pass.h"
 #include <gtest/gtest.h>

--- a/paddle/fluid/framework/ir/mkldnn_placement_pass.cc
+++ b/paddle/fluid/framework/ir/mkldnn_placement_pass.cc
@@ -12,7 +12,7 @@ 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 "paddle/fluid/framework/ir/mkldnn_placement_pass.h"
+#include "paddle/fluid/framework/ir/mkldnn/mkldnn_placement_pass.h"
 #include <string>
 namespace paddle {

--- a/paddle/fluid/framework/ir/mkldnn_placement_pass.h
+++ b/paddle/fluid/framework/ir/mkldnn_placement_pass.h
--- a/paddle/fluid/inference/tests/api/CMakeLists.txt
+++ b/paddle/fluid/inference/tests/api/CMakeLists.txt
@@ -128,9 +128,9 @@ inference_analysis_api_test_with_fake_data(test_analyzer_resnet50
 inference_analysis_api_test_with_fake_data(test_analyzer_mobilenet_depthwise_conv
  "${INFERENCE_DEMO_INSTALL_DIR}/mobilenet_depthwise_conv" analyzer_resnet50_tester.cc "mobilenet_model.tar.gz" SERIAL)
-# bert, max_len=20
+# bert, max_len=20, embedding_dim=128
-set(BERT_INSTALL_DIR "${INFERENCE_DEMO_INSTALL_DIR}/bert20")
+set(BERT_INSTALL_DIR "${INFERENCE_DEMO_INSTALL_DIR}/bert_emb128")
-download_model_and_data(${BERT_INSTALL_DIR} "bert_model.tar.gz" "bert_data_len20.txt.tar.gz")
+download_model_and_data(${BERT_INSTALL_DIR} "bert_emb128_model.tar.gz" "bert_data_len20.txt.tar.gz")
 inference_analysis_api_test(test_analyzer_bert ${BERT_INSTALL_DIR} analyzer_bert_tester.cc SERIAL)
 # anakin

--- a/paddle/fluid/inference/utils/CMakeLists.txt
+++ b/paddle/fluid/inference/utils/CMakeLists.txt
 cc_library(benchmark SRCS benchmark.cc DEPS enforce)
 cc_test(test_benchmark SRCS benchmark_tester.cc DEPS benchmark)
-#cc_binary(visualizer SRCS visualizer.cc DEPS analysis
+cc_binary(visualizer SRCS visualizer.cc DEPS analysis
-#    paddle_pass_builder ir_pass_manager pass graph_viz_pass analysis_passes)
+    paddle_pass_builder ir_pass_manager pass graph_viz_pass analysis_passes)
--- a/paddle/fluid/operators/activation_op.cc
+++ b/paddle/fluid/operators/activation_op.cc
@@ -14,7 +14,7 @@ limitations under the License. */
 #include "paddle/fluid/operators/activation_op.h"
 #include <string>
-#include "paddle/fluid/operators/mkldnn_activation_op.h"
+#include "paddle/fluid/operators/mkldnn/mkldnn_activation_op.h"
 #include "paddle/fluid/platform/port.h"
 namespace paddle {

--- a/paddle/fluid/operators/beam_search_op.cc
+++ b/paddle/fluid/operators/beam_search_op.cc
@@ -51,6 +51,9 @@ class BeamSearchOpMaker : public framework::OpProtoAndCheckerMaker {
    AddOutput("selected_scores",
              "A LoDTensor containing the accumulated scores corresponding to "
              "Output(selected_ids).");
+    AddOutput(
+        "parent_idx",
+        "A Tensor preserving the selected_ids' parent indice in pre_ids.");
    // Attributes stored in AttributeMap
    AddAttr<int>("level", "the level of LoDTensor");

--- a/paddle/fluid/operators/beam_search_op.h
+++ b/paddle/fluid/operators/beam_search_op.h
@@ -41,13 +41,15 @@ class BeamSearchOpKernel : public framework::OpKernel<T> {
    auto selected_ids = context.Output<framework::LoDTensor>("selected_ids");
    auto selected_scores =
        context.Output<framework::LoDTensor>("selected_scores");
+    auto* parent_idx = context.Output<framework::Tensor>("parent_idx");
    PADDLE_ENFORCE_NOT_NULL(selected_ids);
    PADDLE_ENFORCE_NOT_NULL(selected_scores);
+    PADDLE_ENFORCE_NOT_NULL(parent_idx);
    math::BeamSearchFunctor<DeviceContext, T> alg;
    alg(context.template device_context<DeviceContext>(), pre_ids, pre_scores,
-        ids, scores, selected_ids, selected_scores, level, beam_size, end_id,
+        ids, scores, selected_ids, selected_scores, parent_idx, level,
-        is_accumulated);
+        beam_size, end_id, is_accumulated);
  }
 };

--- a/paddle/fluid/operators/detection/CMakeLists.txt
+++ b/paddle/fluid/operators/detection/CMakeLists.txt
@@ -32,6 +32,7 @@ polygon_box_transform_op.cu)
 detection_library(rpn_target_assign_op SRCS rpn_target_assign_op.cc)
 detection_library(generate_proposal_labels_op SRCS generate_proposal_labels_op.cc)
 detection_library(box_clip_op SRCS box_clip_op.cc box_clip_op.cu)
+detection_library(yolov3_loss_op SRCS yolov3_loss_op.cc)
 if(WITH_GPU)
  detection_library(generate_proposals_op SRCS generate_proposals_op.cc generate_proposals_op.cu DEPS memory cub)

--- a/paddle/fluid/operators/detection/box_coder_op.cc
+++ b/paddle/fluid/operators/detection/box_coder_op.cc
@@ -10,6 +10,7 @@ See the License for the specific language governing permissions and
 limitations under the License. */
 #include "paddle/fluid/operators/detection/box_coder_op.h"
+#include <vector>
 namespace paddle {
 namespace operators {
@@ -32,32 +33,57 @@ class BoxCoderOp : public framework::OperatorWithKernel {
    if (ctx->IsRuntime()) {
      PADDLE_ENFORCE_EQ(prior_box_dims.size(), 2,
-                        "The rank of Input of PriorBoxVar must be 2");
+                        "The rank of Input PriorBox must be 2");
      PADDLE_ENFORCE_EQ(prior_box_dims[1], 4,
                        "The shape of PriorBox is [N, 4]");
      if (ctx->HasInput("PriorBoxVar")) {
        auto prior_box_var_dims = ctx->GetInputDim("PriorBoxVar");
-        PADDLE_ENFORCE_EQ(prior_box_dims, prior_box_var_dims);
+        PADDLE_ENFORCE(
+            prior_box_var_dims.size() == 1 || prior_box_var_dims.size() == 2,
+            "Input(PriorBoxVar) of BoxCoderOp should be 1 or 2.");
+        if (prior_box_var_dims.size() == 1) {
+          PADDLE_ENFORCE_EQ(
+              prior_box_var_dims[0], 4,
+              "The 1st dimension of Input(PriorBoxVar) should be 4"
+              "when the rank is 1.");
+        } else {
+          PADDLE_ENFORCE_EQ(
+              prior_box_dims, prior_box_var_dims,
+              "The dimension of Input(PriorBoxVar) should be equal to"
+              "the dimension of Input(PriorBox when the rank is 2.)");
+        }
      }
+    }
-      auto code_type =
+    auto code_type = GetBoxCodeType(ctx->Attrs().Get<std::string>("code_type"));
-          GetBoxCodeType(ctx->Attrs().Get<std::string>("code_type"));
+    int axis = ctx->Attrs().Get<int>("axis");
-      if (code_type == BoxCodeType::kEncodeCenterSize) {
+    if (code_type == BoxCodeType::kEncodeCenterSize) {
-        PADDLE_ENFORCE_EQ(target_box_dims.size(), 2,
+      PADDLE_ENFORCE_EQ(target_box_dims.size(), 2,
-                          "The rank of Input of TargetBox must be 2");
+                        "The rank of Input TargetBox must be 2");
-        PADDLE_ENFORCE_EQ(target_box_dims[1], 4,
+      PADDLE_ENFORCE_EQ(target_box_dims[1], 4,
-                          "The shape of TargetBox is [M, 4]");
+                        "The shape of TargetBox is [M, 4]");
-      } else if (code_type == BoxCodeType::kDecodeCenterSize) {
+      ctx->SetOutputDim(
-        PADDLE_ENFORCE_EQ(target_box_dims.size(), 3,
+          "OutputBox",
-                          "The rank of Input of TargetBox must be 3");
+          framework::make_ddim({target_box_dims[0], prior_box_dims[0], 4}));
+    } else if (code_type == BoxCodeType::kDecodeCenterSize) {
+      PADDLE_ENFORCE_EQ(target_box_dims.size(), 3,
+                        "The rank of Input TargetBox must be 3");
+      if (axis == 0) {
        PADDLE_ENFORCE_EQ(target_box_dims[1], prior_box_dims[0]);
-        PADDLE_ENFORCE_EQ(target_box_dims[2], prior_box_dims[1]);
+      } else if (axis == 1) {
+        PADDLE_ENFORCE_EQ(target_box_dims[0], prior_box_dims[0]);
+      } else {
+        PADDLE_THROW("axis must be 0 or 1.");
      }
+      PADDLE_ENFORCE_EQ(target_box_dims[2], prior_box_dims[1]);
+      ctx->ShareDim("TargetBox", /*->*/ "OutputBox");
+    }
+    if (code_type == BoxCodeType::kDecodeCenterSize && axis == 1) {
+      ctx->ShareLoD("PriorBox", /*->*/ "OutputBox");
+    } else {
+      ctx->ShareLoD("TargetBox", /*->*/ "OutputBox");
    }
-    ctx->SetOutputDim(
-        "OutputBox",
-        framework::make_ddim({target_box_dims[0], prior_box_dims[0], 4}));
-    ctx->ShareLoD("TargetBox", /*->*/ "OutputBox");
  }
 };
@@ -100,6 +126,21 @@ class BoxCoderOpMaker : public framework::OpProtoAndCheckerMaker {
                  "(bool, default true) "
                  "whether treat the priorbox as a noramlized box")
        .SetDefault(true);
+    AddAttr<int>("axis",
+                 "(int, default 0)"
+                 "which axis in PriorBox to broadcast for box decode,"
+                 "for example, if axis is 0 and TargetBox has shape"
+                 "[N, M, 4] and PriorBox has shape [M, 4], then PriorBox "
+                 "will broadcast to [N, M, 4] for decoding. It is only valid"
+                 "when code type is decode_center_size")
+        .SetDefault(0)
+        .InEnum({0, 1});
+    AddAttr<std::vector<float>>(
+        "variance",
+        "(vector<float>, default {}),"
+        "variance of prior box with shape [4]. PriorBoxVar and variance can"
+        "not be provided at the same time.")
+        .SetDefault(std::vector<float>{});
    AddOutput("OutputBox",
              "(LoDTensor or Tensor) "
              "When code_type is 'encode_center_size', the output tensor of "
@@ -138,7 +179,11 @@ where `tx`, `ty`, `tw`, `th` denote the target box's center coordinates, width
 and height respectively. Similarly, `px`, `py`, `pw`, `ph` denote the
 priorbox's (anchor) center coordinates, width and height. `pxv`, `pyv`, `pwv`,
 `phv` denote the variance of the priorbox and `ox`, `oy`, `ow`, `oh` denote the
-encoded/decoded coordinates, width and height.
+encoded/decoded coordinates, width and height. 
+During Box Decoding, two modes for broadcast are supported. Say target box has 
+shape [N, M, 4], and the shape of prior box can be [N, 4] or [M, 4]. Then prior
+box will broadcast to target box along the assigned axis. 
 )DOC");
  }
 };

--- a/paddle/fluid/operators/detection/box_coder_op.cu
+++ b/paddle/fluid/operators/detection/box_coder_op.cu
@@ -9,6 +9,9 @@ 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 <thrust/device_vector.h>
+#include <thrust/host_vector.h>
+#include "paddle/fluid/memory/memcpy.h"
 #include "paddle/fluid/operators/detection/box_coder_op.h"
 #include "paddle/fluid/platform/cuda_primitives.h"
@@ -16,11 +19,11 @@ namespace paddle {
 namespace operators {
 template <typename T>
-__global__ void EncodeCenterSizeKernel(const T* prior_box_data,
+__global__ void EncodeCenterSizeKernel(
-                                       const T* prior_box_var_data,
+    const T* prior_box_data, const T* prior_box_var_data,
-                                       const T* target_box_data, const int row,
+    const T* target_box_data, const int row, const int col, const int len,
-                                       const int col, const int len,
+    const bool normalized, const T prior_box_var_size, const float* variance,
-                                       const bool normalized, T* output) {
+    const int var_size, T* output) {
  const int idx = threadIdx.x + blockIdx.x * blockDim.x;
  if (idx < row * col) {
    const int row_idx = idx / col;
@@ -30,11 +33,9 @@ __global__ void EncodeCenterSizeKernel(const T* prior_box_data,
    T prior_box_height = prior_box_data[col_idx * len + 3] -
                         prior_box_data[col_idx * len + 1] +
                         (normalized == false);
-    T prior_box_center_x =
+    T prior_box_center_x = prior_box_data[col_idx * len] + prior_box_width / 2;
-        (prior_box_data[col_idx * len + 2] + prior_box_data[col_idx * len]) / 2;
+    T prior_box_center_y =
-    T prior_box_center_y = (prior_box_data[col_idx * len + 3] +
+        prior_box_data[col_idx * len + 1] + prior_box_height / 2;
-                            prior_box_data[col_idx * len + 1]) /
-                           2;
    T target_box_center_x =
        (target_box_data[row_idx * len + 2] + target_box_data[row_idx * len]) /
@@ -55,58 +56,73 @@ __global__ void EncodeCenterSizeKernel(const T* prior_box_data,
    output[idx * len + 2] = log(fabs(target_box_width / prior_box_width));
    output[idx * len + 3] = log(fabs(target_box_height / prior_box_height));
    if (prior_box_var_data) {
-      output[idx * len] /= prior_box_var_data[col_idx * len];
+      int prior_var_offset = 0;
-      output[idx * len + 1] /= prior_box_var_data[col_idx * len + 1];
+      if (prior_box_var_size == 2) {
-      output[idx * len + 2] /= prior_box_var_data[col_idx * len + 2];
+        prior_var_offset = col_idx * len;
-      output[idx * len + 3] /= prior_box_var_data[col_idx * len + 3];
+      }
+      output[idx * len] /= prior_box_var_data[prior_var_offset];
+      output[idx * len + 1] /= prior_box_var_data[prior_var_offset + 1];
+      output[idx * len + 2] /= prior_box_var_data[prior_var_offset + 2];
+      output[idx * len + 3] /= prior_box_var_data[prior_var_offset + 3];
+    } else if (var_size == 4) {
+      for (int k = 0; k < 4; ++k) {
+        output[idx * len + k] /= static_cast<T>(variance[k]);
+      }
    }
  }
 }
 template <typename T>
-__global__ void DecodeCenterSizeKernel(const T* prior_box_data,
+__global__ void DecodeCenterSizeKernel(
-                                       const T* prior_box_var_data,
+    const T* prior_box_data, const T* prior_box_var_data,
-                                       const T* target_box_data, const int row,
+    const T* target_box_data, const int row, const int col, const int len,
-                                       const int col, const int len,
+    const bool normalized, const T prior_box_var_size, const float* variance,
-                                       const bool normalized, T* output) {
+    const int var_size, const int axis, T* output) {
  const int idx = threadIdx.x + blockIdx.x * blockDim.x;
+  int prior_box_offset = 0;
  if (idx < row * col) {
    const int col_idx = idx % col;
-    T prior_box_width = prior_box_data[col_idx * len + 2] -
+    const int row_idx = idx / col;
-                        prior_box_data[col_idx * len] + (normalized == false);
+    prior_box_offset = axis == 0 ? col_idx * len : row_idx * len;
-    T prior_box_height = prior_box_data[col_idx * len + 3] -
+    T prior_box_width = prior_box_data[prior_box_offset + 2] -
-                         prior_box_data[col_idx * len + 1] +
+                        prior_box_data[prior_box_offset] +
+                        (normalized == false);
+    T prior_box_height = prior_box_data[prior_box_offset + 3] -
+                         prior_box_data[prior_box_offset + 1] +
                         (normalized == false);
    T prior_box_center_x =
-        (prior_box_data[col_idx * len + 2] + prior_box_data[col_idx * len]) / 2;
+        prior_box_data[prior_box_offset] + prior_box_width / 2;
-    T prior_box_center_y = (prior_box_data[col_idx * len + 3] +
+    T prior_box_center_y =
-                            prior_box_data[col_idx * len + 1]) /
+        prior_box_data[prior_box_offset + 1] + prior_box_height / 2;
-                           2;
    T target_box_width, target_box_height;
    T target_box_center_x, target_box_center_y;
+    T box_var_x = T(1), box_var_y = T(1);
+    T box_var_w = T(1), box_var_h = T(1);
    if (prior_box_var_data) {
-      target_box_width = exp(prior_box_var_data[col_idx * len + 2] *
+      int prior_var_offset = 0;
-                             target_box_data[idx * len + 2]) *
+      if (prior_box_var_size == 2) {
-                         prior_box_width;
+        prior_var_offset = axis == 0 ? col_idx * len : row_idx * len;
-      target_box_height = exp(prior_box_var_data[col_idx * len + 3] *
+      }
-                              target_box_data[idx * len + 3]) *
+      box_var_x = prior_box_var_data[prior_var_offset];
-                          prior_box_height;
+      box_var_y = prior_box_var_data[prior_var_offset + 1];
-      target_box_center_x = prior_box_var_data[col_idx * len] *
+      box_var_w = prior_box_var_data[prior_var_offset + 2];
-                                target_box_data[idx * len] * prior_box_width +
+      box_var_h = prior_box_var_data[prior_var_offset + 3];
-                            prior_box_center_x;
+    } else if (var_size == 4) {
-      target_box_center_y = prior_box_var_data[col_idx * len + 1] *
+      box_var_x = static_cast<T>(variance[0]);
-                                target_box_data[idx * len + 1] *
+      box_var_y = static_cast<T>(variance[1]);
-                                prior_box_height +
+      box_var_w = static_cast<T>(variance[2]);
-                            prior_box_center_y;
+      box_var_h = static_cast<T>(variance[3]);
-    } else {
-      target_box_width = exp(target_box_data[idx * len + 2]) * prior_box_width;
-      target_box_height =
-          exp(target_box_data[idx * len + 3]) * prior_box_height;
-      target_box_center_x =
-          target_box_data[idx * len] * prior_box_width + prior_box_center_x;
-      target_box_center_y = target_box_data[idx * len + 1] * prior_box_height +
-                            prior_box_center_y;
    }
+    target_box_width =
+        exp(box_var_w * target_box_data[idx * len + 2]) * prior_box_width;
+    target_box_height =
+        exp(box_var_h * target_box_data[idx * len + 3]) * prior_box_height;
+    target_box_center_x =
+        box_var_x * target_box_data[idx * len] * prior_box_width +
+        prior_box_center_x;
+    target_box_center_y =
+        box_var_y * target_box_data[idx * len + 1] * prior_box_height +
+        prior_box_center_y;
    output[idx * len] = target_box_center_x - target_box_width / 2;
    output[idx * len + 1] = target_box_center_y - target_box_height / 2;
@@ -127,36 +143,64 @@ class BoxCoderCUDAKernel : public framework::OpKernel<T> {
    auto* prior_box_var = context.Input<framework::Tensor>("PriorBoxVar");
    auto* target_box = context.Input<framework::LoDTensor>("TargetBox");
    auto* output_box = context.Output<framework::Tensor>("OutputBox");
+    std::vector<float> variance = context.Attr<std::vector<float>>("variance");
    const T* prior_box_data = prior_box->data<T>();
    const T* target_box_data = target_box->data<T>();
    const T* prior_box_var_data = nullptr;
-    if (prior_box_var) prior_box_var_data = prior_box_var->data<T>();
+    auto prior_box_var_size = 0;
+    if (prior_box_var) {
+      PADDLE_ENFORCE(variance.empty(),
+                     "Input 'PriorBoxVar' and attribute 'variance' should not"
+                     "be used at the same time.");
+      prior_box_var_data = prior_box_var->data<T>();
+      prior_box_var_size = prior_box_var->dims().size();
+    }
+    if (!(variance.empty())) {
+      PADDLE_ENFORCE(static_cast<int>(variance.size()) == 4,
+                     "Size of attribute 'variance' should be 4");
+    }
    if (target_box->lod().size()) {
      PADDLE_ENFORCE_EQ(target_box->lod().size(), 1,
                        "Only support 1 level of LoD.");
    }
+    const int var_size = static_cast<int>(variance.size());
+    auto code_type = GetBoxCodeType(context.Attr<std::string>("code_type"));
+    bool normalized = context.Attr<bool>("box_normalized");
+    int axis = context.Attr<int>("axis");
    auto row = target_box->dims()[0];
    auto col = prior_box->dims()[0];
+    if (code_type == BoxCodeType::kDecodeCenterSize) {
+      col = target_box->dims()[1];
+    }
    auto len = prior_box->dims()[1];
    int block = 512;
    int grid = (row * col + block - 1) / block;
    auto& device_ctx = context.cuda_device_context();
+    auto& allocator =
+        platform::DeviceTemporaryAllocator::Instance().Get(device_ctx);
+    int bytes = var_size * sizeof(float);
+    auto dev_var = allocator.Allocate(bytes);
+    float* dev_var_data = reinterpret_cast<float*>(dev_var->ptr());
+    auto cplace = platform::CPUPlace();
+    const auto gplace = boost::get<platform::CUDAPlace>(context.GetPlace());
+    memory::Copy(gplace, dev_var_data, cplace, &variance[0], bytes,
+                 device_ctx.stream());
    output_box->mutable_data<T>({row, col, len}, context.GetPlace());
    T* output = output_box->data<T>();
-    auto code_type = GetBoxCodeType(context.Attr<std::string>("code_type"));
-    bool normalized = context.Attr<bool>("box_normalized");
    if (code_type == BoxCodeType::kEncodeCenterSize) {
      EncodeCenterSizeKernel<T><<<grid, block, 0, device_ctx.stream()>>>(
          prior_box_data, prior_box_var_data, target_box_data, row, col, len,
-          normalized, output);
+          normalized, prior_box_var_size, dev_var_data, var_size, output);
    } else if (code_type == BoxCodeType::kDecodeCenterSize) {
      DecodeCenterSizeKernel<T><<<grid, block, 0, device_ctx.stream()>>>(
          prior_box_data, prior_box_var_data, target_box_data, row, col, len,
-          normalized, output);
+          normalized, prior_box_var_size, dev_var_data, var_size, axis, output);
    }
  }
 };

