merge develop

test=develop

cmake/cuda.cmake

@@ -139,10 +139,12 @@ endfunction()
 message(STATUS "CUDA detected: " ${CUDA_VERSION})
 if (${CUDA_VERSION} LESS 7.0)
   set(paddle_known_gpu_archs ${paddle_known_gpu_archs})
+  add_definitions("-DPADDLE_CUDA_BINVER=\"60\"")
 elseif (${CUDA_VERSION} LESS 8.0) # CUDA 7.x
   set(paddle_known_gpu_archs ${paddle_known_gpu_archs7})
   list(APPEND CUDA_NVCC_FLAGS "-D_MWAITXINTRIN_H_INCLUDED")
   list(APPEND CUDA_NVCC_FLAGS "-D__STRICT_ANSI__")
+  add_definitions("-DPADDLE_CUDA_BINVER=\"70\"")
 elseif (${CUDA_VERSION} LESS 9.0) # CUDA 8.x
   set(paddle_known_gpu_archs ${paddle_known_gpu_archs8})
   list(APPEND CUDA_NVCC_FLAGS "-D_MWAITXINTRIN_H_INCLUDED")
@@ -150,6 +152,7 @@ elseif (${CUDA_VERSION} LESS 9.0) # CUDA 8.x
   # CUDA 8 may complain that sm_20 is no longer supported. Suppress the
   # warning for now.
   list(APPEND CUDA_NVCC_FLAGS "-Wno-deprecated-gpu-targets")
+  add_definitions("-DPADDLE_CUDA_BINVER=\"80\"")
 endif()
 
 include_directories(${CUDA_INCLUDE_DIRS})

cmake/cudnn.cmake

@@ -89,6 +89,7 @@ if(CUDNN_FOUND)
         if(NOT CUDNN_MAJOR_VERSION)
             set(CUDNN_VERSION "???")
         else()
+            add_definitions("-DPADDLE_CUDNN_BINVER=\"${CUDNN_MAJOR_VERSION}\"")
             math(EXPR CUDNN_VERSION
                 "${CUDNN_MAJOR_VERSION} * 1000 +
                  ${CUDNN_MINOR_VERSION} * 100 + ${CUDNN_PATCHLEVEL_VERSION}")

cmake/external/cub.cmake

@@ -32,4 +32,4 @@ endif()
 
 add_dependencies(cub extern_cub)
 
-LIST(APPEND externl_project_dependencies cub)
+LIST(APPEND external_project_dependencies cub)

cmake/external/dlpack.cmake

@@ -28,4 +28,4 @@ endif()
 
 add_dependencies(dlpack extern_dlpack)
 
-LIST(APPEND externl_project_dependencies dlpack)
+LIST(APPEND external_project_dependencies dlpack)

cmake/operators.cmake

@@ -110,7 +110,7 @@ function(op_library TARGET)
     # Define operators that don't need pybind here.
     foreach(manual_pybind_op "compare_op" "logical_op" "nccl_op"
 "tensor_array_read_write_op" "tensorrt_engine_op" "conv_fusion_op"
-"fusion_transpose_flatten_concat_op")
+"fusion_transpose_flatten_concat_op" "fusion_conv_inception_op")
         if ("${TARGET}" STREQUAL "${manual_pybind_op}")
             set(pybind_flag 1)
         endif()

paddle/fluid/framework/CMakeLists.txt

@@ -78,7 +78,7 @@ if (WITH_GPU)
 endif()
 cc_test(var_type_traits_test SRCS var_type_traits_test.cc DEPS var_type_traits)
 
-cc_library(scope SRCS scope.cc DEPS glog threadpool var_type_traits)
+cc_library(scope SRCS scope.cc DEPS glog threadpool xxhash var_type_traits)
 cc_library(scope_pool SRCS scope_pool.cc DEPS scope)
 cc_test(scope_test SRCS scope_test.cc DEPS scope)
 cc_test(variable_test SRCS variable_test.cc DEPS tensor var_type_traits)
@@ -189,9 +189,9 @@ cc_library(parallel_executor SRCS parallel_executor.cc DEPS
         fast_threaded_ssa_graph_executor variable_helper)
 
