test=develop, Merge branch 'develop' of https://github.com/PaddlePaddle/Paddle into feature/imperative

Author: JiabinYang

CMakeLists.txt

@@ -212,7 +212,7 @@ endif()
 if (WITH_JEMALLOC)
     find_package(JeMalloc REQUIRED)
     include_directories(${JEMALLOC_INCLUDE_DIR})
-    add_definitions(-DWITH_JEMALLOC)
+    add_definitions(-DPADDLE_WITH_JEMALLOC)
 endif()
 
 include(generic)            # simplify cmake module
@@ -276,9 +276,3 @@ add_subdirectory(paddle)
 if(WITH_PYTHON)
     add_subdirectory(python)
 endif()
-
-if(WITH_DOC)
-    find_package(Sphinx REQUIRED)
-    find_python_module(recommonmark REQUIRED)
-    add_subdirectory(doc)
-endif()

Dockerfile

@@ -11,12 +11,10 @@ RUN /bin/bash -c 'if [[ -n ${UBUNTU_MIRROR} ]]; then sed -i 's#http://archive.ub
 # ENV variables
 ARG WITH_GPU
 ARG WITH_AVX
-ARG WITH_DOC
 
 ENV WOBOQ OFF
 ENV WITH_GPU=${WITH_GPU:-ON}
 ENV WITH_AVX=${WITH_AVX:-ON}
-ENV WITH_DOC=${WITH_DOC:-OFF}
 
 ENV HOME /root
 # Add bash enhancements

cmake/FindSphinx.cmake

@@ -325,6 +325,7 @@ paddle.fluid.layers.iou_similarity ArgSpec(args=['x', 'y', 'name'], varargs=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.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.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))
 paddle.fluid.layers.exponential_decay ArgSpec(args=['learning_rate', 'decay_steps', 'decay_rate', 'staircase'], varargs=None, keywords=None, defaults=(False,))

paddle/fluid/API.spec

@@ -1,4 +1,3 @@
-
 #windows treat symbolic file as a real file, which is different with unix
 #We create a hidden file and compile it instead of origin source file.
 function(windows_symbolic TARGET)
@@ -207,3 +206,24 @@ endif (NOT WIN32)
 
 cc_library(dlpack_tensor SRCS dlpack_tensor.cc DEPS tensor dlpack)
 cc_test(dlpack_tensor_test SRCS dlpack_tensor_test.cc DEPS dlpack_tensor glog)
+
+# Get the current working branch
+execute_process(
+  COMMAND git rev-parse --abbrev-ref HEAD
+  WORKING_DIRECTORY ${CMAKE_SOURCE_DIR}
+  OUTPUT_VARIABLE PADDLE_BRANCH
+  OUTPUT_STRIP_TRAILING_WHITESPACE
+)
+
+# Get the latest abbreviated commit hash of the working branch
+execute_process(
+  COMMAND git log -1 --format=%h
+  WORKING_DIRECTORY ${CMAKE_SOURCE_DIR}
+  OUTPUT_VARIABLE PADDLE_COMMIT
+  OUTPUT_STRIP_TRAILING_WHITESPACE
+)
+
+message(STATUS "commit: ${PADDLE_COMMIT}")
+message(STATUS "branch: ${PADDLE_BRANCH}")
+
+configure_file(commit.h.in commit.h)

paddle/fluid/framework/CMakeLists.txt

@@ -0,0 +1,21 @@
+#pragma once
+
+#include <string>
+
+namespace paddle {
+namespace framework {
+
+static std::string paddle_commit() {
+  return "@PADDLE_COMMIT@";
+}
+
+static std::string paddle_compile_branch() {
+  return "@PADDLE_BRANCH@";
+}
+
+static std::string paddle_version() {
+  return "@PADDLE_VERSION@";
+}
+
+}  // namespace framework
+}  // namespace paddle