--- a/paddle/fluid/operators/detection/box_coder_op.h
+++ b/paddle/fluid/operators/detection/box_coder_op.h
@@ -11,6 +11,7 @@ limitations under the License. */
 #pragma once
 #include <string>
+#include <vector>
 #include "paddle/fluid/framework/op_registry.h"
 #include "paddle/fluid/operators/math/math_function.h"
@@ -34,7 +35,8 @@ class BoxCoderKernel : public framework::OpKernel<T> {
  void EncodeCenterSize(const framework::Tensor* target_box,
                        const framework::Tensor* prior_box,
                        const framework::Tensor* prior_box_var,
-                        const bool normalized, T* output) const {
+                        const bool normalized,
+                        const std::vector<float> variance, T* output) const {
    int64_t row = target_box->dims()[0];
    int64_t col = prior_box->dims()[0];
    int64_t len = prior_box->dims()[1];
@@ -53,10 +55,9 @@ class BoxCoderKernel : public framework::OpKernel<T> {
        T prior_box_height = prior_box_data[j * len + 3] -
                             prior_box_data[j * len + 1] +
                             (normalized == false);
-        T prior_box_center_x =
+        T prior_box_center_x = prior_box_data[j * len] + prior_box_width / 2;
-            (prior_box_data[j * len + 2] + prior_box_data[j * len]) / 2;
        T prior_box_center_y =
-            (prior_box_data[j * len + 3] + prior_box_data[j * len + 1]) / 2;
+            prior_box_data[j * len + 1] + prior_box_height / 2;
        T target_box_center_x =
            (target_box_data[i * len + 2] + target_box_data[i * len]) / 2;
@@ -78,10 +79,18 @@ class BoxCoderKernel : public framework::OpKernel<T> {
        output[offset + 3] =
            std::log(std::fabs(target_box_height / prior_box_height));
        if (prior_box_var) {
-          output[offset] /= prior_box_var_data[j * len];
+          int prior_var_offset = 0;
-          output[offset + 1] /= prior_box_var_data[j * len + 1];
+          if (prior_box_var->dims().size() == 2) {
-          output[offset + 2] /= prior_box_var_data[j * len + 2];
+            prior_var_offset = j * len;
-          output[offset + 3] /= prior_box_var_data[j * len + 3];
+          }
+          output[offset] /= prior_box_var_data[prior_var_offset];
+          output[offset + 1] /= prior_box_var_data[prior_var_offset + 1];
+          output[offset + 2] /= prior_box_var_data[prior_var_offset + 2];
+          output[offset + 3] /= prior_box_var_data[prior_var_offset + 3];
+        } else if (!(variance.empty())) {
+          for (int k = 0; k < 4; ++k) {
+            output[offset + k] /= static_cast<T>(variance[k]);
+          }
        }
      }
    }
@@ -89,58 +98,71 @@ class BoxCoderKernel : public framework::OpKernel<T> {
  void DecodeCenterSize(const framework::Tensor* target_box,
                        const framework::Tensor* prior_box,
                        const framework::Tensor* prior_box_var,
-                        const bool normalized, T* output) const {
+                        const bool normalized, const int axis,
+                        const std::vector<float> variance, T* output) const {
    int64_t row = target_box->dims()[0];
-    int64_t col = prior_box->dims()[0];
+    int64_t col = target_box->dims()[1];
-    int64_t len = prior_box->dims()[1];
+    int64_t len = target_box->dims()[2];
    auto* target_box_data = target_box->data<T>();
    auto* prior_box_data = prior_box->data<T>();
    const T* prior_box_var_data = nullptr;
    if (prior_box_var) prior_box_var_data = prior_box_var->data<T>();
+    int prior_box_offset = 0;
 #ifdef PADDLE_WITH_MKLML
 #pragma omp parallel for collapse(2)
 #endif
    for (int64_t i = 0; i < row; ++i) {
      for (int64_t j = 0; j < col; ++j) {
        size_t offset = i * col * len + j * len;
-        T prior_box_width = prior_box_data[j * len + 2] -
+        if (axis == 0) {
-                            prior_box_data[j * len] + (normalized == false);
+          prior_box_offset = j * len;
-        T prior_box_height = prior_box_data[j * len + 3] -
+        } else if (axis == 1) {
-                             prior_box_data[j * len + 1] +
+          prior_box_offset = i * len;
+        }
+        T prior_box_width = prior_box_data[prior_box_offset + 2] -
+                            prior_box_data[prior_box_offset] +
+                            (normalized == false);
+        T prior_box_height = prior_box_data[prior_box_offset + 3] -
+                             prior_box_data[prior_box_offset + 1] +
                             (normalized == false);
        T prior_box_center_x =
-            (prior_box_data[j * len + 2] + prior_box_data[j * len]) / 2;
+            prior_box_data[prior_box_offset] + prior_box_width / 2;
        T prior_box_center_y =
-            (prior_box_data[j * len + 3] + prior_box_data[j * len + 1]) / 2;
+            prior_box_data[prior_box_offset + 1] + prior_box_height / 2;
        T target_box_center_x = 0, target_box_center_y = 0;
        T target_box_width = 0, target_box_height = 0;
+        T box_var_x = T(1), box_var_y = T(1);
+        T box_var_w = T(1), box_var_h = T(1);
        if (prior_box_var) {
-          target_box_center_x = prior_box_var_data[j * len] *
+          int prior_var_offset = 0;
-                                    target_box_data[offset] * prior_box_width +
+          if (prior_box_var->dims().size() == 2) {
-                                prior_box_center_x;
+            if (axis == 0)
-          target_box_center_y = prior_box_var_data[j * len + 1] *
+              prior_var_offset = j * len;
-                                    target_box_data[offset + 1] *
+            else if (axis == 1)
-                                    prior_box_height +
+              prior_var_offset = i * len;
-                                prior_box_center_y;
+          }
-          target_box_width = std::exp(prior_box_var_data[j * len + 2] *
+          box_var_x = prior_box_var_data[prior_var_offset];
-                                      target_box_data[offset + 2]) *
+          box_var_y = prior_box_var_data[prior_var_offset + 1];
-                             prior_box_width;
+          box_var_w = prior_box_var_data[prior_var_offset + 2];
-          target_box_height = std::exp(prior_box_var_data[j * len + 3] *
+          box_var_h = prior_box_var_data[prior_var_offset + 3];
-                                       target_box_data[offset + 3]) *
+        } else if (!(variance.empty())) {
-                              prior_box_height;
+          box_var_x = static_cast<T>(variance[0]);
-        } else {
+          box_var_y = static_cast<T>(variance[1]);
-          target_box_center_x =
+          box_var_w = static_cast<T>(variance[2]);
-              target_box_data[offset] * prior_box_width + prior_box_center_x;
+          box_var_h = static_cast<T>(variance[3]);
-          target_box_center_y = target_box_data[offset + 1] * prior_box_height +
-                                prior_box_center_y;
-          target_box_width =
-              std::exp(target_box_data[offset + 2]) * prior_box_width;
-          target_box_height =
-              std::exp(target_box_data[offset + 3]) * prior_box_height;
        }
+        target_box_center_x =
+            box_var_x * target_box_data[offset] * prior_box_width +
+            prior_box_center_x;
+        target_box_center_y =
+            box_var_y * target_box_data[offset + 1] * prior_box_height +
+            prior_box_center_y;
+        target_box_width =
+            std::exp(box_var_w * target_box_data[offset + 2]) * prior_box_width;
+        target_box_height = std::exp(box_var_h * target_box_data[offset + 3]) *
+                            prior_box_height;
        output[offset] = target_box_center_x - target_box_width / 2;
        output[offset + 1] = target_box_center_y - target_box_height / 2;
@@ -157,26 +179,40 @@ class BoxCoderKernel : public framework::OpKernel<T> {
    auto* prior_box_var = context.Input<framework::Tensor>("PriorBoxVar");
    auto* target_box = context.Input<framework::LoDTensor>("TargetBox");
    auto* output_box = context.Output<framework::Tensor>("OutputBox");
+    std::vector<float> variance = context.Attr<std::vector<float>>("variance");
+    const int axis = context.Attr<int>("axis");
    if (target_box->lod().size()) {
      PADDLE_ENFORCE_EQ(target_box->lod().size(), 1UL,
                        "Only support 1 level of LoD.");
    }
+    if (prior_box_var) {
+      PADDLE_ENFORCE(variance.empty(),
+                     "Input 'PriorBoxVar' and attribute 'variance' should not"
+                     "be used at the same time.");
+    }
+    if (!(variance.empty())) {
+      PADDLE_ENFORCE(static_cast<int>(variance.size()) == 4,
+                     "Size of attribute 'variance' should be 4");
+    }
+    auto code_type = GetBoxCodeType(context.Attr<std::string>("code_type"));
+    bool normalized = context.Attr<bool>("box_normalized");
    auto row = target_box->dims()[0];
    auto col = prior_box->dims()[0];
+    if (code_type == BoxCodeType::kDecodeCenterSize) {
+      col = target_box->dims()[1];
+    }
    auto len = prior_box->dims()[1];
    output_box->mutable_data<T>({row, col, len}, context.GetPlace());
-    auto code_type = GetBoxCodeType(context.Attr<std::string>("code_type"));
-    bool normalized = context.Attr<bool>("box_normalized");
    T* output = output_box->data<T>();
    if (code_type == BoxCodeType::kEncodeCenterSize) {
      EncodeCenterSize(target_box, prior_box, prior_box_var, normalized,
-                       output);
+                       variance, output);
    } else if (code_type == BoxCodeType::kDecodeCenterSize) {
-      DecodeCenterSize(target_box, prior_box, prior_box_var, normalized,
+      DecodeCenterSize(target_box, prior_box, prior_box_var, normalized, axis,
-                       output);
+                       variance, output);
    }
  }
 };

--- a/paddle/fluid/operators/yolov3_loss_op.cc
+++ b/paddle/fluid/operators/yolov3_loss_op.cc
@@ -9,7 +9,7 @@
   See the License for the specific language governing permissions and
   limitations under the License. */
-#include "paddle/fluid/operators/yolov3_loss_op.h"
+#include "paddle/fluid/operators/detection/yolov3_loss_op.h"
 #include "paddle/fluid/framework/op_registry.h"
 namespace paddle {
@@ -29,23 +29,33 @@ class Yolov3LossOp : public framework::OperatorWithKernel {
                   "Input(GTLabel) of Yolov3LossOp should not be null.");
    PADDLE_ENFORCE(ctx->HasOutput("Loss"),
                   "Output(Loss) of Yolov3LossOp should not be null.");
+    PADDLE_ENFORCE(
+        ctx->HasOutput("ObjectnessMask"),
+        "Output(ObjectnessMask) of Yolov3LossOp should not be null.");
+    PADDLE_ENFORCE(ctx->HasOutput("GTMatchMask"),
+                   "Output(GTMatchMask) of Yolov3LossOp should not be null.");
    auto dim_x = ctx->GetInputDim("X");
    auto dim_gtbox = ctx->GetInputDim("GTBox");
    auto dim_gtlabel = ctx->GetInputDim("GTLabel");
    auto anchors = ctx->Attrs().Get<std::vector<int>>("anchors");
+    int anchor_num = anchors.size() / 2;
+    auto anchor_mask = ctx->Attrs().Get<std::vector<int>>("anchor_mask");
+    int mask_num = anchor_mask.size();
    auto class_num = ctx->Attrs().Get<int>("class_num");
    PADDLE_ENFORCE_EQ(dim_x.size(), 4, "Input(X) should be a 4-D tensor.");
    PADDLE_ENFORCE_EQ(dim_x[2], dim_x[3],
                      "Input(X) dim[3] and dim[4] should be euqal.");
-    PADDLE_ENFORCE_EQ(dim_x[1], anchors.size() / 2 * (5 + class_num),
+    PADDLE_ENFORCE_EQ(
-                      "Input(X) dim[1] should be equal to (anchor_number * (5 "
+        dim_x[1], mask_num * (5 + class_num),
-                      "+ class_num)).");
+        "Input(X) dim[1] should be equal to (anchor_mask_number * (5 "
+        "+ class_num)).");
    PADDLE_ENFORCE_EQ(dim_gtbox.size(), 3,
                      "Input(GTBox) should be a 3-D tensor");
    PADDLE_ENFORCE_EQ(dim_gtbox[2], 4, "Input(GTBox) dim[2] should be 5");
    PADDLE_ENFORCE_EQ(dim_gtlabel.size(), 2,
-                      "Input(GTBox) should be a 2-D tensor");
+                      "Input(GTLabel) should be a 2-D tensor");
    PADDLE_ENFORCE_EQ(dim_gtlabel[0], dim_gtbox[0],
                      "Input(GTBox) and Input(GTLabel) dim[0] should be same");
    PADDLE_ENFORCE_EQ(dim_gtlabel[1], dim_gtbox[1],
@@ -54,11 +64,22 @@ class Yolov3LossOp : public framework::OperatorWithKernel {
                      "Attr(anchors) length should be greater then 0.");
    PADDLE_ENFORCE_EQ(anchors.size() % 2, 0,
                      "Attr(anchors) length should be even integer.");
+    for (size_t i = 0; i < anchor_mask.size(); i++) {
+      PADDLE_ENFORCE_LT(
+          anchor_mask[i], anchor_num,
+          "Attr(anchor_mask) should not crossover Attr(anchors).");
+    }
    PADDLE_ENFORCE_GT(class_num, 0,
                      "Attr(class_num) should be an integer greater then 0.");
-    std::vector<int64_t> dim_out({1});
+    std::vector<int64_t> dim_out({dim_x[0]});
    ctx->SetOutputDim("Loss", framework::make_ddim(dim_out));
+    std::vector<int64_t> dim_obj_mask({dim_x[0], mask_num, dim_x[2], dim_x[3]});
+    ctx->SetOutputDim("ObjectnessMask", framework::make_ddim(dim_obj_mask));
+    std::vector<int64_t> dim_gt_match_mask({dim_gtbox[0], dim_gtbox[1]});
+    ctx->SetOutputDim("GTMatchMask", framework::make_ddim(dim_gt_match_mask));
  }
 protected:
@@ -73,11 +94,11 @@ class Yolov3LossOpMaker : public framework::OpProtoAndCheckerMaker {
 public:
  void Make() override {
    AddInput("X",
-             "The input tensor of YOLO v3 loss operator, "
+             "The input tensor of YOLOv3 loss operator, "
             "This is a 4-D tensor with shape of [N, C, H, W]."
             "H and W should be same, and the second dimention(C) stores"
             "box locations, confidence score and classification one-hot"
-             "key of each anchor box");
+             "keys of each anchor box");
    AddInput("GTBox",
             "The input tensor of ground truth boxes, "
             "This is a 3-D tensor with shape of [N, max_box_num, 5], "
@@ -89,32 +110,39 @@ class Yolov3LossOpMaker : public framework::OpProtoAndCheckerMaker {
    AddInput("GTLabel",
             "The input tensor of ground truth label, "
             "This is a 2-D tensor with shape of [N, max_box_num], "
-             "and each element shoudl be an integer to indicate the "
+             "and each element should be an integer to indicate the "
             "box class id.");
    AddOutput("Loss",
              "The output yolov3 loss tensor, "
-              "This is a 1-D tensor with shape of [1]");
+              "This is a 1-D tensor with shape of [N]");
+    AddOutput("ObjectnessMask",
+              "This is an intermediate tensor with shape of [N, M, H, W], "
+              "M is the number of anchor masks. This parameter caches the "
+              "mask for calculate objectness loss in gradient kernel.")
+        .AsIntermediate();
+    AddOutput("GTMatchMask",
+              "This is an intermediate tensor with shape of [N, B], "
+              "B is the max box number of GT boxes. This parameter caches "
+              "matched mask index of each GT boxes for gradient calculate.")
+        .AsIntermediate();
    AddAttr<int>("class_num", "The number of classes to predict.");
    AddAttr<std::vector<int>>("anchors",
                              "The anchor width and height, "
-                              "it will be parsed pair by pair.");
+                              "it will be parsed pair by pair.")
+        .SetDefault(std::vector<int>{});
+    AddAttr<std::vector<int>>("anchor_mask",
+                              "The mask index of anchors used in "
+                              "current YOLOv3 loss calculation.")
+        .SetDefault(std::vector<int>{});
+    AddAttr<int>("downsample_ratio",
+                 "The downsample ratio from network input to YOLOv3 loss "
+                 "input, so 32, 16, 8 should be set for the first, second, "
+                 "and thrid YOLOv3 loss operators.")
+        .SetDefault(32);
    AddAttr<float>("ignore_thresh",
-                   "The ignore threshold to ignore confidence loss.");
+                   "The ignore threshold to ignore confidence loss.")
-    AddAttr<float>("loss_weight_xy", "The weight of x, y location loss.")
+        .SetDefault(0.7);
-        .SetDefault(1.0);
-    AddAttr<float>("loss_weight_wh", "The weight of w, h location loss.")
-        .SetDefault(1.0);
-    AddAttr<float>(
-        "loss_weight_conf_target",
-        "The weight of confidence score loss in locations with target object.")
-        .SetDefault(1.0);
-    AddAttr<float>("loss_weight_conf_notarget",
-                   "The weight of confidence score loss in locations without "
-                   "target object.")
-        .SetDefault(1.0);
-    AddAttr<float>("loss_weight_class", "The weight of classification loss.")
-        .SetDefault(1.0);
    AddComment(R"DOC(
         This operator generate yolov3 loss by given predict result and ground
         truth boxes.
@@ -147,17 +175,28 @@ class Yolov3LossOpMaker : public framework::OpProtoAndCheckerMaker {
         thresh, the confidence score loss of this anchor box will be ignored.
         Therefore, the yolov3 loss consist of three major parts, box location loss,
-         confidence score loss, and classification loss. The MSE loss is used for 
+         confidence score loss, and classification loss. The L2 loss is used for 
-         box location, and binary cross entropy loss is used for confidence score 
+         box coordinates (w, h), and sigmoid cross entropy loss is used for box 
-         loss and classification loss.
+         coordinates (x, y), confidence score loss and classification loss.
+         Each groud truth box find a best matching anchor box in all anchors, 
+         prediction of this anchor box will incur all three parts of losses, and
+         prediction of anchor boxes with no GT box matched will only incur objectness
+         loss.
+         In order to trade off box coordinate losses between big boxes and small 
+         boxes, box coordinate losses will be mutiplied by scale weight, which is
+         calculated as follow.
+         $$
+         weight_{box} = 2.0 - t_w * t_h
+         $$
         Final loss will be represented as follow.
         $$
-         loss = \loss_weight_{xy} * loss_{xy} + \loss_weight_{wh} * loss_{wh}
+         loss = (loss_{xy} + loss_{wh}) * weight_{box}
-              + \loss_weight_{conf_target} * loss_{conf_target}
+              + loss_{conf} + loss_{class}
-              + \loss_weight_{conf_notarget} * loss_{conf_notarget}
-              + \loss_weight_{class} * loss_{class}
         $$
         )DOC");
  }
@@ -196,6 +235,8 @@ class Yolov3LossGradMaker : public framework::SingleGradOpDescMaker {
    op->SetInput("GTBox", Input("GTBox"));
    op->SetInput("GTLabel", Input("GTLabel"));
    op->SetInput(framework::GradVarName("Loss"), OutputGrad("Loss"));
+    op->SetInput("ObjectnessMask", Output("ObjectnessMask"));
+    op->SetInput("GTMatchMask", Output("GTMatchMask"));
    op->SetAttrMap(Attrs());