 if(WITH_PSLIB)
-    cc_library(async_executor SRCS async_executor.cc data_feed.cc data_feed_factory.cc executor_thread_worker.cc DEPS op_registry device_context scope framework_proto glog lod_rank_table feed_fetch_method graph_to_program_pass async_executor_proto variable_helper pslib_brpc pslib)
+    cc_library(async_executor SRCS async_executor.cc data_feed.cc data_feed_factory.cc executor_thread_worker.cc DEPS op_registry device_context scope framework_proto glog lod_rank_table feed_fetch_method graph_to_program_pass async_executor_proto variable_helper pslib_brpc pslib timer)
 else()
-    cc_library(async_executor SRCS async_executor.cc data_feed.cc data_feed_factory.cc executor_thread_worker.cc DEPS op_registry device_context scope framework_proto glog lod_rank_table feed_fetch_method graph_to_program_pass async_executor_proto variable_helper)
+    cc_library(async_executor SRCS async_executor.cc data_feed.cc data_feed_factory.cc executor_thread_worker.cc DEPS op_registry device_context scope framework_proto glog lod_rank_table feed_fetch_method graph_to_program_pass async_executor_proto variable_helper timer)
 endif(WITH_PSLIB)
 
 

paddle/fluid/framework/async_executor.cc

@@ -304,9 +304,14 @@ void AsyncExecutor::RunFromFile(const ProgramDesc& main_program,
 
   // start executing ops in multiple threads
   for (int thidx = 0; thidx < actual_thread_num; ++thidx) {
+    if (debug) {
+      threads.push_back(std::thread(&ExecutorThreadWorker::TrainFilesWithTimer,
+                                    workers[thidx].get()));
+    } else {
       threads.push_back(
           std::thread(&ExecutorThreadWorker::TrainFiles, workers[thidx].get()));
     }
+  }
 
   for (auto& th : threads) {
     th.join();

paddle/fluid/framework/details/all_reduce_op_handle.cc

@@ -50,7 +50,7 @@ void AllReduceOpHandle::RunImpl() {
 
 // FIXME(typhoonzero): If scope0(global scope) have NCCL_ID_VAR,
 // this is a distributed or inter-process call, find a better way.
-#ifdef PADDLE_WITH_CUDA
+#if defined(PADDLE_WITH_CUDA) && !defined(_WIN32)
   if (NoDummyInputSize() == 1 &&
       local_scopes_[0]->FindLocalVar(NCCL_ID_VARNAME) == nullptr) {
 #else

paddle/fluid/framework/details/execution_strategy.h

@@ -25,7 +25,7 @@ struct ExecutionStrategy {
   size_t num_threads_{0};
   bool use_cuda_{true};
   bool allow_op_delay_{false};
-  size_t num_iteration_per_drop_scope_{100};
+  size_t num_iteration_per_drop_scope_{1};
   ExecutorType type_{kDefault};
   bool dry_run_{false};
 };

paddle/fluid/framework/details/scope_buffered_ssa_graph_executor.cc

@@ -64,20 +64,26 @@ FeedFetchList ScopeBufferedSSAGraphExecutor::Run(
   }
 
   platform::RecordEvent e("ScopeBufferedSSAGraphExecutorAfterRun", nullptr);
-  drop_scope_counter_ += 1;
+  ++drop_scope_counter_;
 
-  if (!fetch_tensors.empty() ||
-      drop_scope_counter_ == strategy_.num_iteration_per_drop_scope_) {
-    drop_scope_counter_ = 0;
-    // Wait All computational streams
-    for (auto p : places_) {
-      platform::DeviceContextPool::Instance().Get(p)->Wait();
+  bool stream_end = false;
+  if (!fetch_tensors.empty()) {
+    WaitComputationalStreams();
+    stream_end = true;
+  }
+
+  if (drop_scope_counter_ == strategy_.num_iteration_per_drop_scope_) {
+    if (!stream_end) {
+      WaitComputationalStreams();
     }
+
     for (auto &scope : local_scopes_) {
       auto &local_scope =
           *scope->Var(details::kLocalExecScopeName)->GetMutable<Scope *>();
       scope->DeleteScope(local_scope);
     }
+
+    drop_scope_counter_ = 0;
   }
   if (eptr) {
     std::rethrow_exception(eptr);

paddle/fluid/framework/details/scope_buffered_ssa_graph_executor.h

@@ -47,6 +47,14 @@ class ScopeBufferedSSAGraphExecutor : public SSAGraphExecutor {
 