paddle/fluid/framework/commit.h.in

@@ -156,6 +156,8 @@ class Autograd {
       for (auto it : candidate->pre_ops_) {
         for (OpBase* pre_op : it.second) {
           if (!pre_op) continue;
+          VLOG(5) << "op dep " << candidate->op_desc_->Type() << " <---- "
+                  << it.first << " <---- " << pre_op->op_desc_->Type();
           if (visited.find(pre_op) == visited.end()) {
             visited.insert(pre_op);
             queue.push_back(pre_op);

paddle/fluid/imperative/layer.cc

@@ -28,6 +28,7 @@
 #include "paddle/fluid/framework/var_desc.h"
 #include "paddle/fluid/platform/enforce.h"
 #include "paddle/fluid/platform/device_context.h"
+#include "paddle/fluid/operators/math/math_function.h"
 
 #include "paddle/fluid/imperative/type_defs.h"
 
@@ -140,16 +141,24 @@ class VarBase {
   void RunBackward();
 
   void TrackPreOp(OpBase* pre_op, const std::string& pre_op_out_name,
-                  int pre_op_out_idx, bool stop_gradient) {
+                  int pre_op_out_idx, bool pre_op_stop_gradient) {
     pre_op_ = pre_op;
     pre_op_out_name_ = pre_op_out_name;
     pre_op_out_idx_ = pre_op_out_idx;
-    stop_gradient_ = stop_gradient;
+    if (pre_op_stop_gradient) {
+      stop_gradient_ = pre_op_stop_gradient;
+    }
   }
 
   void ClearGradient() {
-    delete grads_;
-    grads_ = new VarBase(true);
+    VLOG(1) << "clear gradient of " << var_desc_->Name();
+    if (grads_ && grads_->var_ && grads_->var_->IsInitialized()) {
+      auto grads_t = grads_->var_->GetMutable<framework::LoDTensor>();
+      operators::math::set_constant(
+          *(platform::DeviceContextPool::Instance().Get(
+              grads_->var_->Get<framework::LoDTensor>().place())),
+          grads_t, 0.0);
+    }
   }
 
   framework::LoDTensor& GradValue();

paddle/fluid/imperative/layer.h

@@ -85,11 +85,12 @@ void Tracer::Trace(OpBase* op, const VarBasePtrMap& inputs,
   op->input_vars_ = inputs;
   for (auto it : op->input_vars_) {
     auto& invars = invars_map[it.first];
+    invars.reserve(it.second.size());
     for (VarBase* inp : it.second) {
       PADDLE_ENFORCE_NOT_NULL(inp->var_, "op %s input %s nullptr",
                               op->op_desc_->Type(), inp->var_desc_->Name());
 
-      invars.push_back(inp->var_);
+      invars.emplace_back(inp->var_);
       vars[inp->var_desc_->Name()] = inp;
       if (inp->PreOp()) {
         op->pre_ops_[it.first].push_back(inp->PreOp());
@@ -106,9 +107,10 @@ void Tracer::Trace(OpBase* op, const VarBasePtrMap& inputs,
   for (auto it : op->output_vars_) {
     auto& outvars = outvars_map[it.first];
     const std::vector<VarBase*>& outputs = it.second;
+    outvars.reserve(outputs.size());
     for (size_t i = 0; i < outputs.size(); ++i) {
       VarBase* out = outputs[i];
-      outvars.push_back(out->var_);
+      outvars.emplace_back(out->var_);
       vars[out->var_desc_->Name()] = out;
 
       framework::VarDesc* var_desc = block->FindVar(out->var_desc_->Name());

paddle/fluid/imperative/tracer.cc

@@ -132,7 +132,7 @@ struct Argument {
   DECL_ARGUMENT_FIELD(tensorrt_workspace_size, TensorRtWorkspaceSize, int);
   DECL_ARGUMENT_FIELD(tensorrt_min_subgraph_size, TensorRtMinSubgraphSize, int);
   DECL_ARGUMENT_FIELD(tensorrt_precision_mode, TensorRtPrecisionMode,
-                      contrib::AnalysisConfig::Precision);
+                      AnalysisConfig::Precision);
 
   // Memory optimized related.
   DECL_ARGUMENT_FIELD(enable_memory_optim, EnableMemoryOptim, bool);