--- a/paddle/fluid/operators/detection/yolov3_loss_op.h
+++ b/paddle/fluid/operators/detection/yolov3_loss_op.h
+/* Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserve.
+   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
+   http://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. */
+#pragma once
+#include <algorithm>
+#include <vector>
+#include "paddle/fluid/framework/op_registry.h"
+#include "paddle/fluid/operators/math/math_function.h"
+namespace paddle {
+namespace operators {
+using Tensor = framework::Tensor;
+template <typename T, size_t D, int MajorType = Eigen::RowMajor,
+          typename IndexType = Eigen::DenseIndex>
+using EigenTensor = framework::EigenTensor<T, D, MajorType, IndexType>;
+template <typename T, int MajorType = Eigen::RowMajor,
+          typename IndexType = Eigen::DenseIndex>
+using EigenVector = framework::EigenVector<T, MajorType, IndexType>;
+template <typename T>
+static inline bool LessEqualZero(T x) {
+  return x < 1e-6;
+}
+template <typename T>
+static T SigmoidCrossEntropy(T x, T label) {
+  return (x > 0 ? x : 0.0) - x * label + std::log(1.0 + std::exp(-std::abs(x)));
+}
+template <typename T>
+static T L2Loss(T x, T y) {
+  return 0.5 * (y - x) * (y - x);
+}
+template <typename T>
+static T SigmoidCrossEntropyGrad(T x, T label) {
+  return 1.0 / (1.0 + std::exp(-x)) - label;
+}
+template <typename T>
+static T L2LossGrad(T x, T y) {
+  return x - y;
+}
+static int GetMaskIndex(std::vector<int> mask, int val) {
+  for (size_t i = 0; i < mask.size(); i++) {
+    if (mask[i] == val) {
+      return i;
+    }
+  }
+  return -1;
+}
+template <typename T>
+struct Box {
+  T x, y, w, h;
+};
+template <typename T>
+static inline T sigmoid(T x) {
+  return 1.0 / (1.0 + std::exp(-x));
+}
+template <typename T>
+static inline Box<T> GetYoloBox(const T* x, std::vector<int> anchors, int i,
+                                int j, int an_idx, int grid_size,
+                                int input_size, int index, int stride) {
+  Box<T> b;
+  b.x = (i + sigmoid<T>(x[index])) / grid_size;
+  b.y = (j + sigmoid<T>(x[index + stride])) / grid_size;
+  b.w = std::exp(x[index + 2 * stride]) * anchors[2 * an_idx] / input_size;
+  b.h = std::exp(x[index + 3 * stride]) * anchors[2 * an_idx + 1] / input_size;
+  return b;
+}
+template <typename T>
+static inline Box<T> GetGtBox(const T* gt, int batch, int max_boxes, int idx) {
+  Box<T> b;
+  b.x = gt[(batch * max_boxes + idx) * 4];
+  b.y = gt[(batch * max_boxes + idx) * 4 + 1];
+  b.w = gt[(batch * max_boxes + idx) * 4 + 2];
+  b.h = gt[(batch * max_boxes + idx) * 4 + 3];
+  return b;
+}
+template <typename T>
+static inline T BoxOverlap(T c1, T w1, T c2, T w2) {
+  T l1 = c1 - w1 / 2.0;
+  T l2 = c2 - w2 / 2.0;
+  T left = l1 > l2 ? l1 : l2;
+  T r1 = c1 + w1 / 2.0;
+  T r2 = c2 + w2 / 2.0;
+  T right = r1 < r2 ? r1 : r2;
+  return right - left;
+}
+template <typename T>
+static inline T CalcBoxIoU(Box<T> b1, Box<T> b2) {
+  T w = BoxOverlap(b1.x, b1.w, b2.x, b2.w);
+  T h = BoxOverlap(b1.y, b1.h, b2.y, b2.h);
+  T inter_area = (w < 0 || h < 0) ? 0.0 : w * h;
+  T union_area = b1.w * b1.h + b2.w * b2.h - inter_area;
+  return inter_area / union_area;
+}
+static inline int GetEntryIndex(int batch, int an_idx, int hw_idx, int an_num,
+                                int an_stride, int stride, int entry) {
+  return (batch * an_num + an_idx) * an_stride + entry * stride + hw_idx;
+}
+template <typename T>
+static void CalcBoxLocationLoss(T* loss, const T* input, Box<T> gt,
+                                std::vector<int> anchors, int an_idx,
+                                int box_idx, int gi, int gj, int grid_size,
+                                int input_size, int stride) {
+  T tx = gt.x * grid_size - gi;
+  T ty = gt.y * grid_size - gj;
+  T tw = std::log(gt.w * input_size / anchors[2 * an_idx]);
+  T th = std::log(gt.h * input_size / anchors[2 * an_idx + 1]);
+  T scale = (2.0 - gt.w * gt.h);
+  loss[0] += SigmoidCrossEntropy<T>(input[box_idx], tx) * scale;
+  loss[0] += SigmoidCrossEntropy<T>(input[box_idx + stride], ty) * scale;
+  loss[0] += L2Loss<T>(input[box_idx + 2 * stride], tw) * scale;
+  loss[0] += L2Loss<T>(input[box_idx + 3 * stride], th) * scale;
+}
+template <typename T>
+static void CalcBoxLocationLossGrad(T* input_grad, const T loss, const T* input,
+                                    Box<T> gt, std::vector<int> anchors,
+                                    int an_idx, int box_idx, int gi, int gj,
+                                    int grid_size, int input_size, int stride) {
+  T tx = gt.x * grid_size - gi;
+  T ty = gt.y * grid_size - gj;
+  T tw = std::log(gt.w * input_size / anchors[2 * an_idx]);
+  T th = std::log(gt.h * input_size / anchors[2 * an_idx + 1]);
+  T scale = (2.0 - gt.w * gt.h);
+  input_grad[box_idx] =
+      SigmoidCrossEntropyGrad<T>(input[box_idx], tx) * scale * loss;
+  input_grad[box_idx + stride] =
+      SigmoidCrossEntropyGrad<T>(input[box_idx + stride], ty) * scale * loss;
+  input_grad[box_idx + 2 * stride] =
+      L2LossGrad<T>(input[box_idx + 2 * stride], tw) * scale * loss;
+  input_grad[box_idx + 3 * stride] =
+      L2LossGrad<T>(input[box_idx + 3 * stride], th) * scale * loss;
+}
+template <typename T>
+static inline void CalcLabelLoss(T* loss, const T* input, const int index,
+                                 const int label, const int class_num,
+                                 const int stride) {
+  for (int i = 0; i < class_num; i++) {
+    T pred = input[index + i * stride];
+    loss[0] += SigmoidCrossEntropy<T>(pred, (i == label) ? 1.0 : 0.0);
+  }
+}
+template <typename T>
+static inline void CalcLabelLossGrad(T* input_grad, const T loss,
+                                     const T* input, const int index,
+                                     const int label, const int class_num,
+                                     const int stride) {
+  for (int i = 0; i < class_num; i++) {
+    T pred = input[index + i * stride];
+    input_grad[index + i * stride] =
+        SigmoidCrossEntropyGrad<T>(pred, (i == label) ? 1.0 : 0.0) * loss;
+  }
+}
+template <typename T>
+static inline void CalcObjnessLoss(T* loss, const T* input, const T* objness,
+                                   const int n, const int an_num, const int h,
+                                   const int w, const int stride,
+                                   const int an_stride) {
+  for (int i = 0; i < n; i++) {
+    for (int j = 0; j < an_num; j++) {
+      for (int k = 0; k < h; k++) {
+        for (int l = 0; l < w; l++) {
+          T obj = objness[k * w + l];
+          if (obj > 1e-5) {
+            // positive sample: obj = 1
+            loss[i] += SigmoidCrossEntropy<T>(input[k * w + l], 1.0);
+          } else if (obj > -0.5) {
+            // negetive sample: obj = 0
+            loss[i] += SigmoidCrossEntropy<T>(input[k * w + l], 0.0);
+          }
+        }
+      }
+      objness += stride;
+      input += an_stride;
+    }
+  }
+}
+template <typename T>
+static inline void CalcObjnessLossGrad(T* input_grad, const T* loss,
+                                       const T* input, const T* objness,
+                                       const int n, const int an_num,
+                                       const int h, const int w,
+                                       const int stride, const int an_stride) {
+  for (int i = 0; i < n; i++) {
+    for (int j = 0; j < an_num; j++) {
+      for (int k = 0; k < h; k++) {
+        for (int l = 0; l < w; l++) {
+          T obj = objness[k * w + l];
+          if (obj > 1e-5) {
+            input_grad[k * w + l] =
+                SigmoidCrossEntropyGrad<T>(input[k * w + l], 1.0) * loss[i];
+          } else if (obj > -0.5) {
+            input_grad[k * w + l] =
+                SigmoidCrossEntropyGrad<T>(input[k * w + l], 0.0) * loss[i];
+          }
+        }
+      }
+      objness += stride;
+      input += an_stride;
+      input_grad += an_stride;
+    }
+  }
+}
+template <typename T>
+static void inline GtValid(bool* valid, const T* gtbox, const int n,
+                           const int b) {
+  for (int i = 0; i < n; i++) {
+    for (int j = 0; j < b; j++) {
+      if (LessEqualZero(gtbox[j * 4 + 2]) || LessEqualZero(gtbox[j * 4 + 3])) {
+        valid[j] = false;
+      } else {
+        valid[j] = true;
+      }
+    }
+    valid += b;
+    gtbox += b * 4;
+  }
+}
+template <typename T>
+class Yolov3LossKernel : public framework::OpKernel<T> {
+ public:
+  void Compute(const framework::ExecutionContext& ctx) const override {
+    auto* input = ctx.Input<Tensor>("X");
+    auto* gt_box = ctx.Input<Tensor>("GTBox");
+    auto* gt_label = ctx.Input<Tensor>("GTLabel");
+    auto* loss = ctx.Output<Tensor>("Loss");
+    auto* objness_mask = ctx.Output<Tensor>("ObjectnessMask");
+    auto* gt_match_mask = ctx.Output<Tensor>("GTMatchMask");
+    auto anchors = ctx.Attr<std::vector<int>>("anchors");
+    auto anchor_mask = ctx.Attr<std::vector<int>>("anchor_mask");
+    int class_num = ctx.Attr<int>("class_num");
+    float ignore_thresh = ctx.Attr<float>("ignore_thresh");
+    int downsample_ratio = ctx.Attr<int>("downsample_ratio");
+    const int n = input->dims()[0];
+    const int h = input->dims()[2];
+    const int w = input->dims()[3];
+    const int an_num = anchors.size() / 2;
+    const int mask_num = anchor_mask.size();
+    const int b = gt_box->dims()[1];
+    int input_size = downsample_ratio * h;
+    const int stride = h * w;
+    const int an_stride = (class_num + 5) * stride;
+    const T* input_data = input->data<T>();
+    const T* gt_box_data = gt_box->data<T>();
+    const int* gt_label_data = gt_label->data<int>();
+    T* loss_data = loss->mutable_data<T>({n}, ctx.GetPlace());
+    memset(loss_data, 0, loss->numel() * sizeof(T));
+    T* obj_mask_data =
+        objness_mask->mutable_data<T>({n, mask_num, h, w}, ctx.GetPlace());
+    memset(obj_mask_data, 0, objness_mask->numel() * sizeof(T));
+    int* gt_match_mask_data =
+        gt_match_mask->mutable_data<int>({n, b}, ctx.GetPlace());
+    // calc valid gt box mask, avoid calc duplicately in following code
+    Tensor gt_valid_mask;
+    bool* gt_valid_mask_data =
+        gt_valid_mask.mutable_data<bool>({n, b}, ctx.GetPlace());
+    GtValid<T>(gt_valid_mask_data, gt_box_data, n, b);
+    for (int i = 0; i < n; i++) {
+      for (int j = 0; j < mask_num; j++) {
+        for (int k = 0; k < h; k++) {
+          for (int l = 0; l < w; l++) {
+            // each predict box find a best match gt box, if overlap is bigger
+            // then ignore_thresh, ignore the objectness loss.
+            int box_idx =
+                GetEntryIndex(i, j, k * w + l, mask_num, an_stride, stride, 0);
+            Box<T> pred = GetYoloBox(input_data, anchors, l, k, anchor_mask[j],
+                                     h, input_size, box_idx, stride);
+            T best_iou = 0;
+            for (int t = 0; t < b; t++) {
+              if (!gt_valid_mask_data[i * b + t]) {
+                continue;
+              }
+              Box<T> gt = GetGtBox(gt_box_data, i, b, t);
+              T iou = CalcBoxIoU(pred, gt);
+              if (iou > best_iou) {
+                best_iou = iou;
+              }
+            }
+            // If best IoU is bigger then ignore_thresh,
+            // ignore the objectness loss.
+            if (best_iou > ignore_thresh) {
+              int obj_idx = (i * mask_num + j) * stride + k * w + l;
+              obj_mask_data[obj_idx] = static_cast<T>(-1);
+            }
+            // all losses should be calculated if best IoU
+            // is bigger then truth thresh, but currently,
+            // truth thresh is an unreachable value as 1.0.
+          }
+        }
+      }
+      for (int t = 0; t < b; t++) {
+        if (!gt_valid_mask_data[i * b + t]) {
+          gt_match_mask_data[i * b + t] = -1;
+          continue;
+        }
+        Box<T> gt = GetGtBox(gt_box_data, i, b, t);
+        int gi = static_cast<int>(gt.x * w);
+        int gj = static_cast<int>(gt.y * h);
+        Box<T> gt_shift = gt;
+        gt_shift.x = 0.0;
+        gt_shift.y = 0.0;
+        T best_iou = 0.0;
+        int best_n = 0;
+        // each gt box find a best match anchor box as positive sample,
+        // for positive sample, all losses should be calculated, and for
+        // other samples, only objectness loss is required.
+        for (int an_idx = 0; an_idx < an_num; an_idx++) {
+          Box<T> an_box;
+          an_box.x = 0.0;
+          an_box.y = 0.0;
+          an_box.w = anchors[2 * an_idx] / static_cast<T>(input_size);
+          an_box.h = anchors[2 * an_idx + 1] / static_cast<T>(input_size);
+          float iou = CalcBoxIoU<T>(an_box, gt_shift);
+          if (iou > best_iou) {
+            best_iou = iou;
+            best_n = an_idx;
+          }
+        }
+        int mask_idx = GetMaskIndex(anchor_mask, best_n);
+        gt_match_mask_data[i * b + t] = mask_idx;
+        if (mask_idx >= 0) {
+          int box_idx = GetEntryIndex(i, mask_idx, gj * w + gi, mask_num,
+                                      an_stride, stride, 0);
+          CalcBoxLocationLoss<T>(loss_data + i, input_data, gt, anchors, best_n,
+                                 box_idx, gi, gj, h, input_size, stride);
+          int obj_idx = (i * mask_num + mask_idx) * stride + gj * w + gi;
+          obj_mask_data[obj_idx] = 1.0;
+          int label = gt_label_data[i * b + t];
+          int label_idx = GetEntryIndex(i, mask_idx, gj * w + gi, mask_num,
+                                        an_stride, stride, 5);
+          CalcLabelLoss<T>(loss_data + i, input_data, label_idx, label,
+                           class_num, stride);
+        }
+      }
+    }
+    CalcObjnessLoss<T>(loss_data, input_data + 4 * stride, obj_mask_data, n,
+                       mask_num, h, w, stride, an_stride);
+  }
+};
+template <typename T>
+class Yolov3LossGradKernel : public framework::OpKernel<T> {
+ public:
+  void Compute(const framework::ExecutionContext& ctx) const override {
+    auto* input = ctx.Input<Tensor>("X");
+    auto* gt_box = ctx.Input<Tensor>("GTBox");
+    auto* gt_label = ctx.Input<Tensor>("GTLabel");
+    auto* input_grad = ctx.Output<Tensor>(framework::GradVarName("X"));
+    auto* loss_grad = ctx.Input<Tensor>(framework::GradVarName("Loss"));
+    auto* objness_mask = ctx.Input<Tensor>("ObjectnessMask");
+    auto* gt_match_mask = ctx.Input<Tensor>("GTMatchMask");
+    auto anchors = ctx.Attr<std::vector<int>>("anchors");
+    auto anchor_mask = ctx.Attr<std::vector<int>>("anchor_mask");
+    int class_num = ctx.Attr<int>("class_num");
+    int downsample_ratio = ctx.Attr<int>("downsample_ratio");
+    const int n = input_grad->dims()[0];
+    const int c = input_grad->dims()[1];
+    const int h = input_grad->dims()[2];
+    const int w = input_grad->dims()[3];
+    const int mask_num = anchor_mask.size();
+    const int b = gt_match_mask->dims()[1];
+    int input_size = downsample_ratio * h;
+    const int stride = h * w;
+    const int an_stride = (class_num + 5) * stride;
+    const T* input_data = input->data<T>();
+    const T* gt_box_data = gt_box->data<T>();
+    const int* gt_label_data = gt_label->data<int>();
+    const T* loss_grad_data = loss_grad->data<T>();
+    const T* obj_mask_data = objness_mask->data<T>();
+    const int* gt_match_mask_data = gt_match_mask->data<int>();
+    T* input_grad_data =
+        input_grad->mutable_data<T>({n, c, h, w}, ctx.GetPlace());
+    memset(input_grad_data, 0, input_grad->numel() * sizeof(T));
+    for (int i = 0; i < n; i++) {
+      for (int t = 0; t < b; t++) {
+        int mask_idx = gt_match_mask_data[i * b + t];
+        if (mask_idx >= 0) {
+          Box<T> gt = GetGtBox(gt_box_data, i, b, t);
+          int gi = static_cast<int>(gt.x * w);
+          int gj = static_cast<int>(gt.y * h);
+          int box_idx = GetEntryIndex(i, mask_idx, gj * w + gi, mask_num,
+                                      an_stride, stride, 0);
+          CalcBoxLocationLossGrad<T>(
+              input_grad_data, loss_grad_data[i], input_data, gt, anchors,
+              anchor_mask[mask_idx], box_idx, gi, gj, h, input_size, stride);
+          int label = gt_label_data[i * b + t];
+          int label_idx = GetEntryIndex(i, mask_idx, gj * w + gi, mask_num,
+                                        an_stride, stride, 5);
+          CalcLabelLossGrad<T>(input_grad_data, loss_grad_data[i], input_data,
+                               label_idx, label, class_num, stride);
+        }
+      }
+    }
+    CalcObjnessLossGrad<T>(input_grad_data + 4 * stride, loss_grad_data,
+                           input_data + 4 * stride, obj_mask_data, n, mask_num,
+                           h, w, stride, an_stride);
+  }
+};
+}  // namespace operators
+}  // namespace paddle
--- a/paddle/fluid/operators/dropout_op.cu
+++ b/paddle/fluid/operators/dropout_op.cu
@@ -114,4 +114,5 @@ REGISTER_OP_CUDA_KERNEL(
    ops::GPUDropoutKernel<plat::CUDADeviceContext, double>);
 REGISTER_OP_CUDA_KERNEL(
    dropout_grad, ops::DropoutGradKernel<plat::CUDADeviceContext, float>,
+    ops::DropoutGradKernel<plat::CUDADeviceContext, plat::float16>,
    ops::DropoutGradKernel<plat::CUDADeviceContext, double>);
--- a/paddle/fluid/operators/elementwise/elementwise_add_mkldnn_op.cc
+++ b/paddle/fluid/operators/elementwise/elementwise_add_mkldnn_op.cc
--- a/paddle/fluid/operators/elementwise/elementwise_mul_mkldnn_op.cc
+++ b/paddle/fluid/operators/elementwise/elementwise_mul_mkldnn_op.cc
--- a/paddle/fluid/operators/gather_op.cu
+++ b/paddle/fluid/operators/gather_op.cu
@@ -31,7 +31,7 @@ class GatherOpCUDAKernel : public framework::OpKernel<T> {
    auto *output = ctx.Output<Tensor>("Out");
    output->mutable_data<T>(ctx.GetPlace());
+    if (x->numel() == 0) return;
    GPUGather<T>(ctx.device_context(), *x, *index, output);
  }
 };
@@ -45,14 +45,13 @@ class GatherGradOpCUDAKernel : public framework::OpKernel<T> {
    auto *Index = ctx.Input<Tensor>("Index");
    auto *dX = ctx.Output<Tensor>(framework::GradVarName("X"));
    auto *dO = ctx.Input<Tensor>(framework::GradVarName("Out"));
-    auto *x = ctx.Input<Tensor>("X");
    dX->mutable_data<T>(ctx.GetPlace());
    auto dxt = framework::EigenVector<T>::Flatten(*dX);
    auto &place = *ctx.template device_context<platform::CUDADeviceContext>()
                       .eigen_device();
    dxt.device(place) = dxt.constant(static_cast<T>(0));
+    if (dO->numel() == 0) return;
    GPUScatterAssign<T>(ctx.device_context(), *dO, *Index, dX);
  }
 };
@@ -61,11 +60,14 @@ class GatherGradOpCUDAKernel : public framework::OpKernel<T> {
 }  // namespace paddle
 namespace ops = paddle::operators;
+namespace plat = paddle::platform;
 REGISTER_OP_CUDA_KERNEL(gather, ops::GatherOpCUDAKernel<float>,
                        ops::GatherOpCUDAKernel<double>,
                        ops::GatherOpCUDAKernel<int64_t>,
-                        ops::GatherOpCUDAKernel<int>);
+                        ops::GatherOpCUDAKernel<int>,
+                        ops::GatherOpCUDAKernel<plat::float16>);
 REGISTER_OP_CUDA_KERNEL(gather_grad, ops::GatherGradOpCUDAKernel<float>,
                        ops::GatherGradOpCUDAKernel<double>,
                        ops::GatherGradOpCUDAKernel<int64_t>,
-                        ops::GatherGradOpCUDAKernel<int>);
+                        ops::GatherGradOpCUDAKernel<int>,
+                        ops::GatherGradOpCUDAKernel<plat::float16>);
--- a/paddle/fluid/operators/gather_op.h
+++ b/paddle/fluid/operators/gather_op.h
@@ -35,7 +35,7 @@ class GatherOpKernel : public framework::OpKernel<T> {
    auto *output = ctx.Output<Tensor>("Out");
    output->mutable_data<T>(ctx.GetPlace());
+    if (x->numel() == 0) return;
    CPUGather<T>(ctx.device_context(), *x, *index, output);
  }
 };
@@ -56,7 +56,7 @@ class GatherGradientOpKernel : public framework::OpKernel<T> {
    auto &place = *ctx.template device_context<platform::CPUDeviceContext>()
                       .eigen_device();
    dxt.device(place) = dxt.constant(static_cast<T>(0));
+    if (dO->numel() == 0) return;
    ScatterAssign<T>(ctx.device_context(), *dO, *Index, dX);
  }
 };

--- a/paddle/fluid/operators/lookup_table_op.cu
+++ b/paddle/fluid/operators/lookup_table_op.cu
@@ -17,6 +17,7 @@ limitations under the License. */
 #include "paddle/fluid/operators/lookup_table_op.h"
 #include "paddle/fluid/platform/assert.h"
 #include "paddle/fluid/platform/cuda_primitives.h"
+#include "paddle/fluid/platform/float16.h"
 namespace paddle {
 namespace operators {
@@ -193,8 +194,11 @@ class LookupTableGradCUDAKernel : public framework::OpKernel<T> {
 }  // namespace paddle
 namespace ops = paddle::operators;
+namespace plat = paddle::platform;
 REGISTER_OP_CUDA_KERNEL(lookup_table, ops::LookupTableCUDAKernel<float>,
-                        ops::LookupTableCUDAKernel<double>);
+                        ops::LookupTableCUDAKernel<double>,
+                        ops::LookupTableCUDAKernel<plat::float16>);
 REGISTER_OP_CUDA_KERNEL(lookup_table_grad,
                        ops::LookupTableGradCUDAKernel<float>,
-                        ops::LookupTableGradCUDAKernel<double>);
+                        ops::LookupTableGradCUDAKernel<double>,
+                        ops::LookupTableGradCUDAKernel<plat::float16>);
--- a/paddle/fluid/operators/math/beam_search.cc
+++ b/paddle/fluid/operators/math/beam_search.cc
@@ -29,8 +29,9 @@ class BeamSearchFunctor<platform::CPUDeviceContext, T> {
                  const framework::LoDTensor *ids,
                  const framework::LoDTensor *scores,
                  framework::LoDTensor *selected_ids,
-                  framework::LoDTensor *selected_scores, size_t level,
+                  framework::LoDTensor *selected_scores,
-                  size_t beam_size, int end_id, bool is_accumulated) {
+                  framework::Tensor *parent_idx, size_t level, size_t beam_size,
+                  int end_id, bool is_accumulated) {
    auto abs_lod = framework::ToAbsOffset(scores->lod());
    auto &high_level = abs_lod[level];
@@ -57,11 +58,13 @@ class BeamSearchFunctor<platform::CPUDeviceContext, T> {
        std::vector<int64_t>({static_cast<int>(num_instances), 1}));
    selected_ids->Resize(dims);
    selected_scores->Resize(dims);
+    parent_idx->Resize({static_cast<int64_t>(num_instances)});
    auto *selected_ids_data =
        selected_ids->mutable_data<int64_t>(platform::CPUPlace());
    auto *selected_scores_data =
        selected_scores->mutable_data<float>(platform::CPUPlace());
+    auto *parent_idx_data = parent_idx->mutable_data<int>(platform::CPUPlace());
    // fill in data
    std::vector<size_t> low_level;
@@ -69,6 +72,7 @@ class BeamSearchFunctor<platform::CPUDeviceContext, T> {
    for (auto &items : selected_items) {
      low_level.push_back(low_offset);
      for (auto &item : items) {
+        parent_idx_data[low_offset] = static_cast<int>(low_level.size() - 1);
        selected_ids_data[low_offset] = item.id;
        selected_scores_data[low_offset] = item.score;
        low_offset++;