   FeedFetchList Run(const std::vector<std::string>& fetch_tensors) override;
 
+ private:
+  inline void WaitComputationalStreams() {
+    // Wait All computational streams
+    for (auto p : places_) {
+      platform::DeviceContextPool::Instance().Get(p)->Wait();
+    }
+  }
+
  private:
   size_t drop_scope_counter_{0};
 

paddle/fluid/framework/executor_thread_worker.cc

@@ -29,6 +29,7 @@ limitations under the License. */
 #include "paddle/fluid/inference/io.h"
 #include "paddle/fluid/platform/cpu_helper.h"
 #include "paddle/fluid/platform/place.h"
+#include "paddle/fluid/platform/timer.h"
 #include "paddle/fluid/pybind/pybind.h"
 namespace paddle {
 namespace framework {
@@ -180,6 +181,7 @@ void ExecutorThreadWorker::SetDevice() {
   return;
 #else
   static unsigned concurrency_cap = std::thread::hardware_concurrency();
+  LOG(WARNING) << "concurrency capacity " << concurrency_cap;
   int thread_id = this->thread_id_;
 
   if (static_cast<unsigned>(thread_id) < concurrency_cap) {
@@ -238,6 +240,55 @@ static void print_fetch_var(Scope* scope, const std::string& var_name) {
   VLOG(1) << "print_fetch_var: unrecognized data type:" << tensor.type();
 }
 
+void ExecutorThreadWorker::TrainFilesWithTimer() {
+  platform::SetNumThreads(1);
+  SetDevice();
+  thread_reader_->Start();
+  std::vector<double> op_total_time;
+  std::vector<std::string> op_name;
+  for (auto& op : ops_) {
+    op_name.push_back(op->Type());
+  }
+  op_total_time.resize(ops_.size());
+  for (size_t i = 0; i < op_total_time.size(); ++i) {
+    op_total_time[i] = 0.0;
+  }
+  platform::Timer timeline;
+  double total_time = 0.0;
+  double read_time = 0.0;
+  int cur_batch;
+  int batch_cnt = 0;
+  timeline.Start();
+  while ((cur_batch = thread_reader_->Next()) > 0) {
+    timeline.Pause();
+    read_time += timeline.ElapsedSec();
+    total_time += timeline.ElapsedSec();
+    for (size_t i = 0; i < ops_.size(); ++i) {
+      timeline.Start();
+      ops_[i]->Run(*thread_scope_, place_);
+      timeline.Pause();
+      op_total_time[i] += timeline.ElapsedSec();
+      total_time += timeline.ElapsedSec();
+    }
+    ++batch_cnt;
+    thread_scope_->DropKids();
+    if (thread_id_ == 0) {
+      if (batch_cnt > 0 && batch_cnt % 1000 == 0) {
+        for (size_t i = 0; i < ops_.size(); ++i) {
+          fprintf(stderr, "op_name:[%zu][%s], op_mean_time:[%fs]\n", i,
+                  op_name[i].c_str(), op_total_time[i] / batch_cnt);
+        }
+        fprintf(stderr, "mean read time: %fs\n", read_time / batch_cnt);
+        int fetch_var_num = fetch_var_names_.size();
+        for (int i = 0; i < fetch_var_num; ++i) {
+          print_fetch_var(thread_scope_, fetch_var_names_[i]);
+        }
+      }
+    }
+    timeline.Start();
+  }
+}
+
 void ExecutorThreadWorker::TrainFiles() {
   platform::SetNumThreads(1);
 
@@ -320,10 +371,12 @@ void AsyncExecutorThreadWorker::SetPSlibPtr(
     std::shared_ptr<paddle::distributed::PSlib> pslib_ptr) {
   _pslib_ptr = pslib_ptr;
 }
+
 void AsyncExecutorThreadWorker::SetPullDenseThread(
     std::shared_ptr<DensePullThread> dpt) {
   _pull_dense_thread = dpt;
 }
+
 void AsyncExecutorThreadWorker::TrainOneNetwork() {
   PrepareParams();
 

paddle/fluid/framework/executor_thread_worker.h

@@ -155,6 +155,8 @@ class ExecutorThreadWorker {
   void SetDataFeed(const std::shared_ptr<DataFeed>& datafeed);
   // A multi-thread training function
   virtual void TrainFiles();
+  // with timer log
+  virtual void TrainFilesWithTimer();
   // set fetch variable names from python interface assigned by users
   void SetFetchVarNames(const std::vector<std::string>& fetch_var_names);
 #ifdef PADDLE_WITH_PSLIB

paddle/fluid/framework/rw_lock.h

@@ -16,6 +16,8 @@ limitations under the License. */
 
 #if !defined(_WIN32)
 #include <pthread.h>
+#else
+#include <mutex>  // NOLINT
 #endif            // !_WIN32
 