--- a/paddle/fluid/operators/math/beam_search.cu
+++ b/paddle/fluid/operators/math/beam_search.cu
@@ -157,10 +157,10 @@ __device__ __forceinline__ bool PruneEndBeams(Triple* top_beam_local,
 }
 __device__ __forceinline__ void WriteBack(
-    int64_t* selected_ids, float* selected_scores, size_t* selected_offsets,
+    int64_t* selected_ids, float* selected_scores, int* parent_idx,
-    Triple* top_beam_local, const int seq_offset_start,
+    size_t* selected_offsets, Triple* top_beam_local,
-    const int seq_offset_end, const int selected_seq_start,
+    const int seq_offset_start, const int seq_offset_end,
-    const int selected_seq_length) {
+    const int selected_seq_start, const int selected_seq_length) {
  const int tid = threadIdx.x;  // use 1 thread only for each sequence
  int global_index = selected_seq_start;
  for (int global_offset = seq_offset_start; global_offset < seq_offset_end;
@@ -171,6 +171,7 @@ __device__ __forceinline__ void WriteBack(
        selected_ids[global_index] =
            static_cast<int64_t>(top_beam_local[local_index].id);
        selected_scores[global_index] = top_beam_local[local_index].score;
+        parent_idx[global_index] = static_cast<int>(global_offset);
        global_index++;
      }
    }
@@ -180,11 +181,11 @@ __device__ __forceinline__ void WriteBack(
 template <int MaxLength, int MaxThreadsPerSeq, int MaxSeqs>
 __device__ void BeamSearchDetails(
-    int64_t* selected_ids, float* selected_scores, size_t* selected_offsets,
+    int64_t* selected_ids, float* selected_scores, int* parent_idx,
-    const int64_t* pre_ids, const float* pre_scores, const int64_t* ids,
+    size_t* selected_offsets, const int64_t* pre_ids, const float* pre_scores,
-    const float* scores, const int seq_offset_start, const int seq_offset_end,
+    const int64_t* ids, const float* scores, const int seq_offset_start,
-    const int seq_width, int beam_size, int end_id, bool is_accumulated,
+    const int seq_offset_end, const int seq_width, int beam_size, int end_id,
-    int num_used_threads) {
+    bool is_accumulated, int num_used_threads) {
  __shared__ Triple top_beam[MaxLength];
  int num_items = 0;
@@ -228,15 +229,15 @@ __device__ void BeamSearchDetails(
      selected_offsets[0] = 0;
    }
-    WriteBack(selected_ids, selected_scores, selected_offsets, top_beam_local,
+    WriteBack(selected_ids, selected_scores, parent_idx, selected_offsets,
-              seq_offset_start, seq_offset_end, selected_seq_start,
+              top_beam_local, seq_offset_start, seq_offset_end,
-              selected_seq_length);
+              selected_seq_start, selected_seq_length);
  }
 }
 template <int MaxLength, int MaxThreadsPerSeq, int MaxSeqs>
 __global__ void BeamSearchKernel(int64_t* selected_ids, float* selected_scores,
-                                 size_t* selected_offsets,
+                                 int* parent_idx, size_t* selected_offsets,
                                 const int64_t* pre_ids,
                                 const float* pre_scores, const int64_t* ids,
                                 const float* scores, const size_t* seq_offsets,
@@ -250,24 +251,25 @@ __global__ void BeamSearchKernel(int64_t* selected_ids, float* selected_scores,
  int seq_offset_end = static_cast<int>(seq_offsets[seq_id + 1]);
  BeamSearchDetails<MaxLength, MaxThreadsPerSeq, MaxSeqs>(
-      selected_ids, selected_scores, selected_offsets, pre_ids, pre_scores, ids,
+      selected_ids, selected_scores, parent_idx, selected_offsets, pre_ids,
-      scores, seq_offset_start, seq_offset_end, seq_width, beam_size, end_id,
+      pre_scores, ids, scores, seq_offset_start, seq_offset_end, seq_width,
-      is_accumulated, num_used_threads);
+      beam_size, end_id, is_accumulated, num_used_threads);
 }
 template <int MaxLength, int MaxThreadsPerSeq>
 __global__ void BeamSearchKernelSingle(
-    int64_t* selected_ids, float* selected_scores, size_t* selected_offsets,
+    int64_t* selected_ids, float* selected_scores, int* parent_idx,
-    const int64_t* pre_ids, const float* pre_scores, const int64_t* ids,
+    size_t* selected_offsets, const int64_t* pre_ids, const float* pre_scores,
-    const float* scores, const int seq_length, const int seq_width,
+    const int64_t* ids, const float* scores, const int seq_length,
-    int beam_size, int end_id, bool is_accumulated, int num_used_threads) {
+    const int seq_width, int beam_size, int end_id, bool is_accumulated,
+    int num_used_threads) {
  const int seq_offset_start = 0;
  const int seq_offset_end = seq_length;
  BeamSearchDetails<MaxLength, MaxThreadsPerSeq, 1>(
-      selected_ids, selected_scores, selected_offsets, pre_ids, pre_scores, ids,
+      selected_ids, selected_scores, parent_idx, selected_offsets, pre_ids,
-      scores, seq_offset_start, seq_offset_end, seq_width, beam_size, end_id,
+      pre_scores, ids, scores, seq_offset_start, seq_offset_end, seq_width,
-      is_accumulated, num_used_threads);
+      beam_size, end_id, is_accumulated, num_used_threads);
 }
 static inline int GetNumUsedThreads(const int max_threads_per_seq,
@@ -300,8 +302,9 @@ class BeamSearchFunctor<platform::CUDADeviceContext, T> {
                  const framework::LoDTensor* ids,
                  const framework::LoDTensor* scores,
                  framework::LoDTensor* selected_ids,
-                  framework::LoDTensor* selected_scores, size_t level,
+                  framework::LoDTensor* selected_scores,
-                  size_t beam_size, int end_id, bool is_accumulated) {
+                  framework::Tensor* parent_idx, size_t level, size_t beam_size,
+                  int end_id, bool is_accumulated) {
    auto abs_lod = framework::ToAbsOffset(scores->lod());
    const int64_t* pre_ids_data = pre_ids->data<int64_t>();
@@ -322,6 +325,8 @@ class BeamSearchFunctor<platform::CUDADeviceContext, T> {
        selected_ids->mutable_data<int64_t>(selected_dims, context.GetPlace());
    float* selected_scores_data =
        selected_scores->mutable_data<float>(selected_dims, context.GetPlace());
+    int* parent_idx_data = parent_idx->mutable_data<int>(
+        {static_cast<int64_t>(num_seqs * beam_size)}, context.GetPlace());
    framework::LoD selected_lod(2);
    selected_lod[0].assign(abs_lod[level].begin(), abs_lod[level].end());
@@ -339,9 +344,9 @@ class BeamSearchFunctor<platform::CUDADeviceContext, T> {
        CUDA_LAUNCH_KERNEL_HELPER(
            BeamSearchKernelSingle<kPowerOfTwoDim, kMaxThreadsPerSeq><<<
                1, kMaxThreadsPerSeq, 0, context.stream()>>>(
-                selected_ids_data, selected_scores_data, selected_offsets,
+                selected_ids_data, selected_scores_data, parent_idx_data,
-                pre_ids_data, pre_scores_data, ids_data, scores_data,
+                selected_offsets, pre_ids_data, pre_scores_data, ids_data,
-                seq_length, static_cast<int>(seq_width),
+                scores_data, seq_length, static_cast<int>(seq_width),
                static_cast<int>(beam_size), static_cast<int>(end_id),
                is_accumulated, num_used_threads));
      }
@@ -357,9 +362,9 @@ class BeamSearchFunctor<platform::CUDADeviceContext, T> {
        CUDA_LAUNCH_KERNEL_HELPER(
            BeamSearchKernel<kPowerOfTwoDim, kMaxThreadsPerSeq, kMaxSeqs><<<
                1, num_seqs * kMaxThreadsPerSeq, 0, context.stream()>>>(
-                selected_ids_data, selected_scores_data, selected_offsets,
+                selected_ids_data, selected_scores_data, parent_idx_data,
-                pre_ids_data, pre_scores_data, ids_data, scores_data,
+                selected_offsets, pre_ids_data, pre_scores_data, ids_data,
-                seq_offsets, static_cast<int>(num_seqs),
+                scores_data, seq_offsets, static_cast<int>(num_seqs),
                static_cast<int>(seq_width), static_cast<int>(beam_size),
                end_id, is_accumulated, num_used_threads));
      }
@@ -379,6 +384,7 @@ class BeamSearchFunctor<platform::CUDADeviceContext, T> {
          {static_cast<int64_t>(selected_lod[1].back()), 1});
      selected_ids->Resize(final_selected_dims);
      selected_scores->Resize(final_selected_dims);
+      parent_idx->Resize({static_cast<int64_t>(selected_lod[1].back())});
    }
  }
 };

--- a/paddle/fluid/operators/math/beam_search.h
+++ b/paddle/fluid/operators/math/beam_search.h
@@ -104,14 +104,12 @@ class BeamSearchFunctor {
   * Return false if all the input tensor is empty, in machine translation task
   * that means no candidates is provided, and the task will stop running.
   */
-  void operator()(const DeviceContext& context,
+  void operator()(
-                  const framework::LoDTensor* pre_ids,
+      const DeviceContext& context, const framework::LoDTensor* pre_ids,
-                  const framework::LoDTensor* pre_scores,
+      const framework::LoDTensor* pre_scores, const framework::LoDTensor* ids,
-                  const framework::LoDTensor* ids,
+      const framework::LoDTensor* scores, framework::LoDTensor* selected_ids,
-                  const framework::LoDTensor* scores,
+      framework::LoDTensor* selected_scores, framework::Tensor* parent_idx,
-                  framework::LoDTensor* selected_ids,
+      size_t level, size_t beam_size, int end_id, bool is_accumulated);
-                  framework::LoDTensor* selected_scores, size_t level,
-                  size_t beam_size, int end_id, bool is_accumulated);
 };
 }  // namespace math

--- a/paddle/fluid/operators/math/beam_search_test.cc
+++ b/paddle/fluid/operators/math/beam_search_test.cc
@@ -93,13 +93,14 @@ void TestBeamSearch() {
  paddle::framework::LoDTensor selected_ids;
  paddle::framework::LoDTensor selected_scores;
+  paddle::framework::LoDTensor parent_idx;
  size_t level = 0;
  size_t beam_size = 2;
  int end_id = 0;
  paddle::operators::math::BeamSearchFunctor<DeviceContext, float> beamsearch;
  beamsearch(*context, &pre_ids, &pre_scores, &ids, &scores, &selected_ids,
-             &selected_scores, level, beam_size, end_id, true);
+             &selected_scores, &parent_idx, level, beam_size, end_id, true);
  ASSERT_EQ(selected_ids.lod(), selected_scores.lod());

--- a/paddle/fluid/operators/activation_mkldnn_op.cc
+++ b/paddle/fluid/operators/activation_mkldnn_op.cc
--- a/paddle/fluid/operators/batch_norm_mkldnn_op.cc
+++ b/paddle/fluid/operators/batch_norm_mkldnn_op.cc
--- a/paddle/fluid/operators/concat_mkldnn_op.cc
+++ b/paddle/fluid/operators/concat_mkldnn_op.cc
--- a/paddle/fluid/operators/conv_mkldnn_op.cc
+++ b/paddle/fluid/operators/conv_mkldnn_op.cc
--- a/paddle/fluid/operators/conv_transpose_mkldnn_op.cc
+++ b/paddle/fluid/operators/conv_transpose_mkldnn_op.cc
--- a/paddle/fluid/operators/dequantize_mkldnn_op.cc
+++ b/paddle/fluid/operators/dequantize_mkldnn_op.cc
--- a/paddle/fluid/operators/fc_mkldnn_op.cc
+++ b/paddle/fluid/operators/fc_mkldnn_op.cc
--- a/paddle/fluid/operators/gaussian_random_mkldnn_op.cc
+++ b/paddle/fluid/operators/gaussian_random_mkldnn_op.cc
--- a/paddle/fluid/operators/lrn_mkldnn_op.cc
+++ b/paddle/fluid/operators/lrn_mkldnn_op.cc
--- a/paddle/fluid/operators/mkldnn_activation_op.h
+++ b/paddle/fluid/operators/mkldnn_activation_op.h
--- a/paddle/fluid/operators/pool_mkldnn_op.cc
+++ b/paddle/fluid/operators/pool_mkldnn_op.cc
--- a/paddle/fluid/operators/quantize_mkldnn_op.cc
+++ b/paddle/fluid/operators/quantize_mkldnn_op.cc
--- a/paddle/fluid/operators/softmax_mkldnn_op.cc
+++ b/paddle/fluid/operators/softmax_mkldnn_op.cc
--- a/paddle/fluid/operators/sum_mkldnn_op.cc
+++ b/paddle/fluid/operators/sum_mkldnn_op.cc
--- a/paddle/fluid/operators/transpose_mkldnn_op.cc
+++ b/paddle/fluid/operators/transpose_mkldnn_op.cc
--- a/paddle/fluid/operators/ngraph/ngraph_bridge.cc
+++ b/paddle/fluid/operators/ngraph/ngraph_bridge.cc
@@ -38,6 +38,8 @@ std::map<std::string,
        {"mean_grad", NG_OPS::BuildMeanGradNode},
        {"mul", NG_OPS::BuildMulNode},
        {"mul_grad", NG_OPS::BuildMulGradNode},
+        {"pool2d", NG_OPS::BuildPool2dNode},
+        {"pool2d_grad", NG_OPS::BuildPool2dGradNode},
        {"softmax", NG_OPS::BuildSoftmaxNode},
        {"softmax_grad", NG_OPS::BuildSoftmaxGradNode},
        {"scale", NG_OPS::BuildScaleNode},

--- a/paddle/fluid/operators/ngraph/ngraph_ops.h
+++ b/paddle/fluid/operators/ngraph/ngraph_ops.h
@@ -26,6 +26,7 @@ limitations under the License. */
 #include "ops/fill_constant_op.h"
 #include "ops/mean_op.h"
 #include "ops/mul_op.h"
+#include "ops/pool2d_op.h"
 #include "ops/scale_op.h"
 #include "ops/softmax_op.h"
 #include "ops/top_k_op.h"
--- a/paddle/fluid/operators/ngraph/ops/pool2d_op.h
+++ b/paddle/fluid/operators/ngraph/ops/pool2d_op.h
+/*Copyright (c) 2018 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
+    http://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. */
+#pragma once
+#include <string>
+#include <vector>
+#include "ngraph/ngraph.hpp"
+#include "paddle/fluid/platform/ngraph_helper.h"
+namespace paddle {
+namespace operators {
+namespace ngraphs {
+void BuildPool2dNode(
+    const std::shared_ptr<paddle::framework::OperatorBase>& op,
+    std::shared_ptr<
+        std::unordered_map<std::string, std::shared_ptr<ngraph::Node>>>
+        ngb_node_map) {
+  auto op_attrs = paddle::framework::AttrReader(op->Attrs());
+  auto x = paddle::platform::GetInputNode(op, "X", ngb_node_map);
+  auto x_shape = x->get_shape();
+  std::string pooling_type = op_attrs.Get<std::string>("pooling_type");
+  std::vector<int> ksize = op_attrs.Get<std::vector<int>>("ksize");
+  std::vector<int> strides = op_attrs.Get<std::vector<int>>("strides");
+  std::vector<int> paddings = op_attrs.Get<std::vector<int>>("paddings");
+  PADDLE_ENFORCE_EQ(x_shape.size() - 2, ksize.size(),
+                    "Handling 2d pooling only");
+  if (op_attrs.Get<bool>("global_pooling")) {
+    for (size_t i = 0; i < ksize.size(); ++i) {
+      paddings[i] = 0;
+      ksize[i] = static_cast<int>(x_shape.at(i + 2));
+    }
+  }
+  ngraph::Shape ng_padding_below{static_cast<size_t>(paddings.at(0)),
+                                 static_cast<size_t>(paddings.at(1))};
+  ngraph::Shape ng_padding_above{static_cast<size_t>(paddings.at(0)),
+                                 static_cast<size_t>(paddings.at(1))};
+  ngraph::Shape ng_ksize_shape{static_cast<size_t>(ksize.at(0)),
+                               static_cast<size_t>(ksize.at(1))};
+  ngraph::Strides ng_strides{static_cast<size_t>(strides.at(0)),
+                             static_cast<size_t>(strides.at(1))};
+  auto ComputeCeiledOutput = [](size_t in, size_t k, size_t p, size_t s) {
+    return (in - k + 2 * p) / s + 1;
+  };
+  if (op_attrs.Get<bool>("ceil_mode")) {
+    auto dummy_out = paddle::platform::GetOutputNode(op, "Out", ngb_node_map);
+    auto dummpy_shape = dummy_out->get_shape();
+    for (size_t i = 0; i < ng_padding_above.size(); ++i) {
+      auto desired_size = ComputeCeiledOutput(x_shape[i + 2], ksize[i],
+                                              paddings[i], strides[i]);
+      if (desired_size != dummpy_shape[i + 2]) {
+        ng_padding_above[i] += strides[i];
+      }
+    }
+  }
+  bool padding_exclusive = op_attrs.Get<bool>("exclusive");
+  if (pooling_type == "max") {
+    auto pool2d = std::make_shared<ngraph::op::MaxPool>(
+        x, ng_ksize_shape, ng_strides, ng_padding_below, ng_padding_above);
+    paddle::platform::SetOutputNode(op, "Out", pool2d, ngb_node_map);
+  } else if (pooling_type == "avg") {
+    std::shared_ptr<ngraph::Node> pool2d;
+    if (op_attrs.Get<bool>("adaptive")) {
+      auto ComputeAdaptive = [](size_t in, size_t k) {
+        return std::floor(in / k);
+      };
+      ng_strides[0] = x_shape.size() == 4
+                          ? ComputeAdaptive(x_shape[3], ksize[0])
+                          : ng_strides[0];
+      ng_strides[1] = x_shape.size() == 4
+                          ? ComputeAdaptive(x_shape[3], ksize[0])
+                          : ng_strides[1];
+      pool2d =
+          std::make_shared<ngraph::op::AvgPool>(x, ng_ksize_shape, ng_strides);
+    } else {
+      pool2d = std::make_shared<ngraph::op::AvgPool>(
+          x, ng_ksize_shape, ng_strides, ng_padding_below, ng_padding_above,
+          !padding_exclusive);
+    }
+    paddle::platform::SetOutputNode(op, "Out", pool2d, ngb_node_map);
+  } else {
+    PADDLE_THROW("Support max and avg pooling only");
+  }
+}
+void BuildPool2dGradNode(
+    const std::shared_ptr<paddle::framework::OperatorBase>& op,
+    std::shared_ptr<
+        std::unordered_map<std::string, std::shared_ptr<ngraph::Node>>>
+        ngb_node_map) {
+  auto op_attrs = paddle::framework::AttrReader(op->Attrs());
+  auto out = paddle::platform::GetInputNode(op, "Out", ngb_node_map);
+  auto dout = paddle::platform::GetInputNode(op, "Out@GRAD", ngb_node_map);
+  auto x = paddle::platform::GetInputNode(op, "X", ngb_node_map);
+  auto x_shape = x->get_shape();
+  std::string pooling_type = op_attrs.Get<std::string>("pooling_type");
+  std::vector<int> ksize = op_attrs.Get<std::vector<int>>("ksize");
+  std::vector<int> strides = op_attrs.Get<std::vector<int>>("strides");
+  std::vector<int> paddings = op_attrs.Get<std::vector<int>>("paddings");
+  PADDLE_ENFORCE_EQ(x_shape.size() - 2, ksize.size(),
+                    "Handling 2d pooling only");
+  if (op_attrs.Get<bool>("global_pooling")) {
+    for (size_t i = 0; i < ksize.size(); ++i) {
+      paddings[i] = 0;
+      ksize[i] = static_cast<int>(x_shape.at(i + 2));
+    }
+  }
+  ngraph::Shape ng_padding_below{static_cast<size_t>(paddings.at(0)),
+                                 static_cast<size_t>(paddings.at(1))};
+  ngraph::Shape ng_padding_above{static_cast<size_t>(paddings.at(0)),
+                                 static_cast<size_t>(paddings.at(1))};
+  ngraph::Shape ng_ksize_shape{static_cast<size_t>(ksize.at(0)),
+                               static_cast<size_t>(ksize.at(1))};
+  ngraph::Strides ng_strides{static_cast<size_t>(strides.at(0)),
+                             static_cast<size_t>(strides.at(1))};
+  bool padding_exclusive = op_attrs.Get<bool>("exclusive");
+  if (pooling_type == "max") {
+    auto pool2d_grad = std::make_shared<ngraph::op::MaxPoolBackprop>(
+        x, dout, out, ng_ksize_shape, ng_strides, ng_padding_below,
+        ng_padding_above);
+    paddle::platform::SetOutputNode(op, "X@GRAD", pool2d_grad, ngb_node_map);
+  } else if (pooling_type == "avg") {
+    std::shared_ptr<ngraph::Node> pool2d_grad;
+    if (op_attrs.Get<bool>("adaptive")) {
+      auto ComputeAdaptive = [](size_t in, size_t k) {
+        return std::floor(in / k);
+      };
+      ng_strides[0] = x_shape.size() == 4
+                          ? ComputeAdaptive(x_shape[3], ksize[0])
+                          : ng_strides[0];
+      ng_strides[1] = x_shape.size() == 4
+                          ? ComputeAdaptive(x_shape[3], ksize[0])
+                          : ng_strides[1];
+      pool2d_grad = std::make_shared<ngraph::op::AvgPoolBackprop>(
+          x->get_shape(), dout, ng_ksize_shape, ng_strides, ng_padding_below,
+          ng_padding_above, !padding_exclusive);
+    } else {
+      pool2d_grad = std::make_shared<ngraph::op::AvgPoolBackprop>(
+          x->get_shape(), dout, ng_ksize_shape, ng_strides, ng_padding_below,
+          ng_padding_above, !padding_exclusive);
+    }
+    paddle::platform::SetOutputNode(op, "X@GRAD", pool2d_grad, ngb_node_map);
+  } else {
+    PADDLE_THROW("Support max and avg pooling only");
+  }
+}
+}  // namespace ngraphs
+}  // namespace operators
+}  // namespace paddle
--- a/paddle/fluid/operators/reshape_op.cc
+++ b/paddle/fluid/operators/reshape_op.cc
@@ -330,6 +330,7 @@ class Reshape2GradOp : public framework::OperatorWithKernel {
 }  // namespace operators
 }  // namespace paddle
 namespace ops = paddle::operators;
+namespace plat = paddle::platform;
 REGISTER_OPERATOR(reshape, ops::ReshapeOp, ops::ReshapeOpMaker,
                  paddle::framework::DefaultGradOpDescMaker<true>);
@@ -356,16 +357,20 @@ REGISTER_OP_CPU_KERNEL_FUNCTOR(reshape2_grad, float, ops::ReshapeGradKernel,
 #ifdef PADDLE_WITH_CUDA
 REGISTER_OP_CUDA_KERNEL_FUNCTOR(reshape, float, ops::ReshapeKernel, double,
                                ops::ReshapeKernel, int, ops::ReshapeKernel,
-                                int64_t, ops::ReshapeKernel);
+                                int64_t, ops::ReshapeKernel, plat::float16,
+                                ops::ReshapeKernel);
 REGISTER_OP_CUDA_KERNEL_FUNCTOR(reshape_grad, float, ops::ReshapeGradKernel,
                                double, ops::ReshapeGradKernel, int,
                                ops::ReshapeGradKernel, int64_t,
+                                ops::ReshapeGradKernel, plat::float16,
                                ops::ReshapeGradKernel);
 REGISTER_OP_CUDA_KERNEL_FUNCTOR(reshape2, float, ops::ReshapeKernel, double,
                                ops::ReshapeKernel, int, ops::ReshapeKernel,
-                                int64_t, ops::ReshapeKernel);
+                                int64_t, ops::ReshapeKernel, plat::float16,
+                                ops::ReshapeKernel);
 REGISTER_OP_CUDA_KERNEL_FUNCTOR(reshape2_grad, float, ops::ReshapeGradKernel,
                                double, ops::ReshapeGradKernel, int,
                                ops::ReshapeGradKernel, int64_t,
+                                ops::ReshapeGradKernel, plat::float16,
                                ops::ReshapeGradKernel);
 #endif
--- a/paddle/fluid/operators/slice_op.cc
+++ b/paddle/fluid/operators/slice_op.cc
@@ -54,6 +54,9 @@ class SliceOp : public framework::OperatorWithKernel {
      out_dims[axes[i]] = end - start;
    }
    ctx->SetOutputDim("Out", out_dims);
+    if (axes[0] != 0) {
+      ctx->ShareLoD("Input", /*->*/ "Out");
+    }
  }
 protected:

--- a/paddle/fluid/operators/stack_op.cu
+++ b/paddle/fluid/operators/stack_op.cu
@@ -17,13 +17,16 @@
 namespace plat = paddle::platform;
 namespace ops = paddle::operators;
-REGISTER_OP_CUDA_KERNEL(stack, ops::StackKernel<plat::CUDADeviceContext, float>,
+REGISTER_OP_CUDA_KERNEL(
-                        ops::StackKernel<plat::CUDADeviceContext, double>,
+    stack, ops::StackKernel<plat::CUDADeviceContext, float>,
-                        ops::StackKernel<plat::CUDADeviceContext, int>,
+    ops::StackKernel<plat::CUDADeviceContext, double>,
-                        ops::StackKernel<plat::CUDADeviceContext, int64_t>);
+    ops::StackKernel<plat::CUDADeviceContext, int>,
+    ops::StackKernel<plat::CUDADeviceContext, int64_t>,
+    ops::StackKernel<plat::CUDADeviceContext, plat::float16>);
-REGISTER_OP_CUDA_KERNEL(stack_grad,
+REGISTER_OP_CUDA_KERNEL(
-                        ops::StackGradKernel<plat::CUDADeviceContext, float>,
+    stack_grad, ops::StackGradKernel<plat::CUDADeviceContext, float>,
-                        ops::StackGradKernel<plat::CUDADeviceContext, double>,
+    ops::StackGradKernel<plat::CUDADeviceContext, double>,
-                        ops::StackGradKernel<plat::CUDADeviceContext, int>,
+    ops::StackGradKernel<plat::CUDADeviceContext, int>,
-                        ops::StackGradKernel<plat::CUDADeviceContext, int64_t>);
+    ops::StackGradKernel<plat::CUDADeviceContext, int64_t>,
+    ops::StackGradKernel<plat::CUDADeviceContext, plat::float16>);
--- a/paddle/fluid/operators/transpose_op.cu.cc
+++ b/paddle/fluid/operators/transpose_op.cu.cc
@@ -15,19 +15,27 @@ limitations under the License. */
 #include "paddle/fluid/operators/transpose_op.h"
 namespace ops = paddle::operators;
+namespace plat = paddle::platform;
 REGISTER_OP_CUDA_KERNEL(
    transpose, ops::TransposeKernel<paddle::platform::CUDADeviceContext, float>,
-    ops::TransposeKernel<paddle::platform::CUDADeviceContext, double>);
+    ops::TransposeKernel<paddle::platform::CUDADeviceContext, double>,
+    ops::TransposeKernel<paddle::platform::CUDADeviceContext, plat::float16>);
 REGISTER_OP_CUDA_KERNEL(
    transpose_grad,
    ops::TransposeGradKernel<paddle::platform::CUDADeviceContext, float>,
-    ops::TransposeGradKernel<paddle::platform::CUDADeviceContext, double>);
+    ops::TransposeGradKernel<paddle::platform::CUDADeviceContext, double>,
+    ops::TransposeGradKernel<paddle::platform::CUDADeviceContext,
+                             plat::float16>);
 REGISTER_OP_CUDA_KERNEL(
    transpose2,
    ops::TransposeKernel<paddle::platform::CUDADeviceContext, float>,
-    ops::TransposeKernel<paddle::platform::CUDADeviceContext, double>);
+    ops::TransposeKernel<paddle::platform::CUDADeviceContext, double>,
+    ops::TransposeKernel<paddle::platform::CUDADeviceContext, plat::float16>);
 REGISTER_OP_CUDA_KERNEL(
    transpose2_grad,
    ops::TransposeGradKernel<paddle::platform::CUDADeviceContext, float>,
-    ops::TransposeGradKernel<paddle::platform::CUDADeviceContext, double>);
+    ops::TransposeGradKernel<paddle::platform::CUDADeviceContext, double>,
+    ops::TransposeGradKernel<paddle::platform::CUDADeviceContext,
+                             plat::float16>);
--- a/paddle/fluid/operators/yolov3_loss_op.h
+++ b/paddle/fluid/operators/yolov3_loss_op.h
--- a/python/paddle/fluid/contrib/__init__.py
+++ b/python/paddle/fluid/contrib/__init__.py
@@ -22,6 +22,8 @@ from . import op_frequence
 from .op_frequence import *
 from . import quantize
 from .quantize import *
+from . import int8_inference
+from .int8_inference import *
 from . import reader
 from .reader import *
 from . import slim
@@ -34,6 +36,7 @@ __all__ += decoder.__all__
 __all__ += memory_usage_calc.__all__
 __all__ += op_frequence.__all__
 __all__ += quantize.__all__
+__all__ += int8_inference.__all__
 __all__ += reader.__all__
 __all__ += slim.__all__
 __all__ += utils.__all__
--- a/python/paddle/fluid/contrib/int8_inference/__init__.py
+++ b/python/paddle/fluid/contrib/int8_inference/__init__.py
@@ -11,3 +11,10 @@
 # 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.
+from __future__ import print_function
+from . import utility
+from .utility import *
+__all__ = utility.__all__
--- a/python/paddle/fluid/contrib/int8_inference/utility.py
+++ b/python/paddle/fluid/contrib/int8_inference/utility.py
@@ -11,11 +11,15 @@
 # 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.
-import paddle.fluid.core as core
+from paddle.fluid import core
 import numpy as np
 import math
 import os
-import paddle.fluid as fluid
+from paddle.fluid.executor import global_scope
+from paddle.fluid import io
+__all__ = ['Calibrator']
 class Calibrator(object):
@@ -76,8 +80,7 @@ class Calibrator(object):
        '''
        for i in self.sampling_program.list_vars():
            if i.name in self.sampling_vars:
-                np_data = np.array(fluid.global_scope().find_var(i.name)
+                np_data = np.array(global_scope().find_var(i.name).get_tensor())
-                                   .get_tensor())
                if i.name not in self._sampling_data:
                    self._sampling_data[i.name] = []
                self._sampling_data[i.name].append(np_data)
@@ -86,9 +89,9 @@ class Calibrator(object):
        '''
        Save the quantized model to the disk.
        '''
-        fluid.io.save_inference_model(self.output, self.feed_var_names,
+        io.save_inference_model(self.output, self.feed_var_names,
-                                      self.fetch_list, self.exe,
+                                self.fetch_list, self.exe,
-                                      self.sampling_program)
+                                self.sampling_program)
    def __display_debug(self):
        if self.debug:

--- a/python/paddle/fluid/contrib/tests/test_calibration.py
+++ b/python/paddle/fluid/contrib/tests/test_calibration.py
@@ -19,15 +19,12 @@ import sys
 import random
 import paddle
 import paddle.fluid as fluid
-import argparse
 import functools
 import contextlib
-import paddle.fluid.profiler as profiler
 from paddle.dataset.common import download
 from PIL import Image, ImageEnhance
 import math
-sys.path.append('..')
+import paddle.fluid.contrib.int8_inference.utility as int8_utility
-import int8_inference.utility as int8_utility
 random.seed(0)
 np.random.seed(0)
@@ -43,7 +40,7 @@ img_mean = np.array([0.485, 0.456, 0.406]).reshape((3, 1, 1))
 img_std = np.array([0.229, 0.224, 0.225]).reshape((3, 1, 1))
-# TODO(guomingz): Remove duplicated code from line 45 ~ line 114
+# TODO(guomingz): Remove duplicated code from resize_short, crop_image, process_image, _reader_creator
 def resize_short(img, target_size):
    percent = float(target_size) / min(img.size[0], img.size[1])
    resized_width = int(round(img.size[0] * percent))
@@ -123,16 +120,37 @@ class TestCalibrationForResnet50(unittest.TestCase):
        self.cache_folder = os.path.expanduser('~/.cache/paddle/dataset/' +
                                               self.int8_download)
-        data_url = 'http://paddle-inference-dist.cdn.bcebos.com/int8/calibration_test_data.tar.gz'
+        data_urls = []
-        data_md5 = '1b6c1c434172cca1bf9ba1e4d7a3157d'
+        data_md5s = []
-        self.data_cache_folder = self.download_data(data_url, data_md5, "data")
+        self.data_cache_folder = ''
+        if os.environ.get('DATASET') == 'full':
+            data_urls.append(
+                'https://paddle-inference-dist.bj.bcebos.com/int8/ILSVRC2012_img_val.tar.gz.partaa'
+            )
+            data_md5s.append('60f6525b0e1d127f345641d75d41f0a8')
+            data_urls.append(
+                'https://paddle-inference-dist.bj.bcebos.com/int8/ILSVRC2012_img_val.tar.gz.partab'
+            )
+            data_md5s.append('1e9f15f64e015e58d6f9ec3210ed18b5')
+            self.data_cache_folder = self.download_data(data_urls, data_md5s,
+                                                        "full_data", False)
+        else:
+            data_urls.append(
+                'http://paddle-inference-dist.cdn.bcebos.com/int8/calibration_test_data.tar.gz'
+            )
+            data_md5s.append('1b6c1c434172cca1bf9ba1e4d7a3157d')
+            self.data_cache_folder = self.download_data(data_urls, data_md5s,
+                                                        "small_data", False)
        # reader/decorator.py requires the relative path to the data folder
        cmd = 'rm -rf {0} && ln -s {1} {0}'.format("data",
                                                   self.data_cache_folder)
        os.system(cmd)
-        self.iterations = 50
+        self.batch_size = 1
+        self.sample_iterations = 50
+        self.infer_iterations = 50000 if os.environ.get(
+            'DATASET') == 'full' else 50
    def cache_unzipping(self, target_folder, zip_path):
        if not os.path.exists(target_folder):
@@ -140,20 +158,44 @@ class TestCalibrationForResnet50(unittest.TestCase):
                                                          zip_path)
            os.system(cmd)
-    def download_data(self, data_url, data_md5, folder_name):
+    def download_data(self, data_urls, data_md5s, folder_name, is_model=True):
-        download(data_url, self.int8_download, data_md5)
        data_cache_folder = os.path.join(self.cache_folder, folder_name)
-        file_name = data_url.split('/')[-1]
+        zip_path = ''
-        zip_path = os.path.join(self.cache_folder, file_name)
+        if os.environ.get('DATASET') == 'full':
+            file_names = []
+            for i in range(0, len(data_urls)):
+                download(data_urls[i], self.int8_download, data_md5s[i])
+                file_names.append(data_urls[i].split('/')[-1])
+            zip_path = os.path.join(self.cache_folder,
+                                    'full_imagenet_val.tar.gz')
+            if not os.path.exists(zip_path):
+                cat_command = 'cat'
+                for file_name in file_names:
+                    cat_command += ' ' + os.path.join(self.cache_folder,
+                                                      file_name)
+                cat_command += ' > ' + zip_path
+                os.system(cat_command)
+        if os.environ.get('DATASET') != 'full' or is_model:
+            download(data_urls[0], self.int8_download, data_md5s[0])
+            file_name = data_urls[0].split('/')[-1]
+            zip_path = os.path.join(self.cache_folder, file_name)
+        print('Data is downloaded at {0}').format(zip_path)
        self.cache_unzipping(data_cache_folder, zip_path)
        return data_cache_folder
-    def download_resnet50_model(self):
+    def download_model(self):
        # resnet50 fp32 data
-        data_url = 'http://paddle-inference-dist.cdn.bcebos.com/int8/resnet50_int8_model.tar.gz'
+        data_urls = [
-        data_md5 = '4a5194524823d9b76da6e738e1367881'
+            'http://paddle-inference-dist.cdn.bcebos.com/int8/resnet50_int8_model.tar.gz'
-        self.model_cache_folder = self.download_data(data_url, data_md5,
+        ]
+        data_md5s = ['4a5194524823d9b76da6e738e1367881']
+        self.model_cache_folder = self.download_data(data_urls, data_md5s,
                                                     "resnet50_fp32")
+        self.model = "ResNet-50"
+        self.algo = "direct"
    def run_program(self, model_path, generate_int8=False, algo='direct'):
        image_shape = [3, 224, 224]
@@ -169,17 +211,17 @@ class TestCalibrationForResnet50(unittest.TestCase):
        t = fluid.transpiler.InferenceTranspiler()
        t.transpile(infer_program, fluid.CPUPlace())
-        val_reader = paddle.batch(val(), batch_size=1)
+        val_reader = paddle.batch(val(), self.batch_size)
+        iterations = self.infer_iterations
        if generate_int8:
            int8_model = os.path.join(os.getcwd(), "calibration_out")
+            iterations = self.sample_iterations
            if os.path.exists(int8_model):
                os.system("rm -rf " + int8_model)
                os.system("mkdir " + int8_model)
-            print("Start calibration ...")
            calibrator = int8_utility.Calibrator(
                program=infer_program,
                pretrained_model=model_path,
@@ -191,6 +233,7 @@ class TestCalibrationForResnet50(unittest.TestCase):
        test_info = []
        cnt = 0
+        periods = []
        for batch_id, data in enumerate(val_reader()):
            image = np.array(
                [x[0].reshape(image_shape) for x in data]).astype("float32")
@@ -202,21 +245,28 @@ class TestCalibrationForResnet50(unittest.TestCase):
                if op.has_attr("use_mkldnn"):
                    op._set_attr("use_mkldnn", True)
+            t1 = time.time()
            _, acc1, _ = exe.run(
                running_program,
                feed={feed_dict[0]: image,
                      feed_dict[1]: label},
                fetch_list=fetch_targets)
+            t2 = time.time()
+            period = t2 - t1
+            periods.append(period)
            if generate_int8:
                calibrator.sample_data()
            test_info.append(np.mean(acc1) * len(data))
            cnt += len(data)
-            if batch_id != self.iterations - 1:
+            if (batch_id + 1) % 100 == 0:
-                continue
+                print("{0} images,".format(batch_id + 1))
+                sys.stdout.flush()
-            break
+            if (batch_id + 1) == iterations:
+                break
        if generate_int8:
            calibrator.save_int8_model()
@@ -225,32 +275,49 @@ class TestCalibrationForResnet50(unittest.TestCase):
                "Calibration is done and the corresponding files are generated at {}".
                format(os.path.abspath("calibration_out")))
        else:
-            return np.sum(test_info) / cnt
+            throughput = cnt / np.sum(periods)
+            latency = np.average(periods)
+            acc1 = np.sum(test_info) / cnt
+            return (throughput, latency, acc1)
    def test_calibration(self):
-        self.download_resnet50_model()
+        self.download_model()
-        fp32_acc1 = self.run_program(self.model_cache_folder + "/model")
+        print("Start FP32 inference for {0} on {1} images ...").format(
-        self.run_program(self.model_cache_folder + "/model", True)
+            self.model, self.infer_iterations)
-        int8_acc1 = self.run_program("calibration_out")
+        (fp32_throughput, fp32_latency,
+         fp32_acc1) = self.run_program(self.model_cache_folder + "/model")
+        print("Start INT8 calibration for {0} on {1} images ...").format(
+            self.model, self.sample_iterations)
+        self.run_program(
+            self.model_cache_folder + "/model", True, algo=self.algo)
+        print("Start INT8 inference for {0} on {1} images ...").format(
+            self.model, self.infer_iterations)
+        (int8_throughput, int8_latency,
+         int8_acc1) = self.run_program("calibration_out")
        delta_value = np.abs(fp32_acc1 - int8_acc1)
        self.assertLess(delta_value, 0.01)
+        print(
+            "FP32 {0}: batch_size {1}, throughput {2} images/second, latency {3} second, accuracy {4}".
+            format(self.model, self.batch_size, fp32_throughput, fp32_latency,
+                   fp32_acc1))
+        print(
+            "INT8 {0}: batch_size {1}, throughput {2} images/second, latency {3} second, accuracy {4}".
+            format(self.model, self.batch_size, int8_throughput, int8_latency,
+                   int8_acc1))
+        sys.stdout.flush()
 class TestCalibrationForMobilenetv1(TestCalibrationForResnet50):
-    def download_mobilenetv1_model(self):
+    def download_model(self):
        # mobilenetv1 fp32 data
-        data_url = 'http://paddle-inference-dist.cdn.bcebos.com/int8/mobilenetv1_int8_model.tar.gz'
+        data_urls = [
-        data_md5 = '13892b0716d26443a8cdea15b3c6438b'
+            'http://paddle-inference-dist.cdn.bcebos.com/int8/mobilenetv1_int8_model.tar.gz'
-        self.model_cache_folder = self.download_data(data_url, data_md5,
+        ]
+        data_md5s = ['13892b0716d26443a8cdea15b3c6438b']
+        self.model_cache_folder = self.download_data(data_urls, data_md5s,
                                                     "mobilenetv1_fp32")
+        self.model = "MobileNet-V1"
-    def test_calibration(self):
+        self.algo = "KL"
-        self.download_mobilenetv1_model()
-        fp32_acc1 = self.run_program(self.model_cache_folder + "/model")
-        self.run_program(self.model_cache_folder + "/model", True, algo='KL')
-        int8_acc1 = self.run_program("calibration_out")
-        delta_value = np.abs(fp32_acc1 - int8_acc1)
-        self.assertLess(delta_value, 0.01)
 if __name__ == '__main__':

--- a/python/paddle/fluid/initializer.py
+++ b/python/paddle/fluid/initializer.py
@@ -366,17 +366,40 @@ class TruncatedNormalInitializer(Initializer):
        # Initialization Ops should be prepended and not appended
        if self._seed == 0:
            self._seed = block.program.random_seed
+        # to be compatible of fp16 initalizers
+        if var.dtype == VarDesc.VarType.FP16:
+            out_dtype = VarDesc.VarType.FP32
+            out_var = block.create_var(
+                name=unique_name.generate(".".join(
+                    ['truncated_gaussian_random', 'tmp'])),
+                shape=var.shape,
+                dtype=out_dtype,
+                type=VarDesc.VarType.LOD_TENSOR,
+                persistable=False)
+        else:
+            out_dtype = var.dtype
+            out_var = var
        op = block._prepend_op(
            type="truncated_gaussian_random",
-            outputs={"Out": var},
+            outputs={"Out": out_var},
            attrs={
                "shape": var.shape,
-                "dtype": int(var.dtype),
+                "dtype": out_dtype,
                "mean": self._mean,
                "std": self._std_dev,
                "seed": self._seed
            },
            stop_gradient=True)
+        if var.dtype == VarDesc.VarType.FP16:
+            block.append_op(
+                type="cast",
+                inputs={"X": out_var},
+                outputs={"Out": var},
+                attrs={"in_dtype": out_var.dtype,
+                       "out_dtype": var.dtype})
        var.op = op
        return op

--- a/python/paddle/fluid/layers/detection.py
+++ b/python/paddle/fluid/layers/detection.py
@@ -347,19 +347,107 @@ def box_coder(prior_box,
              target_box,
              code_type="encode_center_size",
              box_normalized=True,
-              name=None):
+              name=None,
+              axis=0):
    """
-    ${comment}
+    **Box Coder Layer**
+    Encode/Decode the target bounding box with the priorbox information.
+    The Encoding schema described below:
+    .. math::
+        ox = (tx - px) / pw / pxv
+        oy = (ty - py) / ph / pyv
+        ow = \log(\abs(tw / pw)) / pwv 
+        oh = \log(\abs(th / ph)) / phv 
+    The Decoding schema described below:
+    .. math::
+        ox = (pw * pxv * tx * + px) - tw / 2
+        oy = (ph * pyv * ty * + py) - th / 2
+        ow = \exp(pwv * tw) * pw + tw / 2
+        oh = \exp(phv * th) * ph + th / 2   
+    where `tx`, `ty`, `tw`, `th` denote the target box's center coordinates, 
+    width and height respectively. Similarly, `px`, `py`, `pw`, `ph` denote 
+    the priorbox's (anchor) center coordinates, width and height. `pxv`, 
+    `pyv`, `pwv`, `phv` denote the variance of the priorbox and `ox`, `oy`, 
+    `ow`, `oh` denote the encoded/decoded coordinates, width and height. 
+    During Box Decoding, two modes for broadcast are supported. Say target 
+    box has shape [N, M, 4], and the shape of prior box can be [N, 4] or 
+    [M, 4]. Then prior box will broadcast to target box along the 
+    assigned axis. 
    Args:
-        prior_box(${prior_box_type}): ${prior_box_comment}
+        prior_box(Variable): Box list prior_box is a 2-D Tensor with shape 
-        prior_box_var(${prior_box_var_type}): ${prior_box_var_comment}
+                             [M, 4] holds M boxes, each box is represented as
-        target_box(${target_box_type}): ${target_box_comment}
+                             [xmin, ymin, xmax, ymax], [xmin, ymin] is the 
-        code_type(${code_type_type}): ${code_type_comment}
+                             left top coordinate of the anchor box, if the 
-        box_normalized(${box_normalized_type}): ${box_normalized_comment}
+                             input is image feature map, they are close to 
+                             the origin of the coordinate system. [xmax, ymax]
+                             is the right bottom coordinate of the anchor box.       
+        prior_box_var(Variable|list): prior_box_var supports two types of input. 
+                              One is variable with shape [M, 4] holds M group.
+                              The other one is list consist of 4 elements 
+                              shared by all boxes. 
+        target_box(Variable): This input can be a 2-D LoDTensor with shape 
+                              [N, 4] when code_type is 'encode_center_size'. 
+                              This input also can be a 3-D Tensor with shape 
+                              [N, M, 4] when code_type is 'decode_center_size'. 
+                              Each box is represented as  
+                              [xmin, ymin, xmax, ymax]. This tensor can 
+                              contain LoD information to represent a batch 
+                              of inputs. 
+        code_type(string): The code type used with the target box. It can be
+                           encode_center_size or decode_center_size
+        box_normalized(int): Whether treat the priorbox as a noramlized box.
+                             Set true by default.
+        name(string): The name of box coder.
+        axis(int): Which axis in PriorBox to broadcast for box decode, 
+                   for example, if axis is 0 and TargetBox has shape
+                   [N, M, 4] and PriorBox has shape [M, 4], then PriorBox
+                   will broadcast to [N, M, 4] for decoding. It is only valid
+                   when code type is decode_center_size. Set 0 by default. 
    Returns:
-        output_box(${output_box_type}): ${output_box_comment}
+        output_box(Variable): When code_type is 'encode_center_size', the 
+                              output tensor of box_coder_op with shape 
+                              [N, M, 4] representing the result of N target 
+                              boxes encoded with M Prior boxes and variances. 
+                              When code_type is 'decode_center_size', 
+                              N represents the batch size and M represents 
+                              the number of deocded boxes.
+    Examples:
+        .. code-block:: python
+            prior_box = fluid.layers.data(name='prior_box', 
+                                          shape=[512, 4], 
+                                          dtype='float32',
+                                          append_batch_size=False)
+            target_box = fluid.layers.data(name='target_box',
+                                           shape=[512,81,4],
+                                           dtype='float32',
+                                           append_batch_size=False)
+            output = fluid.layers.box_coder(prior_box=prior_box,
+                                            prior_box_var=[0.1,0.1,0.2,0.2],
+                                            target_box=target_box,
+                                            code_type="decode_center_size",
+                                            box_normalized=False,
+                                            axis=1)
    """
    helper = LayerHelper("box_coder", **locals())
@@ -370,15 +458,22 @@ def box_coder(prior_box,
        output_box = helper.create_variable(
            name=name, dtype=prior_box.dtype, persistable=False)
+    inputs = {"PriorBox": prior_box, "TargetBox": target_box}
+    attrs = {
+        "code_type": code_type,
+        "box_normalized": box_normalized,
+        "axis": axis
+    }
+    if isinstance(prior_box_var, Variable):
+        inputs['PriorBoxVar'] = prior_box_var
+    elif isinstance(prior_box_var, list):
+        attrs['variance'] = prior_box_var
+    else:
+        raise TypeError("Input variance of box_coder must be Variable or lisz")
    helper.append_op(
        type="box_coder",
-        inputs={
+        inputs=inputs,
-            "PriorBox": prior_box,
+        attrs=attrs,
-            "PriorBoxVar": prior_box_var,
-            "TargetBox": target_box
-        },
-        attrs={"code_type": code_type,
-               "box_normalized": box_normalized},
        outputs={"OutputBox": output_box})
    return output_box
@@ -414,13 +509,10 @@ def yolov3_loss(x,
                gtbox,
                gtlabel,
                anchors,
+                anchor_mask,
                class_num,
                ignore_thresh,
-                loss_weight_xy=None,
+                downsample_ratio,
-                loss_weight_wh=None,
-                loss_weight_conf_target=None,
-                loss_weight_conf_notarget=None,
-                loss_weight_class=None,
                name=None):
    """
    ${comment}
@@ -432,16 +524,13 @@ def yolov3_loss(x,
                          and x, y, w, h should be relative value of input image.
                          N is the batch number and B is the max box number in 
                          an image.
-        gtlabel (Variable): class id of ground truth boxes, shoud be ins shape
+        gtlabel (Variable): class id of ground truth boxes, shoud be in shape
                            of [N, B].
        anchors (list|tuple): ${anchors_comment}
+        anchor_mask (list|tuple): ${anchor_mask_comment}
        class_num (int): ${class_num_comment}
        ignore_thresh (float): ${ignore_thresh_comment}
-        loss_weight_xy (float|None): ${loss_weight_xy_comment}
+        downsample_ratio (int): ${downsample_ratio_comment}
-        loss_weight_wh (float|None): ${loss_weight_wh_comment}
-        loss_weight_conf_target (float|None): ${loss_weight_conf_target_comment}
-        loss_weight_conf_notarget (float|None): ${loss_weight_conf_notarget_comment}
-        loss_weight_class (float|None): ${loss_weight_class_comment}
        name (string): the name of yolov3 loss
    Returns:
@@ -461,9 +550,10 @@ def yolov3_loss(x,
        x = fluid.layers.data(name='x', shape=[255, 13, 13], dtype='float32')
        gtbox = fluid.layers.data(name='gtbox', shape=[6, 5], dtype='float32')
        gtlabel = fluid.layers.data(name='gtlabel', shape=[6, 1], dtype='int32')
-        anchors = [10, 13, 16, 30, 33, 23]
+        anchors = [10, 13, 16, 30, 33, 23, 30, 61, 62, 45, 59, 119, 116, 90, 156, 198, 373, 326]
-        loss = fluid.layers.yolov3_loss(x=x, gtbox=gtbox, class_num=80
+        anchors = [0, 1, 2]
-                                        anchors=anchors, ignore_thresh=0.5)
+        loss = fluid.layers.yolov3_loss(x=x, gtbox=gtbox, class_num=80, anchors=anchors, 
+                                        ignore_thresh=0.5, downsample_ratio=32)
    """
    helper = LayerHelper('yolov3_loss', **locals())
@@ -475,6 +565,8 @@ def yolov3_loss(x,
        raise TypeError("Input gtlabel of yolov3_loss must be Variable")
    if not isinstance(anchors, list) and not isinstance(anchors, tuple):
        raise TypeError("Attr anchors of yolov3_loss must be list or tuple")
+    if not isinstance(anchor_mask, list) and not isinstance(anchor_mask, tuple):
+        raise TypeError("Attr anchor_mask of yolov3_loss must be list or tuple")
    if not isinstance(class_num, int):
        raise TypeError("Attr class_num of yolov3_loss must be an integer")
    if not isinstance(ignore_thresh, float):
@@ -487,31 +579,29 @@ def yolov3_loss(x,
        loss = helper.create_variable(
            name=name, dtype=x.dtype, persistable=False)
+    objectness_mask = helper.create_variable_for_type_inference(dtype='int32')
+    gt_match_mask = helper.create_variable_for_type_inference(dtype='int32')
    attrs = {
        "anchors": anchors,
+        "anchor_mask": anchor_mask,
        "class_num": class_num,
        "ignore_thresh": ignore_thresh,
+        "downsample_ratio": downsample_ratio,
    }
-    if loss_weight_xy is not None and isinstance(loss_weight_xy, float):
-        self.attrs['loss_weight_xy'] = loss_weight_xy
-    if loss_weight_wh is not None and isinstance(loss_weight_wh, float):
-        self.attrs['loss_weight_wh'] = loss_weight_wh
-    if loss_weight_conf_target is not None and isinstance(
-            loss_weight_conf_target, float):
-        self.attrs['loss_weight_conf_target'] = loss_weight_conf_target
-    if loss_weight_conf_notarget is not None and isinstance(
-            loss_weight_conf_notarget, float):
-        self.attrs['loss_weight_conf_notarget'] = loss_weight_conf_notarget
-    if loss_weight_class is not None and isinstance(loss_weight_class, float):
-        self.attrs['loss_weight_class'] = loss_weight_class
    helper.append_op(
        type='yolov3_loss',
-        inputs={"X": x,
+        inputs={
-                "GTBox": gtbox,
+            "X": x,
-                "GTLabel": gtlabel},
+            "GTBox": gtbox,
-        outputs={'Loss': loss},
+            "GTLabel": gtlabel,
+        },
+        outputs={
+            'Loss': loss,
+            'ObjectnessMask': objectness_mask,
+            'GTMatchMask': gt_match_mask
+        },
        attrs=attrs)
    return loss