 #include "paddle/fluid/platform/enforce.h"
@@ -29,17 +31,17 @@ struct RWLock {
 
   ~RWLock() { pthread_rwlock_destroy(&lock_); }
 
-  void RDLock() {
+  inline void RDLock() {
     PADDLE_ENFORCE_EQ(pthread_rwlock_rdlock(&lock_), 0,
                       "acquire read lock failed");
   }
 
-  void WRLock() {
+  inline void WRLock() {
     PADDLE_ENFORCE_EQ(pthread_rwlock_wrlock(&lock_), 0,
                       "acquire write lock failed");
   }
 
-  void UNLock() {
+  inline void UNLock() {
     PADDLE_ENFORCE_EQ(pthread_rwlock_unlock(&lock_), 0, "unlock failed");
   }
 
@@ -51,81 +53,46 @@ struct RWLock {
 // https://stackoverflow.com/questions/7125250/making-pthread-rwlock-wrlock-recursive
 // In windows, rw_lock seems like a hack. Use empty object and do nothing.
 struct RWLock {
-  void RDLock() {}
-  void WRLock() {}
-  void UNLock() {}
+  // FIXME(minqiyang): use mutex here to do fake lock
+  inline void RDLock() { mutex_.lock(); }
+
+  inline void WRLock() { mutex_.lock(); }
+
+  inline void UNLock() { mutex_.unlock(); }
+
+ private:
+  std::mutex mutex_;
 };
 #endif
 
-class RWLockGuard {
+class AutoWRLock {
  public:
-  enum Status { kUnLock, kWRLock, kRDLock };
-
-  RWLockGuard(RWLock* rw_lock, Status init_status)
-      : lock_(rw_lock), status_(Status::kUnLock) {
-    switch (init_status) {
-      case Status::kRDLock: {
-        RDLock();
-        break;
-      }
-      case Status::kWRLock: {
-        WRLock();
-        break;
-      }
-      case Status::kUnLock: {
-        break;
-      }
-    }
-  }
+  explicit AutoWRLock(RWLock* rw_lock) : lock_(rw_lock) { Lock(); }
 
-  void WRLock() {
-    switch (status_) {
-      case Status::kUnLock: {
-        lock_->WRLock();
-        status_ = Status::kWRLock;
-        break;
-      }
-      case Status::kWRLock: {
-        break;
-      }
-      case Status::kRDLock: {
-        PADDLE_THROW(
-            "Please unlock read lock first before invoking write lock.");
-        break;
-      }
-    }
-  }
+  ~AutoWRLock() { UnLock(); }
 
-  void RDLock() {
-    switch (status_) {
-      case Status::kUnLock: {
-        lock_->RDLock();
-        status_ = Status::kRDLock;
-        break;
-      }
-      case Status::kRDLock: {
-        break;
-      }
-      case Status::kWRLock: {
-        PADDLE_THROW(
-            "Please unlock write lock first before invoking read lock.");
-        break;
-      }
-    }
-  }
+ private:
+  inline void Lock() { lock_->WRLock(); }
 
-  void UnLock() {
-    if (status_ != Status::kUnLock) {
-      lock_->UNLock();
-      status_ = Status::kUnLock;
-    }
-  }
+  inline void UnLock() { lock_->UNLock(); }
+
+ private:
+  RWLock* lock_;
+};
+
+class AutoRDLock {
+ public:
+  explicit AutoRDLock(RWLock* rw_lock) : lock_(rw_lock) { Lock(); }
+
+  ~AutoRDLock() { UnLock(); }
+
+ private:
+  inline void Lock() { lock_->RDLock(); }
 
-  ~RWLockGuard() { UnLock(); }
+  inline void UnLock() { lock_->UNLock(); }
 
  private:
   RWLock* lock_;
-  Status status_;
 };
 
 }  // namespace framework

paddle/fluid/framework/scope.cc

@@ -47,9 +47,15 @@ DEFINE_bool(fast_eager_deletion_mode, false,
 // the mutex will cause serious performance issue.
 // So the mutex is disabled when `ON_INFER`.
 #ifdef PADDLE_ON_INFERENCE
-#define SCOPE_LOCK_GUARD
+#define SCOPE_KIDS_READER_LOCK
+#define SCOPE_KIDS_WRITER_LOCK
+#define SCOPE_VARS_READER_LOCK
+#define SCOPE_VARS_WRITER_LOCK
 #else
-#define SCOPE_LOCK_GUARD std::lock_guard<std::mutex> lock(mutex_);
+#define SCOPE_KIDS_READER_LOCK AutoRDLock auto_lock(&kids_lock_);
+#define SCOPE_KIDS_WRITER_LOCK AutoWRLock auto_lock(&kids_lock_);
+#define SCOPE_VARS_READER_LOCK AutoRDLock auto_lock(&vars_lock_);
+#define SCOPE_VARS_WRITER_LOCK AutoWRLock auto_lock(&vars_lock_);
 #endif
 
 namespace paddle {
@@ -67,64 +73,69 @@ bool IsFastEagerDeletionModeEnabled() { return FLAGS_fast_eager_deletion_mode; }
 Scope::~Scope() { DropKids(); }
 
 Scope& Scope::NewScope() const {
-  SCOPE_LOCK_GUARD
-  kids_.push_back(new Scope(this));
-  return *kids_.back();
+  Scope* child = new Scope(this);
+  {
+    SCOPE_KIDS_WRITER_LOCK
+    kids_.push_back(child);
+  }
+  return *child;
 }
 
 Variable* Scope::Var(const std::string& name) {
-  SCOPE_LOCK_GUARD
+  SCOPE_VARS_WRITER_LOCK
   return VarInternal(name);
 }
 
 Variable* Scope::Var(std::string* name) {
-  SCOPE_LOCK_GUARD
   auto new_name = string::Sprintf("%p.%d", this, vars_.size());
   if (name != nullptr) {
     *name = new_name;
   }
+  SCOPE_VARS_WRITER_LOCK
   return VarInternal(new_name);
 }
 
 Variable* Scope::FindVar(const std::string& name) const {
-  SCOPE_LOCK_GUARD
+  SCOPE_VARS_READER_LOCK
   return FindVarInternal(name);
 }
 
 Variable* Scope::FindLocalVar(const std::string& name) const {
-  SCOPE_LOCK_GUARD
+  SCOPE_VARS_READER_LOCK
   return FindVarLocally(name);
 }
 
 const Scope* Scope::FindScope(const Variable* var) const {
-  SCOPE_LOCK_GUARD
+  SCOPE_VARS_READER_LOCK
   return FindScopeInternal(var);
 }
 
 void Scope::DropKids() {
-  SCOPE_LOCK_GUARD
+  SCOPE_KIDS_WRITER_LOCK
   for (Scope* s : kids_) delete s;
   kids_.clear();
 }
 
 bool Scope::HasKid(const Scope* scope) const {
-  SCOPE_LOCK_GUARD
+  SCOPE_KIDS_READER_LOCK
   auto it = std::find(this->kids_.begin(), this->kids_.end(), scope);
   return it != this->kids_.end();
 }
 
 std::vector<std::string> Scope::LocalVarNames() const {
-  SCOPE_LOCK_GUARD
   std::vector<std::string> known_vars;
+  {
+    SCOPE_VARS_READER_LOCK
     known_vars.reserve(this->vars_.size());
     for (auto& p : vars_) {
       known_vars.emplace_back(p.first);
     }
+  }
   return known_vars;
 }
 
 void Scope::DeleteScope(Scope* scope) const {
-  SCOPE_LOCK_GUARD
+  SCOPE_KIDS_WRITER_LOCK
   auto it = std::find(this->kids_.begin(), this->kids_.end(), scope);
   PADDLE_ENFORCE(it != this->kids_.end(), "%p Cannot find %p as kid scope",
                  this, scope);
@@ -138,8 +149,8 @@ void Scope::DeleteScope(Scope* scope) const {
 }
 