--- a/python/paddle/fluid/layers/nn.py
+++ b/python/paddle/fluid/layers/nn.py
@@ -3877,7 +3877,8 @@ def beam_search(pre_ids,
                end_id,
                level=0,
                is_accumulated=True,
-                name=None):
+                name=None,
+                return_parent_idx=False):
    """
    Beam search is a classical algorithm for selecting candidate words in a
    machine translation task.
@@ -3933,10 +3934,16 @@ def beam_search(pre_ids,
             accumulated scores.
        name(str|None): A name for this layer(optional). If set None, the layer
                        will be named automatically.
+        return_parent_idx(bool): Whether to return an extra Tensor variable 
+                        preserving the selected_ids' parent indice in pre_ids
+                        in output, which can be used to gather cell states at
+                        the next time step.
    Returns:
-        Variable: The LodTensor pair containing the selected ids and the \
+        Variable: The LodTensor tuple containing the selected ids and the \
-            corresponding scores.
+            corresponding scores. If :attr:`return_parent_idx` is :attr:`True`, \
+            an extra Tensor variable preserving the selected_ids' parent indice \
+            is included.
    Examples:
        .. code-block:: python
@@ -3969,6 +3976,11 @@ def beam_search(pre_ids,
    selected_scores = helper.create_variable_for_type_inference(
        dtype=score_type)
    selected_ids = helper.create_variable_for_type_inference(dtype=id_type)
+    # parent_idx is a tensor used to gather cell states at the next time
+    # step. Though lod in selected_ids can also be used to gather by
+    # sequence_expand, it is not efficient.
+    # gather_op's index input only supports int32 dtype currently
+    parent_idx = helper.create_variable_for_type_inference(dtype="int32")
    helper.append_op(
        type='beam_search',
@@ -3976,6 +3988,7 @@ def beam_search(pre_ids,
        outputs={
            'selected_ids': selected_ids,
            'selected_scores': selected_scores,
+            'parent_idx': parent_idx
        },
        attrs={
            # TODO(ChunweiYan) to assure other value support
@@ -3984,8 +3997,10 @@ def beam_search(pre_ids,
            'end_id': end_id,
            'is_accumulated': is_accumulated,
        })
+    if return_parent_idx:
-    return selected_ids, selected_scores
+        return selected_ids, selected_scores, parent_idx
+    else:
+        return selected_ids, selected_scores
 def beam_search_decode(ids, scores, beam_size, end_id, name=None):

--- a/python/paddle/fluid/layers/ops.py
+++ b/python/paddle/fluid/layers/ops.py
@@ -135,7 +135,7 @@ def thresholded_relu(x, threshold=None):
        if val is not None:
            kwargs[name] = val
-    _thresholded_relu_(**kwargs)
+    return _thresholded_relu_(**kwargs)
 thresholded_relu.__doc__ = _thresholded_relu_.__doc__ + """

--- a/python/paddle/fluid/tests/test_detection.py
+++ b/python/paddle/fluid/tests/test_detection.py
@@ -50,6 +50,19 @@ class TestDetection(unittest.TestCase):
            self.assertEqual(out.shape[-1], 6)
        print(str(program))
+    def test_box_coder_api(self):
+        program = Program()
+        with program_guard(program):
+            x = layers.data(name='x', shape=[4], dtype='float32')
+            y = layers.data(name='z', shape=[4], dtype='float32', lod_level=1)
+            bcoder = layers.box_coder(
+                prior_box=x,
+                prior_box_var=[0.1, 0.2, 0.1, 0.2],
+                target_box=y,
+                code_type='encode_center_size')
+            self.assertIsNotNone(bcoder)
+        print(str(program))
    def test_detection_api(self):
        program = Program()
        with program_guard(program):
@@ -463,8 +476,8 @@ class TestYoloDetection(unittest.TestCase):
            x = layers.data(name='x', shape=[30, 7, 7], dtype='float32')
            gtbox = layers.data(name='gtbox', shape=[10, 4], dtype='float32')
            gtlabel = layers.data(name='gtlabel', shape=[10], dtype='int32')
-            loss = layers.yolov3_loss(x, gtbox, gtlabel, [10, 13, 30, 13], 10,
+            loss = layers.yolov3_loss(x, gtbox, gtlabel, [10, 13, 30, 13],
-                                      0.5)
+                                      [0, 1], 10, 0.7, 32)
            self.assertIsNotNone(loss)

--- a/python/paddle/fluid/tests/unittests/CMakeLists.txt
+++ b/python/paddle/fluid/tests/unittests/CMakeLists.txt
 file(GLOB TEST_OPS RELATIVE "${CMAKE_CURRENT_SOURCE_DIR}" "test_*.py")
 string(REPLACE ".py" "" TEST_OPS "${TEST_OPS}")
-# The MKLDNN tests are skiped when the MKLDNN flag is OFF
-if(NOT WITH_MKLDNN)
-    foreach(src ${TEST_OPS})
-        if(${src} MATCHES ".*_mkldnn_op$")
-            list(REMOVE_ITEM TEST_OPS ${src})
-        endif()
-    endforeach()
-endif(NOT WITH_MKLDNN)
 if(NOT WITH_DISTRIBUTE)
    list(REMOVE_ITEM TEST_OPS test_recv_op)
    list(REMOVE_ITEM TEST_OPS test_dist_transpiler)
@@ -123,3 +114,7 @@ endif()
 if (WITH_NGRAPH)
    add_subdirectory(ngraph)
 endif()
+if (WITH_MKLDNN)
+    add_subdirectory(mkldnn)
+endif()
--- a/python/paddle/fluid/tests/unittests/mkldnn/CMakeLists.txt
+++ b/python/paddle/fluid/tests/unittests/mkldnn/CMakeLists.txt
+file(GLOB TEST_OPS RELATIVE "${CMAKE_CURRENT_SOURCE_DIR}" "test_*.py")
+string(REPLACE ".py" "" TEST_OPS "${TEST_OPS}")
+foreach(TEST_OP ${TEST_OPS})
+    py_test_modules(${TEST_OP} MODULES ${TEST_OP})
+endforeach(TEST_OP)
--- a/python/paddle/fluid/tests/unittests/mkldnn/__init__.py
+++ b/python/paddle/fluid/tests/unittests/mkldnn/__init__.py
+#   Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserve.
+#
+# 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
+#
+#     http://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.
--- a/python/paddle/fluid/tests/unittests/test_activation_mkldnn_op.py
+++ b/python/paddle/fluid/tests/unittests/test_activation_mkldnn_op.py
@@ -17,9 +17,9 @@ from __future__ import print_function
 import unittest
 import numpy as np
 import paddle.fluid.core as core
-from op_test import OpTest
+from paddle.fluid.tests.unittests.op_test import OpTest
 from scipy.special import expit
-from test_activation_op import TestRelu, TestTanh, TestSqrt, TestAbs
+from paddle.fluid.tests.unittests.test_activation_op import TestRelu, TestTanh, TestSqrt, TestAbs
 class TestMKLDNNReluDim2(TestRelu):

--- a/python/paddle/fluid/tests/unittests/test_batch_norm_mkldnn_op.py
+++ b/python/paddle/fluid/tests/unittests/test_batch_norm_mkldnn_op.py
@@ -19,9 +19,9 @@ import numpy as np
 import paddle.fluid.core as core
 from paddle.fluid.op import Operator
 import paddle.fluid as fluid
-from op_test import OpTest
+from paddle.fluid.tests.unittests.op_test import OpTest
 from paddle.fluid.framework import grad_var_name
-from test_batch_norm_op import TestBatchNormOpInference, TestBatchNormOpTraining, _reference_training, _reference_grad
+from paddle.fluid.tests.unittests.test_batch_norm_op import TestBatchNormOpInference, TestBatchNormOpTraining, _reference_training, _reference_grad
 class TestMKLDNNBatchNormOpTraining(TestBatchNormOpTraining):

--- a/python/paddle/fluid/tests/unittests/test_concat_mkldnn_op.py
+++ b/python/paddle/fluid/tests/unittests/test_concat_mkldnn_op.py
@@ -15,7 +15,7 @@
 from __future__ import print_function
 import unittest
-from test_concat_op import TestConcatOp, TestConcatOp2, TestConcatOp3
+from paddle.fluid.tests.unittests.test_concat_op import TestConcatOp, TestConcatOp2, TestConcatOp3
 class TestMKLDNNConcatOp(TestConcatOp):

--- a/python/paddle/fluid/tests/unittests/test_conv2d_int8_mkldnn_op.py
+++ b/python/paddle/fluid/tests/unittests/test_conv2d_int8_mkldnn_op.py
@@ -18,8 +18,8 @@ import unittest
 import numpy as np
 import paddle.fluid.core as core
-from op_test import OpTest
+from paddle.fluid.tests.unittests.op_test import OpTest
-from test_conv2d_op import conv2d_forward_naive, TestConv2dOp
+from paddle.fluid.tests.unittests.test_conv2d_op import conv2d_forward_naive, TestConv2dOp
 def conv2d_forward_refer(input, filter, group, conv_param):

--- a/python/paddle/fluid/tests/unittests/test_conv2d_mkldnn_op.py
+++ b/python/paddle/fluid/tests/unittests/test_conv2d_mkldnn_op.py
@@ -16,7 +16,7 @@ from __future__ import print_function
 import unittest
-from test_conv2d_op import TestConv2dOp, TestWithPad, TestWithStride, TestWithGroup, TestWith1x1, TestWithInput1x1Filter1x1
+from paddle.fluid.tests.unittests.test_conv2d_op import TestConv2dOp, TestWithPad, TestWithStride, TestWithGroup, TestWith1x1, TestWithInput1x1Filter1x1
 class TestMKLDNN(TestConv2dOp):

--- a/python/paddle/fluid/tests/unittests/test_conv2d_transpose_mkldnn_op.py
+++ b/python/paddle/fluid/tests/unittests/test_conv2d_transpose_mkldnn_op.py
@@ -16,7 +16,7 @@ from __future__ import print_function
 import unittest
-from test_conv2d_transpose_op import TestConv2dTransposeOp, TestWithPad, TestWithStride
+from paddle.fluid.tests.unittests.test_conv2d_transpose_op import TestConv2dTransposeOp, TestWithPad, TestWithStride
 class TestMKLDNN(TestConv2dTransposeOp):

--- a/python/paddle/fluid/tests/unittests/test_conv3d_mkldnn_op.py
+++ b/python/paddle/fluid/tests/unittests/test_conv3d_mkldnn_op.py
@@ -16,7 +16,7 @@ from __future__ import print_function
 import unittest
-from test_conv3d_op import TestConv3dOp, TestCase1, TestWithGroup1, TestWithGroup2, TestWith1x1, TestWithInput1x1Filter1x1
+from paddle.fluid.tests.unittests.test_conv3d_op import TestConv3dOp, TestCase1, TestWithGroup1, TestWithGroup2, TestWith1x1, TestWithInput1x1Filter1x1
 class TestMKLDNN(TestConv3dOp):

--- a/python/paddle/fluid/tests/unittests/test_dequantize_mkldnn_op.py
+++ b/python/paddle/fluid/tests/unittests/test_dequantize_mkldnn_op.py
@@ -16,7 +16,7 @@ from __future__ import print_function
 import unittest
 import numpy as np
-from op_test import OpTest
+from paddle.fluid.tests.unittests.op_test import OpTest
 class TestDeQuantizeOp(OpTest):

--- a/python/paddle/fluid/tests/unittests/test_elementwise_add_mkldnn_op.py
+++ b/python/paddle/fluid/tests/unittests/test_elementwise_add_mkldnn_op.py
@@ -16,8 +16,8 @@ from __future__ import print_function
 import unittest
 import numpy as np
 import paddle.fluid.core as core
-from op_test import OpTest
+from paddle.fluid.tests.unittests.op_test import OpTest
-from test_elementwise_add_op import *
+from paddle.fluid.tests.unittests.test_elementwise_add_op import *
 '''
 Some tests differ from the tests defined in test_elementwise_add_op.py
 because MKLDNN does not support tensors of number of dimensions 3.

--- a/python/paddle/fluid/tests/unittests/test_elementwise_mul_mkldnn_op.py
+++ b/python/paddle/fluid/tests/unittests/test_elementwise_mul_mkldnn_op.py
@@ -15,10 +15,10 @@
 from __future__ import print_function
 import unittest
 import numpy as np
-from op_test import OpTest
+from paddle.fluid.tests.unittests.op_test import OpTest
 import paddle.fluid.core as core
 from paddle.fluid.op import Operator
-from test_elementwise_mul_op import *
+from paddle.fluid.tests.unittests.test_elementwise_mul_op import *
 class TestElementwiseMulMKLDNNOp_BroadcastNCHW16c(ElementwiseMulOp):

--- a/python/paddle/fluid/tests/unittests/test_fc_mkldnn_op.py
+++ b/python/paddle/fluid/tests/unittests/test_fc_mkldnn_op.py
@@ -16,7 +16,7 @@ from __future__ import print_function
 import unittest
 import numpy as np
-from op_test import OpTest
+from paddle.fluid.tests.unittests.op_test import OpTest
 def fully_connected_naive(input, weights, bias_data=None):

--- a/python/paddle/fluid/tests/unittests/test_gaussian_random_mkldnn_op.py
+++ b/python/paddle/fluid/tests/unittests/test_gaussian_random_mkldnn_op.py
@@ -16,7 +16,7 @@ from __future__ import print_function
 import unittest
-from test_gaussian_random_op import TestGaussianRandomOp
+from paddle.fluid.tests.unittests.test_gaussian_random_op import TestGaussianRandomOp
 class TestMKLDNN(TestGaussianRandomOp):

--- a/python/paddle/fluid/tests/unittests/test_lrn_mkldnn_op.py
+++ b/python/paddle/fluid/tests/unittests/test_lrn_mkldnn_op.py
@@ -15,7 +15,7 @@
 from __future__ import print_function
 import unittest
-from test_lrn_op import TestLRNOp
+from paddle.fluid.tests.unittests.test_lrn_op import TestLRNOp
 class TestLRNMKLDNNOp(TestLRNOp):

--- a/python/paddle/fluid/tests/unittests/test_pool2d_int8_mkldnn_op.py
+++ b/python/paddle/fluid/tests/unittests/test_pool2d_int8_mkldnn_op.py
@@ -19,8 +19,8 @@ import unittest
 import numpy as np
 import paddle.fluid.core as core
-from op_test import OpTest
+from paddle.fluid.tests.unittests.op_test import OpTest
-from test_pool2d_op import TestPool2D_Op, avg_pool2D_forward_naive, max_pool2D_forward_naive
+from paddle.fluid.tests.unittests.test_pool2d_op import TestPool2D_Op, avg_pool2D_forward_naive, max_pool2D_forward_naive
 class TestPool2dMKLDNNInt8_Op(TestPool2D_Op):

--- a/python/paddle/fluid/tests/unittests/test_pool2d_mkldnn_op.py
+++ b/python/paddle/fluid/tests/unittests/test_pool2d_mkldnn_op.py
@@ -15,7 +15,7 @@
 from __future__ import print_function
 import unittest
-from test_pool2d_op import TestPool2D_Op, TestCase1, TestCase2, TestCase3, TestCase4, TestCase5
+from paddle.fluid.tests.unittests.test_pool2d_op import TestPool2D_Op, TestCase1, TestCase2, TestCase3, TestCase4, TestCase5
 def create_test_mkldnn_class(parent):

--- a/python/paddle/fluid/tests/unittests/test_quantize_mkldnn_op.py
+++ b/python/paddle/fluid/tests/unittests/test_quantize_mkldnn_op.py
@@ -16,7 +16,7 @@ from __future__ import print_function
 import unittest
 import numpy as np
-from op_test import OpTest
+from paddle.fluid.tests.unittests.op_test import OpTest
 class TestQuantizeOp(OpTest):

--- a/python/paddle/fluid/tests/unittests/test_sum_mkldnn_op.py
+++ b/python/paddle/fluid/tests/unittests/test_sum_mkldnn_op.py
@@ -16,7 +16,7 @@ from __future__ import print_function
 import unittest
-from test_sum_op import TestSumOp
+from paddle.fluid.tests.unittests.test_sum_op import TestSumOp
 class TestMKLDNN(TestSumOp):

--- a/python/paddle/fluid/tests/unittests/test_transpose_mkldnn_op.py
+++ b/python/paddle/fluid/tests/unittests/test_transpose_mkldnn_op.py
@@ -16,7 +16,7 @@ from __future__ import print_function
 import unittest
-from test_transpose_op import TestTransposeOp
+from paddle.fluid.tests.unittests.test_transpose_op import TestTransposeOp
 class TestTransposeMKLDNN(TestTransposeOp):

--- a/python/paddle/fluid/tests/unittests/ngraph/test_pool2d_ngraph_op.py
+++ b/python/paddle/fluid/tests/unittests/ngraph/test_pool2d_ngraph_op.py
+#   Copyright (c) 2018 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
+#
+#     http://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.
+from __future__ import print_function
+from paddle.fluid.tests.unittests.test_pool2d_op import *
+class TestNGRAPHPool2D_Op(TestPool2D_Op):
+    def init_test_case(self):
+        super(TestNGRAPHPool2D_Op, self).init_test_case()
+class TestNGRAPHCase1(TestCase1):
+    def init_test_case(self):
+        super(TestNGRAPHCase1, self).init_test_case()
+class TestNGRAPHCase2(TestCase2):
+    def init_test_case(self):
+        super(TestNGRAPHCase2, self).init_test_case()
+class TestNGRAPHCase3(TestCase3):
+    def init_pool_type(self):
+        super(TestNGRAPHCase3, self).init_pool_type()
+class TestNGRAPHCase4(TestCase4):
+    def init_pool_type(self):
+        super(TestNGRAPHCase4, self).init_pool_type()
+class TestNGRAPHCase5(TestCase5):
+    def init_pool_type(self):
+        super(TestNGRAPHCase5, self).init_pool_type()
+if __name__ == '__main__':
+    unittest.main()
--- a/python/paddle/fluid/tests/unittests/test_beam_search_op.py
+++ b/python/paddle/fluid/tests/unittests/test_beam_search_op.py
@@ -38,6 +38,7 @@ class BeamSearchOpTester(unittest.TestCase):
        self._create_pre_ids()
        self.scope.var('selected_ids')
        self.scope.var('selected_scores')
+        self.scope.var('parent_idx')
    def test_run(self):
        op = Operator(
@@ -48,12 +49,14 @@ class BeamSearchOpTester(unittest.TestCase):
            scores='scores',
            selected_ids='selected_ids',
            selected_scores='selected_scores',
+            parent_idx='parent_idx',
            level=0,
            beam_size=2,
            end_id=0, )
        op.run(self.scope, core.CPUPlace())
        selected_ids = self.scope.find_var("selected_ids").get_tensor()
        selected_scores = self.scope.find_var("selected_scores").get_tensor()
+        parent_idx = self.scope.find_var("parent_idx").get_tensor()
        self.assertTrue(
            np.allclose(
                np.array(selected_ids), np.array([4, 2, 3, 8])[:, np.newaxis]))
@@ -62,6 +65,8 @@ class BeamSearchOpTester(unittest.TestCase):
                np.array(selected_scores),
                np.array([0.5, 0.6, 0.9, 0.7])[:, np.newaxis]))
        self.assertEqual(selected_ids.lod(), [[0, 2, 4], [0, 1, 2, 3, 4]])
+        self.assertTrue(
+            np.allclose(np.array(parent_idx), np.array([0, 1, 2, 3])))
    def _create_pre_ids(self):
        np_data = np.array([[1, 2, 3, 4]], dtype='int64')