 void Scope::EraseVars(const std::vector<std::string>& var_names) {
-  SCOPE_LOCK_GUARD
   std::set<std::string> var_set(var_names.begin(), var_names.end());
+  SCOPE_VARS_WRITER_LOCK
   for (auto it = vars_.begin(); it != vars_.end();) {
     if (var_set.find(it->first) != var_set.end()) {
       it = vars_.erase(it);
@@ -151,12 +162,12 @@ void Scope::EraseVars(const std::vector<std::string>& var_names) {
 
 void Scope::Rename(const std::string& origin_name,
                    const std::string& new_name) const {
-  SCOPE_LOCK_GUARD
+  SCOPE_VARS_WRITER_LOCK
   RenameInternal(origin_name, new_name);
 }
 
 std::string Scope::Rename(const std::string& origin_name) const {
-  SCOPE_LOCK_GUARD
+  SCOPE_VARS_WRITER_LOCK
   auto new_name = string::Sprintf("%p.%d", this, vars_.size());
   RenameInternal(origin_name, new_name);
   return new_name;

paddle/fluid/framework/scope.h

@@ -14,12 +14,18 @@ limitations under the License. */
 
 #pragma once
 
+extern "C" {
+#include <xxhash.h>
+}
+
 #include <list>
-#include <mutex>  // NOLINT
+#include <memory>
 #include <string>
 #include <unordered_map>
+#include <utility>
 #include <vector>
 
+#include "paddle/fluid/framework/rw_lock.h"
 #include "paddle/fluid/framework/variable.h"
 #include "paddle/fluid/platform/macros.h"
 
@@ -95,7 +101,14 @@ class Scope {
   std::string Rename(const std::string& origin_name) const;
 
  protected:
-  mutable std::unordered_map<std::string, std::unique_ptr<Variable>> vars_;
+  struct KeyHasher {
+    std::size_t operator()(const std::string& key) const {
+      return XXH32(key.c_str(), key.size(), 1);
+    }
+  };
+
+  mutable std::unordered_map<std::string, std::unique_ptr<Variable>, KeyHasher>
+      vars_;
 
  private:
   // Call Scope::NewScope for a sub-scope.
@@ -124,7 +137,8 @@ class Scope {
   DISABLE_COPY_AND_ASSIGN(Scope);
 
  private:
-  mutable std::mutex mutex_;
+  mutable RWLock kids_lock_;
+  mutable RWLock vars_lock_;
 };
 
 // Generate some debug string about the inherience structure of scope, quite

paddle/fluid/operators/conv_cudnn_op_cache.h

@@ -19,6 +19,10 @@ limitations under the License. */
 #include <vector>
 #include "paddle/fluid/platform/cudnn_helper.h"
 
+DECLARE_uint64(conv_workspace_size_limit);
+DECLARE_bool(cudnn_exhaustive_search);
+DECLARE_int64(cudnn_exhaustive_search_times);
+
 namespace paddle {
 namespace operators {
 
@@ -45,6 +49,7 @@ static constexpr size_t kNUM_CUDNN_BWD_DATA_ALGS = 5;
 template <typename TAlgorithm>
 class AlgorithmsCache {
  public:
+  AlgorithmsCache() : search_times_(0) { hash_.clear(); }
   // Caches the best algorithm for a given
   // combination of tensor dimensions & compute data type.
   TAlgorithm GetAlgorithm(
@@ -54,9 +59,14 @@ class AlgorithmsCache {
       int algorithmFlags,  // can set for different data type
       std::function<TAlgorithm()> gen_func);
 
+  TAlgorithm GetAlgorithm(int64_t area, int search_times, int algorithmFlags,
+                          std::function<TAlgorithm()> gen_func);
+
  private:
   std::unordered_map<int64_t, TAlgorithm> hash_;
   std::mutex mutex_;
+
+  int search_times_;
 };
 
 template <typename TAlgorithm>
@@ -107,5 +117,29 @@ TAlgorithm AlgorithmsCache<TAlgorithm>::GetAlgorithm(
   return hash_[seed];
 }
 