--- a/python/paddle/fluid/tests/unittests/test_box_coder_op.py
+++ b/python/paddle/fluid/tests/unittests/test_box_coder_op.py
@@ -21,80 +21,80 @@ import math
 from op_test import OpTest
-def box_coder(target_box, prior_box, prior_box_var, output_box, code_type,
+def box_decoder(t_box, p_box, pb_v, output_box, norm, axis=0):
-              box_normalized):
+    pb_w = p_box[:, 2] - p_box[:, 0] + (norm == False)
-    prior_box_x = (
+    pb_h = p_box[:, 3] - p_box[:, 1] + (norm == False)
-        (prior_box[:, 2] + prior_box[:, 0]) / 2).reshape(1, prior_box.shape[0])
+    pb_x = pb_w * 0.5 + p_box[:, 0]
-    prior_box_y = (
+    pb_y = pb_h * 0.5 + p_box[:, 1]
-        (prior_box[:, 3] + prior_box[:, 1]) / 2).reshape(1, prior_box.shape[0])
+    shape = (1, p_box.shape[0]) if axis == 0 else (p_box.shape[0], 1)
-    prior_box_width = (
-        (prior_box[:, 2] - prior_box[:, 0])).reshape(1, prior_box.shape[0])
+    pb_w = pb_w.reshape(shape)
-    prior_box_height = (
+    pb_h = pb_h.reshape(shape)
-        (prior_box[:, 3] - prior_box[:, 1])).reshape(1, prior_box.shape[0])
+    pb_x = pb_x.reshape(shape)
-    prior_box_var = prior_box_var.reshape(1, prior_box_var.shape[0],
+    pb_y = pb_y.reshape(shape)
-                                          prior_box_var.shape[1])
-    if not box_normalized:
+    if pb_v.ndim == 2:
-        prior_box_height = prior_box_height + 1
+        pb_v = pb_v.reshape(1, pb_v.shape[0], pb_v.shape[1])
-        prior_box_width = prior_box_width + 1
+    if pb_v.ndim == 1:
+        tb_x = pb_v[0] * t_box[:, :, 0] * pb_w + pb_x
-    if (code_type == "EncodeCenterSize"):
+        tb_y = pb_v[1] * t_box[:, :, 1] * pb_h + pb_y
-        target_box_x = ((target_box[:, 2] + target_box[:, 0]) / 2).reshape(
+        tb_w = np.exp(pb_v[2] * t_box[:, :, 2]) * pb_w
-            target_box.shape[0], 1)
+        tb_h = np.exp(pb_v[3] * t_box[:, :, 3]) * pb_h
-        target_box_y = ((target_box[:, 3] + target_box[:, 1]) / 2).reshape(
+    else:
-            target_box.shape[0], 1)
+        tb_x = pb_v[:, :, 0] * t_box[:, :, 0] * pb_w + pb_x
-        target_box_width = ((target_box[:, 2] - target_box[:, 0])).reshape(
+        tb_y = pb_v[:, :, 1] * t_box[:, :, 1] * pb_h + pb_y
-            target_box.shape[0], 1)
+        tb_w = np.exp(pb_v[:, :, 2] * t_box[:, :, 2]) * pb_w
-        target_box_height = ((target_box[:, 3] - target_box[:, 1])).reshape(
+        tb_h = np.exp(pb_v[:, :, 3] * t_box[:, :, 3]) * pb_h
-            target_box.shape[0], 1)
+    output_box[:, :, 0] = tb_x - tb_w / 2
-        if not box_normalized:
+    output_box[:, :, 1] = tb_y - tb_h / 2
-            target_box_height = target_box_height + 1
+    output_box[:, :, 2] = tb_x + tb_w / 2 - (not norm)
-            target_box_width = target_box_width + 1
+    output_box[:, :, 3] = tb_y + tb_h / 2 - (not norm)
-        output_box[:,:,0] = (target_box_x - prior_box_x) / prior_box_width / \
-                prior_box_var[:,:,0]
+def box_encoder(t_box, p_box, pb_v, output_box, norm):
-        output_box[:,:,1] = (target_box_y - prior_box_y) / prior_box_height / \
+    pb_w = p_box[:, 2] - p_box[:, 0] + (norm == False)
-                prior_box_var[:,:,1]
+    pb_h = p_box[:, 3] - p_box[:, 1] + (norm == False)
-        output_box[:,:,2] = np.log(np.fabs(target_box_width / prior_box_width)) / \
+    pb_x = pb_w * 0.5 + p_box[:, 0]
-                prior_box_var[:,:,2]
+    pb_y = pb_h * 0.5 + p_box[:, 1]
-        output_box[:,:,3] = np.log(np.fabs(target_box_height / prior_box_height)) / \
+    shape = (1, p_box.shape[0])
-                prior_box_var[:,:,3]
+    pb_w = pb_w.reshape(shape)
-    elif (code_type == "DecodeCenterSize"):
+    pb_h = pb_h.reshape(shape)
-        target_box_x = prior_box_var[:,:,0] * target_box[:,:,0] * \
+    pb_x = pb_x.reshape(shape)
-                       prior_box_width + prior_box_x
+    pb_y = pb_y.reshape(shape)
-        target_box_y = prior_box_var[:,:,1] * target_box[:,:,1] * \
-                       prior_box_height + prior_box_y
+    if pb_v.ndim == 2:
-        target_box_width = np.exp(prior_box_var[:,:,2] * target_box[:,:,2]) * \
+        pb_v = pb_v.reshape(1, pb_v.shape[0], pb_v.shape[1])
-                           prior_box_width
+    tb_x = ((t_box[:, 2] + t_box[:, 0]) / 2).reshape(t_box.shape[0], 1)
-        target_box_height = np.exp(prior_box_var[:,:,3] * target_box[:,:,3]) * \
+    tb_y = ((t_box[:, 3] + t_box[:, 1]) / 2).reshape(t_box.shape[0], 1)
-                            prior_box_height
+    tb_w = (t_box[:, 2] - t_box[:, 0]).reshape(t_box.shape[0], 1) + (not norm)
+    tb_h = (t_box[:, 3] - t_box[:, 1]).reshape(t_box.shape[0], 1) + (not norm)
-        output_box[:, :, 0] = target_box_x - target_box_width / 2
+    if pb_v.ndim == 1:
-        output_box[:, :, 1] = target_box_y - target_box_height / 2
+        output_box[:, :, 0] = (tb_x - pb_x) / pb_w / pb_v[0]
-        output_box[:, :, 2] = target_box_x + target_box_width / 2
+        output_box[:, :, 1] = (tb_y - pb_y) / pb_h / pb_v[1]
-        output_box[:, :, 3] = target_box_y + target_box_height / 2
+        output_box[:, :, 2] = np.log(np.fabs(tb_w / pb_w)) / pb_v[2]
-        if not box_normalized:
+        output_box[:, :, 3] = np.log(np.fabs(tb_h / pb_h)) / pb_v[3]
-            output_box[:, :, 2] = output_box[:, :, 2] - 1
+    else:
-            output_box[:, :, 3] = output_box[:, :, 3] - 1
+        output_box[:, :, 0] = (tb_x - pb_x) / pb_w / pb_v[:, :, 0]
+        output_box[:, :, 1] = (tb_y - pb_y) / pb_h / pb_v[:, :, 1]
+        output_box[:, :, 2] = np.log(np.fabs(tb_w / pb_w)) / pb_v[:, :, 2]
-def batch_box_coder(prior_box, prior_box_var, target_box, lod, code_type,
+        output_box[:, :, 3] = np.log(np.fabs(tb_h / pb_h)) / pb_v[:, :, 3]
-                    box_normalized):
-    n = target_box.shape[0]
-    m = prior_box.shape[0]
+def batch_box_coder(p_box, pb_v, t_box, lod, code_type, norm, axis=0):
+    n = t_box.shape[0]
+    m = p_box.shape[0]
+    if code_type == "DecodeCenterSize":
+        m = t_box.shape[1]
    output_box = np.zeros((n, m, 4), dtype=np.float32)
    cur_offset = 0
    for i in range(len(lod)):
        if (code_type == "EncodeCenterSize"):
-            box_coder(target_box[cur_offset:(cur_offset + lod[i]), :],
+            box_encoder(t_box[cur_offset:(cur_offset + lod[i]), :], p_box, pb_v,
-                      prior_box, prior_box_var,
+                        output_box[cur_offset:(cur_offset + lod[i]), :, :],
-                      output_box[cur_offset:(cur_offset + lod[i]), :, :],
+                        norm)
-                      code_type, box_normalized)
        elif (code_type == "DecodeCenterSize"):
-            box_coder(target_box[cur_offset:(cur_offset + lod[i]), :, :],
+            box_decoder(t_box, p_box, pb_v, output_box, norm, axis)
-                      prior_box, prior_box_var,
-                      output_box[cur_offset:(cur_offset + lod[i]), :, :],
-                      code_type, box_normalized)
        cur_offset += lod[i]
    return output_box
@@ -106,9 +106,35 @@ class TestBoxCoderOp(OpTest):
    def setUp(self):
        self.op_type = "box_coder"
        lod = [[1, 1, 1, 1, 1]]
-        prior_box = np.random.random((10, 4)).astype('float32')
+        prior_box = np.random.random((81, 4)).astype('float32')
-        prior_box_var = np.random.random((10, 4)).astype('float32')
+        prior_box_var = np.random.random((81, 4)).astype('float32')
-        target_box = np.random.random((5, 10, 4)).astype('float32')
+        target_box = np.random.random((20, 81, 4)).astype('float32')
+        code_type = "DecodeCenterSize"
+        box_normalized = False
+        output_box = batch_box_coder(prior_box, prior_box_var, target_box,
+                                     lod[0], code_type, box_normalized)
+        self.inputs = {
+            'PriorBox': prior_box,
+            'PriorBoxVar': prior_box_var,
+            'TargetBox': target_box,
+        }
+        self.attrs = {
+            'code_type': 'decode_center_size',
+            'box_normalized': False
+        }
+        self.outputs = {'OutputBox': output_box}
+class TestBoxCoderOpWithOneRankVar(OpTest):
+    def test_check_output(self):
+        self.check_output()
+    def setUp(self):
+        self.op_type = "box_coder"
+        lod = [[1, 1, 1, 1, 1]]
+        prior_box = np.random.random((81, 4)).astype('float32')
+        prior_box_var = np.random.random((4)).astype('float32')
+        target_box = np.random.random((20, 81, 4)).astype('float32')
        code_type = "DecodeCenterSize"
        box_normalized = False
        output_box = batch_box_coder(prior_box, prior_box_var, target_box,
@@ -133,9 +159,9 @@ class TestBoxCoderOpWithoutBoxVar(OpTest):
    def setUp(self):
        self.op_type = "box_coder"
        lod = [[0, 1, 2, 3, 4, 5]]
-        prior_box = np.random.random((10, 4)).astype('float32')
+        prior_box = np.random.random((81, 4)).astype('float32')
-        prior_box_var = np.ones((10, 4)).astype('float32')
+        prior_box_var = np.ones((81, 4)).astype('float32')
-        target_box = np.random.random((5, 10, 4)).astype('float32')
+        target_box = np.random.random((20, 81, 4)).astype('float32')
        code_type = "DecodeCenterSize"
        box_normalized = False
        output_box = batch_box_coder(prior_box, prior_box_var, target_box,
@@ -158,10 +184,10 @@ class TestBoxCoderOpWithLoD(OpTest):
    def setUp(self):
        self.op_type = "box_coder"
-        lod = [[4, 8, 8]]
+        lod = [[10, 20, 20]]
-        prior_box = np.random.random((10, 4)).astype('float32')
+        prior_box = np.random.random((20, 4)).astype('float32')
-        prior_box_var = np.random.random((10, 4)).astype('float32')
+        prior_box_var = np.random.random((20, 4)).astype('float32')
-        target_box = np.random.random((20, 4)).astype('float32')
+        target_box = np.random.random((50, 4)).astype('float32')
        code_type = "EncodeCenterSize"
        box_normalized = True
        output_box = batch_box_coder(prior_box, prior_box_var, target_box,
@@ -176,5 +202,63 @@ class TestBoxCoderOpWithLoD(OpTest):
        self.outputs = {'OutputBox': output_box}
+class TestBoxCoderOpWithAxis(OpTest):
+    def test_check_output(self):
+        self.check_output()
+    def setUp(self):
+        self.op_type = "box_coder"
+        lod = [[1, 1, 1, 1, 1]]
+        prior_box = np.random.random((30, 4)).astype('float32')
+        prior_box_var = np.random.random((4)).astype('float32')
+        target_box = np.random.random((30, 81, 4)).astype('float32')
+        code_type = "DecodeCenterSize"
+        box_normalized = False
+        axis = 1
+        output_box = batch_box_coder(prior_box, prior_box_var, target_box,
+                                     lod[0], code_type, box_normalized, axis)
+        self.inputs = {
+            'PriorBox': prior_box,
+            'PriorBoxVar': prior_box_var,
+            'TargetBox': target_box,
+        }
+        self.attrs = {
+            'code_type': 'decode_center_size',
+            'box_normalized': False,
+            'axis': axis
+        }
+        self.outputs = {'OutputBox': output_box}
+class TestBoxCoderOpWithVariance(OpTest):
+    def test_check_output(self):
+        self.check_output()
+    def setUp(self):
+        self.op_type = "box_coder"
+        lod = [[1, 1, 1, 1, 1]]
+        prior_box = np.random.random((30, 4)).astype('float32')
+        prior_box_var = np.random.random((4)).astype('float32')
+        target_box = np.random.random((30, 81, 4)).astype('float32')
+        code_type = "DecodeCenterSize"
+        box_normalized = False
+        axis = 1
+        output_box = batch_box_coder(prior_box, prior_box_var, target_box,
+                                     lod[0], code_type, box_normalized, axis)
+        self.inputs = {
+            'PriorBox': prior_box,
+            'TargetBox': target_box,
+        }
+        self.attrs = {
+            'code_type': 'decode_center_size',
+            'box_normalized': False,
+            'variance': prior_box_var.astype(np.float).flatten(),
+            'axis': axis
+        }
+        self.outputs = {'OutputBox': output_box}
 if __name__ == '__main__':
    unittest.main()
--- a/python/paddle/fluid/tests/unittests/test_imperative_ptb_rnn.py
+++ b/python/paddle/fluid/tests/unittests/test_imperative_ptb_rnn.py
@@ -268,9 +268,6 @@ class TestImperativePtbRnn(unittest.TestCase):
                sgd.minimize(dy_loss)
                for param in ptb_model.parameters():
                    dy_param_updated[param.name] = param._numpy()
-                # print("dy_loss is {}".format(dy_loss._numpy()))
-                # print("last_hidden is {}".format(last_hidden._numpy()))
-                # print("last_cell is {}".format(last_cell._numpy()))
        with new_program_scope():
            fluid.default_startup_program().random_seed = seed

--- a/python/paddle/fluid/tests/unittests/test_yolov3_loss_op.py
+++ b/python/paddle/fluid/tests/unittests/test_yolov3_loss_op.py
@@ -16,174 +16,179 @@ from __future__ import division
 import unittest
 import numpy as np
+from scipy.special import logit
+from scipy.special import expit
 from op_test import OpTest
 from paddle.fluid import core
-def sigmoid(x):
+def l2loss(x, y):
-    return 1.0 / (1.0 + np.exp(-1.0 * x))
+    return 0.5 * (y - x) * (y - x)
-def mse(x, y, num):
+def sce(x, label):
-    return ((y - x)**2).sum() / num
+    sigmoid_x = expit(x)
+    term1 = label * np.log(sigmoid_x)
+    term2 = (1.0 - label) * np.log(1.0 - sigmoid_x)
+    return -term1 - term2
-def bce(x, y, mask):
+def sigmoid(x):
-    x = x.reshape((-1))
+    return 1.0 / (1.0 + np.exp(-1.0 * x))
-    y = y.reshape((-1))
-    mask = mask.reshape((-1))
-    error_sum = 0.0
-    count = 0
-    for i in range(x.shape[0]):
-        if mask[i] > 0:
-            error_sum += y[i] * np.log(x[i]) + (1 - y[i]) * np.log(1 - x[i])
-            count += 1
-    return error_sum / (-1.0 * count)
+def batch_xywh_box_iou(box1, box2):
+    b1_left = box1[:, :, 0] - box1[:, :, 2] / 2
+    b1_right = box1[:, :, 0] + box1[:, :, 2] / 2
+    b1_top = box1[:, :, 1] - box1[:, :, 3] / 2
+    b1_bottom = box1[:, :, 1] + box1[:, :, 3] / 2
-def box_iou(box1, box2):
+    b2_left = box2[:, :, 0] - box2[:, :, 2] / 2
-    b1_x1 = box1[0] - box1[2] / 2
+    b2_right = box2[:, :, 0] + box2[:, :, 2] / 2
-    b1_x2 = box1[0] + box1[2] / 2
+    b2_top = box2[:, :, 1] - box2[:, :, 3] / 2
-    b1_y1 = box1[1] - box1[3] / 2
+    b2_bottom = box2[:, :, 1] + box2[:, :, 3] / 2
-    b1_y2 = box1[1] + box1[3] / 2
-    b2_x1 = box2[0] - box2[2] / 2
-    b2_x2 = box2[0] + box2[2] / 2
-    b2_y1 = box2[1] - box2[3] / 2
-    b2_y2 = box2[1] + box2[3] / 2
-    b1_area = (b1_x2 - b1_x1) * (b1_y2 - b1_y1)
+    left = np.maximum(b1_left[:, :, np.newaxis], b2_left[:, np.newaxis, :])
-    b2_area = (b2_x2 - b2_x1) * (b2_y2 - b2_y1)
+    right = np.minimum(b1_right[:, :, np.newaxis], b2_right[:, np.newaxis, :])
+    top = np.maximum(b1_top[:, :, np.newaxis], b2_top[:, np.newaxis, :])
+    bottom = np.minimum(b1_bottom[:, :, np.newaxis],
+                        b2_bottom[:, np.newaxis, :])
-    inter_rect_x1 = max(b1_x1, b2_x1)
+    inter_w = np.clip(right - left, 0., 1.)
-    inter_rect_y1 = max(b1_y1, b2_y1)
+    inter_h = np.clip(bottom - top, 0., 1.)
-    inter_rect_x2 = min(b1_x2, b2_x2)
+    inter_area = inter_w * inter_h
-    inter_rect_y2 = min(b1_y2, b2_y2)
-    inter_area = max(inter_rect_x2 - inter_rect_x1, 0) * max(
-        inter_rect_y2 - inter_rect_y1, 0)
-    return inter_area / (b1_area + b2_area + inter_area)
+    b1_area = (b1_right - b1_left) * (b1_bottom - b1_top)
+    b2_area = (b2_right - b2_left) * (b2_bottom - b2_top)
+    union = b1_area[:, :, np.newaxis] + b2_area[:, np.newaxis, :] - inter_area
+    return inter_area / union
-def build_target(gtboxs, gtlabel, attrs, grid_size):
-    n, b, _ = gtboxs.shape
-    ignore_thresh = attrs["ignore_thresh"]
-    anchors = attrs["anchors"]
-    class_num = attrs["class_num"]
-    an_num = len(anchors) // 2
-    obj_mask = np.zeros((n, an_num, grid_size, grid_size)).astype('float32')
-    noobj_mask = np.ones((n, an_num, grid_size, grid_size)).astype('float32')
-    tx = np.zeros((n, an_num, grid_size, grid_size)).astype('float32')
-    ty = np.zeros((n, an_num, grid_size, grid_size)).astype('float32')
-    tw = np.zeros((n, an_num, grid_size, grid_size)).astype('float32')
-    th = np.zeros((n, an_num, grid_size, grid_size)).astype('float32')
-    tconf = np.zeros((n, an_num, grid_size, grid_size)).astype('float32')
-    tcls = np.zeros(
-        (n, an_num, grid_size, grid_size, class_num)).astype('float32')
+def YOLOv3Loss(x, gtbox, gtlabel, attrs):
+    n, c, h, w = x.shape
+    b = gtbox.shape[1]
+    anchors = attrs['anchors']
+    an_num = len(anchors) // 2
+    anchor_mask = attrs['anchor_mask']
+    mask_num = len(anchor_mask)
+    class_num = attrs["class_num"]
+    ignore_thresh = attrs['ignore_thresh']
+    downsample = attrs['downsample']
+    input_size = downsample * h
+    x = x.reshape((n, mask_num, 5 + class_num, h, w)).transpose((0, 1, 3, 4, 2))
+    loss = np.zeros((n)).astype('float32')
+    pred_box = x[:, :, :, :, :4].copy()
+    grid_x = np.tile(np.arange(w).reshape((1, w)), (h, 1))
+    grid_y = np.tile(np.arange(h).reshape((h, 1)), (1, w))
+    pred_box[:, :, :, :, 0] = (grid_x + sigmoid(pred_box[:, :, :, :, 0])) / w
+    pred_box[:, :, :, :, 1] = (grid_y + sigmoid(pred_box[:, :, :, :, 1])) / h
+    x[:, :, :, :, 5:] = np.where(x[:, :, :, :, 5:] < -0.5, x[:, :, :, :, 5:],
+                                 np.ones_like(x[:, :, :, :, 5:]) * 1.0 /
+                                 class_num)
+    mask_anchors = []
+    for m in anchor_mask:
+        mask_anchors.append((anchors[2 * m], anchors[2 * m + 1]))
+    anchors_s = np.array(
+        [(an_w / input_size, an_h / input_size) for an_w, an_h in mask_anchors])
+    anchor_w = anchors_s[:, 0:1].reshape((1, mask_num, 1, 1))
+    anchor_h = anchors_s[:, 1:2].reshape((1, mask_num, 1, 1))
+    pred_box[:, :, :, :, 2] = np.exp(pred_box[:, :, :, :, 2]) * anchor_w
+    pred_box[:, :, :, :, 3] = np.exp(pred_box[:, :, :, :, 3]) * anchor_h
+    pred_box = pred_box.reshape((n, -1, 4))
+    pred_obj = x[:, :, :, :, 4].reshape((n, -1))
+    objness = np.zeros(pred_box.shape[:2]).astype('float32')
+    ious = batch_xywh_box_iou(pred_box, gtbox)
+    ious_max = np.max(ious, axis=-1)
+    objness = np.where(ious_max > ignore_thresh, -np.ones_like(objness),
+                       objness)
+    gtbox_shift = gtbox.copy()
+    gtbox_shift[:, :, 0] = 0
+    gtbox_shift[:, :, 1] = 0
+    anchors = [(anchors[2 * i], anchors[2 * i + 1]) for i in range(0, an_num)]
+    anchors_s = np.array(
+        [(an_w / input_size, an_h / input_size) for an_w, an_h in anchors])
+    anchor_boxes = np.concatenate(
+        [np.zeros_like(anchors_s), anchors_s], axis=-1)
+    anchor_boxes = np.tile(anchor_boxes[np.newaxis, :, :], (n, 1, 1))
+    ious = batch_xywh_box_iou(gtbox_shift, anchor_boxes)
+    iou_matches = np.argmax(ious, axis=-1)
+    gt_matches = iou_matches.copy()
    for i in range(n):
        for j in range(b):
-            if gtboxs[i, j, :].sum() == 0:
+            if gtbox[i, j, 2:].sum() == 0:
+                gt_matches[i, j] = -1
                continue
+            if iou_matches[i, j] not in anchor_mask:
+                gt_matches[i, j] = -1
+                continue
+            an_idx = anchor_mask.index(iou_matches[i, j])
+            gt_matches[i, j] = an_idx
+            gi = int(gtbox[i, j, 0] * w)
+            gj = int(gtbox[i, j, 1] * h)
-            gt_label = gtlabel[i, j]
+            tx = gtbox[i, j, 0] * w - gi
-            gx = gtboxs[i, j, 0] * grid_size
+            ty = gtbox[i, j, 1] * w - gj
-            gy = gtboxs[i, j, 1] * grid_size
+            tw = np.log(gtbox[i, j, 2] * input_size / mask_anchors[an_idx][0])
-            gw = gtboxs[i, j, 2] * grid_size
+            th = np.log(gtbox[i, j, 3] * input_size / mask_anchors[an_idx][1])
-            gh = gtboxs[i, j, 3] * grid_size
+            scale = (2.0 - gtbox[i, j, 2] * gtbox[i, j, 3])
+            loss[i] += sce(x[i, an_idx, gj, gi, 0], tx) * scale
-            gi = int(gx)
+            loss[i] += sce(x[i, an_idx, gj, gi, 1], ty) * scale
-            gj = int(gy)
+            loss[i] += l2loss(x[i, an_idx, gj, gi, 2], tw) * scale
+            loss[i] += l2loss(x[i, an_idx, gj, gi, 3], th) * scale
-            gtbox = [0, 0, gw, gh]
-            max_iou = 0
+            objness[i, an_idx * h * w + gj * w + gi] = 1.0
-            for k in range(an_num):
-                anchor_box = [0, 0, anchors[2 * k], anchors[2 * k + 1]]
+            for label_idx in range(class_num):
-                iou = box_iou(gtbox, anchor_box)
+                loss[i] += sce(x[i, an_idx, gj, gi, 5 + label_idx],
-                if iou > max_iou:
+                               float(label_idx == gtlabel[i, j]))
-                    max_iou = iou
-                    best_an_index = k
+        for j in range(mask_num * h * w):
-                if iou > ignore_thresh:
+            if objness[i, j] > 0:
-                    noobj_mask[i, best_an_index, gj, gi] = 0
+                loss[i] += sce(pred_obj[i, j], 1.0)
+            elif objness[i, j] == 0:
-            obj_mask[i, best_an_index, gj, gi] = 1
+                loss[i] += sce(pred_obj[i, j], 0.0)
-            noobj_mask[i, best_an_index, gj, gi] = 0
-            tx[i, best_an_index, gj, gi] = gx - gi
+    return (loss, objness.reshape((n, mask_num, h, w)).astype('float32'), \
-            ty[i, best_an_index, gj, gi] = gy - gj
+            gt_matches.astype('int32'))
-            tw[i, best_an_index, gj, gi] = np.log(gw / anchors[2 *
-                                                               best_an_index])
-            th[i, best_an_index, gj, gi] = np.log(
-                gh / anchors[2 * best_an_index + 1])
-            tconf[i, best_an_index, gj, gi] = 1
-            tcls[i, best_an_index, gj, gi, gt_label] = 1
-    return (tx, ty, tw, th, tconf, tcls, obj_mask, noobj_mask)
-def YoloV3Loss(x, gtbox, gtlabel, attrs):
-    n, c, h, w = x.shape
-    an_num = len(attrs['anchors']) // 2
-    class_num = attrs["class_num"]
-    x = x.reshape((n, an_num, 5 + class_num, h, w)).transpose((0, 1, 3, 4, 2))
-    pred_x = sigmoid(x[:, :, :, :, 0])
-    pred_y = sigmoid(x[:, :, :, :, 1])
-    pred_w = x[:, :, :, :, 2]
-    pred_h = x[:, :, :, :, 3]
-    pred_conf = sigmoid(x[:, :, :, :, 4])
-    pred_cls = sigmoid(x[:, :, :, :, 5:])
-    tx, ty, tw, th, tconf, tcls, obj_mask, noobj_mask = build_target(
-        gtbox, gtlabel, attrs, x.shape[2])
-    obj_mask_expand = np.tile(
-        np.expand_dims(obj_mask, 4), (1, 1, 1, 1, int(attrs['class_num'])))
-    loss_x = mse(pred_x * obj_mask, tx * obj_mask, obj_mask.sum())
-    loss_y = mse(pred_y * obj_mask, ty * obj_mask, obj_mask.sum())
-    loss_w = mse(pred_w * obj_mask, tw * obj_mask, obj_mask.sum())
-    loss_h = mse(pred_h * obj_mask, th * obj_mask, obj_mask.sum())
-    loss_conf_target = bce(pred_conf * obj_mask, tconf * obj_mask, obj_mask)
-    loss_conf_notarget = bce(pred_conf * noobj_mask, tconf * noobj_mask,
-                             noobj_mask)
-    loss_class = bce(pred_cls * obj_mask_expand, tcls * obj_mask_expand,
-                     obj_mask_expand)
-    return attrs['loss_weight_xy'] * (loss_x + loss_y) \
-            + attrs['loss_weight_wh'] * (loss_w + loss_h) \
-            + attrs['loss_weight_conf_target'] * loss_conf_target \
-            + attrs['loss_weight_conf_notarget'] * loss_conf_notarget \
-            + attrs['loss_weight_class'] * loss_class
 class TestYolov3LossOp(OpTest):
    def setUp(self):
-        self.loss_weight_xy = 1.0
-        self.loss_weight_wh = 1.0
-        self.loss_weight_conf_target = 1.0
-        self.loss_weight_conf_notarget = 1.0
-        self.loss_weight_class = 1.0
        self.initTestCase()
        self.op_type = 'yolov3_loss'
-        x = np.random.random(size=self.x_shape).astype('float32')
+        x = logit(np.random.uniform(0, 1, self.x_shape).astype('float32'))
        gtbox = np.random.random(size=self.gtbox_shape).astype('float32')
-        gtlabel = np.random.randint(0, self.class_num,
+        gtlabel = np.random.randint(0, self.class_num, self.gtbox_shape[:2])
-                                    self.gtbox_shape[:2]).astype('int32')
+        gtmask = np.random.randint(0, 2, self.gtbox_shape[:2])
+        gtbox = gtbox * gtmask[:, :, np.newaxis]
+        gtlabel = gtlabel * gtmask
        self.attrs = {
            "anchors": self.anchors,
+            "anchor_mask": self.anchor_mask,
            "class_num": self.class_num,
            "ignore_thresh": self.ignore_thresh,
-            "loss_weight_xy": self.loss_weight_xy,
+            "downsample": self.downsample,
-            "loss_weight_wh": self.loss_weight_wh,
-            "loss_weight_conf_target": self.loss_weight_conf_target,
-            "loss_weight_conf_notarget": self.loss_weight_conf_notarget,
-            "loss_weight_class": self.loss_weight_class,
        }
-        self.inputs = {'X': x, 'GTBox': gtbox, 'GTLabel': gtlabel}
+        self.inputs = {
+            'X': x,
+            'GTBox': gtbox.astype('float32'),
+            'GTLabel': gtlabel.astype('int32'),
+        }
+        loss, objness, gt_matches = YOLOv3Loss(x, gtbox, gtlabel, self.attrs)
        self.outputs = {
-            'Loss': np.array(
+            'Loss': loss,
-                [YoloV3Loss(x, gtbox, gtlabel, self.attrs)]).astype('float32')
+            'ObjectnessMask': objness,
+            "GTMatchMask": gt_matches
        }
    def test_check_output(self):
@@ -196,19 +201,16 @@ class TestYolov3LossOp(OpTest):
            place, ['X'],
            'Loss',
            no_grad_set=set(["GTBox", "GTLabel"]),
-            max_relative_error=0.06)
+            max_relative_error=0.3)
    def initTestCase(self):
-        self.anchors = [10, 13, 12, 12]
+        self.anchors = [10, 13, 16, 30, 33, 23]
-        self.class_num = 10
+        self.anchor_mask = [1, 2]
+        self.class_num = 5
        self.ignore_thresh = 0.5
-        self.x_shape = (5, len(self.anchors) // 2 * (5 + self.class_num), 7, 7)
+        self.downsample = 32
-        self.gtbox_shape = (5, 10, 4)
+        self.x_shape = (3, len(self.anchor_mask) * (5 + self.class_num), 5, 5)
-        self.loss_weight_xy = 2.5
+        self.gtbox_shape = (3, 5, 4)
-        self.loss_weight_wh = 0.8
-        self.loss_weight_conf_target = 1.5
-        self.loss_weight_conf_notarget = 0.5
-        self.loss_weight_class = 1.2
 if __name__ == "__main__":

--- a/python/paddle/fluid/transpiler/details/__init__.py
+++ b/python/paddle/fluid/transpiler/details/__init__.py
@@ -17,3 +17,4 @@ from __future__ import print_function
 from .program_utils import *
 from .ufind import *
 from .checkport import *
+from .vars_distributed import *
--- a/python/paddle/fluid/transpiler/details/vars_distributed.py
+++ b/python/paddle/fluid/transpiler/details/vars_distributed.py
+# Copyright (c) 2018 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
+#
+#     http://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.
+from __future__ import print_function
+from paddle.fluid.framework import Variable
+class VarStruct(object):
+    """
+    record part properties of a Variable in python.
+    """
+    def __init__(self, name, shape, dtype, type, lod_level, persistable):
+        self.name = name
+        self.shape = shape
+        self.dtype = dtype
+        self.type = type
+        self.lod_level = lod_level
+        self.persistable = persistable
+class VarDistributed(object):
+    """
+    a class to record the var distributed on parameter servers.
+    the class will record the relationship between origin var and slice var.
+    the slice var's properties, such as type/shape/offset/endpoint.
+    """
+    def __init__(self,
+                 origin_var,
+                 slice_var,
+                 is_slice=None,
+                 block_id=None,
+                 offset=None,
+                 vtype=None,
+                 endpoint=None):
+        """
+        Args:
+            origin_var(Variable|VarStruct): origin var properties
+            slice_var(Variable|VarStruct): slice var properties
+            is_slice(bool|None): slice or not, slice_var=True/False and its block size > 8192 are the judgement standard.
+            block_id(int|None): the number about the slice var.
+            offset(int|None): if the slice var is sliced, offset is the numel before the var.
+            vtype(str|None): a tag, such as Optimizer/Param/RemoteProfetch.
+            endpoint(str|None): which parameter the slice var on, such as "127.0.0.1:1001"
+        """
+        if isinstance(origin_var, Variable):
+            self.origin = self.__create_var_struct(origin_var)
+        else:
+            self.origin = origin_var
+        if isinstance(slice_var, Variable):
+            self.slice = self.__create_var_struct(slice_var)
+        else:
+            self.slice = slice_var
+        if self.equal(self.origin, self.slice):
+            self.is_slice = False
+            self.block_id = 0
+            self.offset = 0
+        else:
+            self.is_slice = True
+            self.block_id = 0
+            self.offset = 0
+        if is_slice is not None:
+            self.is_slice = is_slice
+        if block_id is not None:
+            self.block_id = block_id
+        if offset is not None:
+            self.offset = offset
+        self.vtype = vtype
+        self.endpoint = endpoint
+    @staticmethod
+    def __create_var_struct(var):
+        return VarStruct(var.name, var.shape, var.dtype, var.type,
+                         var.lod_level, var.persistable)
+    @staticmethod
+    def equal(var1, var2):
+        """
+        the two var is equal or not.
+        Returns:
+            bool: equal will return True else False
+        """
+        assert isinstance(var1, VarStruct) and isinstance(var2, VarStruct)
+        return var1.name == var2.name and \
+               var1.type == var2.type and \
+               var1.shape == var2.shape and \
+               var1.dtype == var2.dtype and \
+               var1.lod_level == var2.lod_level and \
+               var1.persistable == var2.persistable
+    def __str__(self):
+        origin_var_str = "{name} : fluid.{type}.shape{shape}.astype({dtype})". \
+            format(i="{", e="}", name=self.origin.name, type=self.origin.type,
+                   shape=self.origin.shape, dtype=self.origin.dtype)
+        slice_var_str = "{name} : fluid.{type}.shape{shape}.astype({dtype})" \
+                        ".slice({is_slice}).block({block_id}).offset({offset})". \
+            format(i="{", e="}", name=self.slice.name, type=self.slice.type,
+                   shape=self.slice.shape, dtype=self.slice.dtype,
+                   is_slice=self.is_slice, block_id=self.block_id, offset=self.offset)
+        return "var owned: {}, origin var: ( {} ), slice var: ( {} ), endpoint: {} ".format(
+            self.vtype, origin_var_str, slice_var_str, self.endpoint)
+class VarsDistributed(object):
+    """
+    a gather about VarDistributed with many methods to find distributed vars.
+    through the class, we can get overview about the distributed parameters on parameter servers.
+    this class may centralized and convenient for developer to manage and get variable's distribute.
+    other module can also use this to find variables such io.py.
+    """
+    def __init__(self):
+        self.distributed_vars = []
+    def add_distributed_var(self,
+                            origin_var,
+                            slice_var,
+                            is_slice=None,
+                            block_id=None,
+                            offset=None,
+                            vtype=None,
+                            endpoint=None):
+        """
+        add distributed var in this.
+        Args:
+            origin_var(Variable|VarStruct): origin var properties
+            slice_var(Variable|VarStruct): slice var properties
+            is_slice(bool|None): slice or not, slice_var=True/False and its block size > 8192 are the judgement standard.
+            block_id(int|None): the number about the slice var.
+            offset(int|None): if the slice var is sliced, offset is the numel before the var.
+            vtype(str|None): a tag, such as Optimizer/Param/RemoteProfetch.
+            endpoint(str|None): which parameter the slice var on, such as "127.0.0.1:1001"
+        Returns:
+            None
+        """
+        self.distributed_vars.append(
+            VarDistributed(origin_var, slice_var, is_slice, block_id, offset,
+                           vtype, endpoint))
+    def get_distributed_var_by_slice(self, var_name):
+        """
+        get distributed var by conditions.
+        Args:
+            var_name(str): slice var name, such as "w.traier0.block1"
+        Returns:
+            VarDistributed: distributed var.
+        """
+        for dist_var in self.distributed_vars:
+            if dist_var.slice.name == var_name:
+                return dist_var
+        return None
+    @staticmethod
+    def equal(var1, var2):
+        """
+        the two var is equal or not.
+        Returns:
+            bool: equal will return True else False
+        """
+        return var1.name == var2.name and \
+               var1.type == var2.type and \
+               var1.shape == var2.shape and \
+               var1.dtype == var2.dtype and \
+               var1.lod_level == var2.lod_level and \
+               var1.persistable == var2.persistable
+    def get_distributed_var_by_origin_and_ep(self, origin_var_name, endpoint):
+        """
+        get distributed var by conditions.
+        Args:
+            origin_var_name(str):
+            endpoint(str): the parameter endpoint, such as "127.0.0.1:1001"
+        Returns:
+            VarDistributed: distributed var.
+        """
+        for dist_var in self.distributed_vars:
+            if dist_var.origin.name == origin_var_name and dist_var.endpoint == endpoint:
+                return dist_var
+        return None
+    def get_distributed_vars_by_vtypes(self, vtypes, groupby=False):
+        """
+        get distributed vars by conditions.
+        Args:
+            vtype(str|None): distributed var's vtype, such as "Optimizer", "RemotePrefetch"
+            groupby(bool|False): group by origin var or not.
+        Returns:
+            list: distributed var list.
+            dict: distributed var map when groupby=True
+        """
+        vtype_vars = []
+        for var in self.distributed_vars:
+            if var.vtype in vtypes:
+                vtype_vars.append(var)
+        if not groupby:
+            return vtype_vars
+        params_map = {}
+        for var in vtype_vars:
+            origin_var_name = var.origin.name
+            if origin_var_name in params_map.keys():
+                optimizers = params_map.get(origin_var_name)
+            else:
+                optimizers = []
+            optimizers.append(var)
+            params_map[origin_var_name] = optimizers
+        return params_map
+    def get_distributed_vars_by_ep(self, endpoint, vtype=None):
+        """
+        get distributed vars by conditions.
+        Args:
+            endpoint(str): the parameter server endpoint, such as "127.0.0.1:2001"
+            vtype(str|None): distributed var's vtype, such as "Optimizer", "RemotePrefetch"
+        Returns:
+            list: distributed var list.
+        """
+        endpoint_vars = []
+        for var in self.distributed_vars:
+            if var.endpoint == endpoint:
+                endpoint_vars.append(var)
+        if not vtype:
+            return endpoint_vars
+        vtype_vars = []
+        for var in endpoint_vars:
+            if var.vtype == vtype:
+                vtype_vars.append(var)
+        return vtype_vars
+    def overview(self):
+        """
+        get the overview string about all params on all parameter servers.
+        Returns:
+            Str: overview string.
+        """
+        vars_str = []
+        for var in self.distributed_vars:
+            vars_str.append(str(var))
+        return "\n".join(vars_str)
--- a/python/paddle/fluid/transpiler/distribute_transpiler.py
+++ b/python/paddle/fluid/transpiler/distribute_transpiler.py
@@ -30,19 +30,23 @@ Steps to transpile pserver:
 5. add listen_and_serv op
 """
+import sys
 import math
-import numpy as np
+from functools import reduce
 import collections
+import six
 import logging
+import numpy as np
 from .ps_dispatcher import RoundRobin, PSDispatcher
 from .. import core, framework, unique_name
 from ..framework import Program, default_main_program, \
-    default_startup_program, Block, \
+    default_startup_program, Block, Parameter, grad_var_name
-    Parameter, Variable, grad_var_name
+from .details import wait_server_ready, UnionFind, VarStruct, VarsDistributed
-from .details import *
+from .details import delete_ops, find_op_by_output_arg
 from ..distribute_lookup_table import find_distributed_lookup_table
-from functools import reduce
 LOOKUP_TABLE_TYPE = "lookup_table"
 LOOKUP_TABLE_GRAD_TYPE = "lookup_table_grad"
@@ -62,260 +66,6 @@ def log(*args):
        print(args)
-class VarStruct(object):
-    """
-    record part properties of a Variable in python.
-    """
-    def __init__(self, name, shape, dtype, type, lod_level, persistable):
-        self.name = name
-        self.shape = shape
-        self.dtype = dtype
-        self.type = type
-        self.lod_level = lod_level
-        self.persistable = persistable
-class VarDistributed(object):
-    """
-    a class to record the var distributed on parameter servers.
-    the class will record the relationship between origin var and slice var.
-    the slice var's properties, such as type/shape/offset/endpoint.
-    """
-    def __init__(self,
-                 origin_var,
-                 slice_var,
-                 is_slice=None,
-                 block_id=None,
-                 offset=None,
-                 vtype=None,
-                 endpoint=None):
-        """
-        Args:
-            origin_var(Variable|VarStruct): origin var properties
-            slice_var(Variable|VarStruct): slice var properties
-            is_slice(bool|None): slice or not, slice_var=True/False and its block size > 8192 are the judgement standard.
-            block_id(int|None): the number about the slice var.
-            offset(int|None): if the slice var is sliced, offset is the numel before the var.
-            vtype(str|None): a tag, such as Optimizer/Param/RemoteProfetch.
-            endpoint(str|None): which parameter the slice var on, such as "127.0.0.1:1001"
-        """
-        if isinstance(origin_var, Variable):
-            self.origin = self.__create_var_struct(origin_var)
-        else:
-            self.origin = origin_var
-        if isinstance(slice_var, Variable):
-            self.slice = self.__create_var_struct(slice_var)
-        else:
-            self.slice = slice_var
-        if self.equal(self.origin, self.slice):
-            self.is_slice = False
-            self.block_id = 0
-            self.offset = 0
-        else:
-            self.is_slice = True
-            self.block_id = 0
-            self.offset = 0
-        if is_slice is not None:
-            self.is_slice = is_slice
-        if block_id is not None:
-            self.block_id = block_id
-        if offset is not None:
-            self.offset = offset
-        self.vtype = vtype
-        self.endpoint = endpoint
-    @staticmethod
-    def __create_var_struct(var):
-        return VarStruct(var.name, var.shape, var.dtype, var.type,
-                         var.lod_level, var.persistable)
-    @staticmethod
-    def equal(var1, var2):
-        """
-        the two var is equal or not.
-        Returns:
-            bool: equal will return True else False
-        """
-        assert isinstance(var1, VarStruct) and isinstance(var2, VarStruct)
-        return var1.name == var2.name and \
-               var1.type == var2.type and \
-               var1.shape == var2.shape and \
-               var1.dtype == var2.dtype and \
-               var1.lod_level == var2.lod_level and \
-               var1.persistable == var2.persistable
-    def __str__(self):
-        origin_var_str = "{name} : fluid.{type}.shape{shape}.astype({dtype})". \
-            format(i="{", e="}", name=self.origin.name, type=self.origin.type,
-                   shape=self.origin.shape, dtype=self.origin.dtype)
-        slice_var_str = "{name} : fluid.{type}.shape{shape}.astype({dtype})" \
-                        ".slice({is_slice}).block({block_id}).offset({offset})". \
-            format(i="{", e="}", name=self.slice.name, type=self.slice.type,
-                   shape=self.slice.shape, dtype=self.slice.dtype,
-                   is_slice=self.is_slice, block_id=self.block_id, offset=self.offset)
-        return "var owned: {}, origin var: ( {} ), slice var: ( {} ), endpoint: {} ".format(
-            self.vtype, origin_var_str, slice_var_str, self.endpoint)
-class VarsDistributed(object):
-    """
-    a gather about VarDistributed with many methods to find distributed vars.
-    through the class, we can get overview about the distributed parameters on parameter servers.
-    this class may centralized and convenient for developer to manage and get variable's distribute.
-    other module can also use this to find variables such io.py.
-    """
-    def __init__(self):
-        self.distributed_vars = []
-    def add_distributed_var(self,
-                            origin_var,
-                            slice_var,
-                            is_slice=None,
-                            block_id=None,
-                            offset=None,
-                            vtype=None,
-                            endpoint=None):
-        """
-        add distributed var in this.
-        Args:
-            origin_var(Variable|VarStruct): origin var properties
-            slice_var(Variable|VarStruct): slice var properties
-            is_slice(bool|None): slice or not, slice_var=True/False and its block size > 8192 are the judgement standard.
-            block_id(int|None): the number about the slice var.
-            offset(int|None): if the slice var is sliced, offset is the numel before the var.
-            vtype(str|None): a tag, such as Optimizer/Param/RemoteProfetch.
-            endpoint(str|None): which parameter the slice var on, such as "127.0.0.1:1001"
-        Returns:
-            None
-        """
-        self.distributed_vars.append(
-            VarDistributed(origin_var, slice_var, is_slice, block_id, offset,
-                           vtype, endpoint))
-    def get_distributed_var_by_slice(self, var_name):
-        """
-        get distributed var by conditions.
-        Args:
-            var_name(str): slice var name, such as "w.traier0.block1"
-        Returns:
-            VarDistributed: distributed var.
-        """
-        for dist_var in self.distributed_vars:
-            if dist_var.slice.name == var_name:
-                return dist_var
-        return None
-    @staticmethod
-    def equal(var1, var2):
-        """
-        the two var is equal or not.
-        Returns:
-            bool: equal will return True else False
-        """
-        return var1.name == var2.name and \
-               var1.type == var2.type and \
-               var1.shape == var2.shape and \
-               var1.dtype == var2.dtype and \
-               var1.lod_level == var2.lod_level and \
-               var1.persistable == var2.persistable
-    def get_distributed_var_by_origin_and_ep(self, origin_var_name, endpoint):
-        """
-        get distributed var by conditions.
-        Args:
-            origin_var_name(str):
-            endpoint(str): the parameter endpoint, such as "127.0.0.1:1001"
-        Returns:
-            VarDistributed: distributed var.
-        """
-        for dist_var in self.distributed_vars:
-            if dist_var.origin.name == origin_var_name and dist_var.endpoint == endpoint:
-                return dist_var
-        return None
-    def get_distributed_vars_by_vtypes(self, vtypes, groupby=False):
-        """
-        get distributed vars by conditions.
-        Args:
-            vtype(str|None): distributed var's vtype, such as "Optimizer", "RemotePrefetch"
-            groupby(bool|False): group by origin var or not.
-        Returns:
-            list: distributed var list.
-            dict: distributed var map when groupby=True
-        """
-        vtype_vars = []
-        for var in self.distributed_vars:
-            if var.vtype in vtypes:
-                vtype_vars.append(var)
-        if not groupby:
-            return vtype_vars
-        params_map = {}
-        for var in vtype_vars:
-            origin_var_name = var.origin.name
-            if origin_var_name in params_map.keys():
-                optimizers = params_map.get(origin_var_name)
-            else:
-                optimizers = []
-            optimizers.append(var)
-            params_map[origin_var_name] = optimizers
-        return params_map
-    def get_distributed_vars_by_ep(self, endpoint, vtype=None):
-        """
-        get distributed vars by conditions.
-        Args:
-            endpoint(str): the parameter server endpoint, such as "127.0.0.1:2001"
-            vtype(str|None): distributed var's vtype, such as "Optimizer", "RemotePrefetch"
-        Returns:
-            list: distributed var list.
-        """
-        endpoint_vars = []
-        for var in self.distributed_vars:
-            if var.endpoint == endpoint:
-                endpoint_vars.append(var)
-        if not vtype:
-            return endpoint_vars
-        vtype_vars = []
-        for var in endpoint_vars:
-            if var.vtype == vtype:
-                vtype_vars.append(var)
-        return vtype_vars
-    def overview(self):
-        """
-        get the overview string about all params on all parameter servers.
-        Returns:
-            Str: overview string.
-        """
-        vars_str = []
-        for var in self.distributed_vars:
-            vars_str.append(str(var))
-        return "\n".join(vars_str)
 class VarBlock:
    def __init__(self, varname, offset, size):
        self.varname = varname

--- a/python/setup.py.in
+++ b/python/setup.py.in
@@ -109,6 +109,7 @@ packages=['paddle',
          'paddle.fluid.contrib',
          'paddle.fluid.contrib.decoder',
          'paddle.fluid.contrib.quantize',
+          'paddle.fluid.contrib.int8_inference',
          'paddle.fluid.contrib.reader',
          'paddle.fluid.contrib.slim',
          'paddle.fluid.contrib.slim.core',