+template <typename TAlgorithm>
+TAlgorithm AlgorithmsCache<TAlgorithm>::GetAlgorithm(
+    int64_t area, int search_times, int algorithmFlags,
+    std::function<TAlgorithm()> gen_func) {
+  if (hash_.find(area) != hash_.end()) {
+    return hash_[area];
+  }
+  if (search_times_ < search_times) {
+    auto algo = gen_func();
+    hash_[area] = algo;
+    ++search_times_;
+    return algo;
+  }
+  TAlgorithm algo;
+  int64_t min = static_cast<uint64_t>(INT_MAX);
+  for (const auto& m : hash_) {
+    if (m.first < min) {
+      min = m.first;
+      algo = m.second;
+    }
+  }
+  return algo;
+}
+
 }  // namespace operators
 }  // namespace paddle

paddle/fluid/operators/conv_fusion_op.cc

@@ -28,6 +28,8 @@ namespace operators {
 //         x is Input,
 //         z is ResidualData,
 //         bias is Bias
+// When `split_channels` is set, y will be splitted into multiple outputs,
+// each output has split_channels[i] number of channels.
 class Conv2DFusionOpMaker : public Conv2DOpMaker {
  protected:
   void Apply() override {
@@ -36,8 +38,65 @@ class Conv2DFusionOpMaker : public Conv2DOpMaker {
         "The activation type can be 'identity', 'sigmoid', 'relu', 'relu6' "
         "'relux' , 'tanh', 'band_pass'")
         .SetDefault("relu");
+    AddAttr<std::vector<int>>(
+        "split_channels",
+        "When `split_channels` are set, there will be multiple outputs, the "
+        "output size is equal to the number of `split_channels`.")
+        .SetDefault({});
+    AddOutput("Outputs",
+              "This Outputs is used when setting `split_channels`."
+              "Usually used to fuse conv with same input and same filter size, "
+              "padding, stride, dilation size.")
+        .AsDuplicable()
+        .AsDispensable();
+    AddInput("AlgoCache",
+             "The cache of convolution algorithm, a RAW type variable.")
+        .AsDispensable();
+    AddAttr<int>(
+        "search_times",
+        "The number of exhaustive search times for convolution algorithm.")
+        .SetDefault(-1);
   }
 };
+
+class Conv2DFusionOpInferShape : public framework::InferShapeBase {
+ public:
+  void operator()(framework::InferShapeContext* ctx) const override {
+    PADDLE_ENFORCE(ctx->HasInput("Input"),
+                   "Input(Input) of ConvOp should not be null.");
+    PADDLE_ENFORCE(ctx->HasInput("Filter"),
+                   "Input(Filter) of ConvOp should not be null.");
+    auto in_dims = ctx->GetInputDim("Input");
+    auto filter_dims = ctx->GetInputDim("Filter");
+
+    std::vector<int> strides = ctx->Attrs().Get<std::vector<int>>("strides");
+    std::vector<int> paddings = ctx->Attrs().Get<std::vector<int>>("paddings");
+    std::vector<int> dilations =
+        ctx->Attrs().Get<std::vector<int>>("dilations");
+
+    std::vector<int64_t> oshape({in_dims[0], filter_dims[0]});
+    for (size_t i = 0; i < strides.size(); ++i) {
+      oshape.push_back(ConvOutputSize(in_dims[i + 2], filter_dims[i + 2],
+                                      dilations[i], paddings[i], strides[i]));
+    }
+    PADDLE_ENFORCE(ctx->HasOutput("Output"),
+                   "Output(Output) of ConvOp should not be null.");
+    ctx->SetOutputDim("Output", framework::make_ddim(oshape));
+    std::vector<int> channels =
+        ctx->Attrs().Get<std::vector<int>>("split_channels");
+    if (channels.size()) {
+      PADDLE_ENFORCE(ctx->HasOutputs("Outputs"),
+                     "Output(Outputs) of ConvOp should not be null.");
+      std::vector<framework::DDim> oshapes;
+      oshapes.reserve(channels.size());
+      for (size_t i = 0; i < channels.size(); ++i) {
+        oshapes.push_back({oshape[0], channels[i], oshape[2], oshape[3]});
+      }
+      ctx->SetOutputsDim("Outputs", oshapes);
+    }
+  }
+};
+
 // TODO(qingqing): add gradient operator for conv2d_fusion
 
 }  // namespace operators
@@ -45,4 +104,5 @@ class Conv2DFusionOpMaker : public Conv2DOpMaker {
 
 namespace ops = paddle::operators;
 REGISTER_OPERATOR(conv2d_fusion, ops::ConvOp, ops::Conv2DFusionOpMaker,
-                  ops::ConvOpInferVarType, paddle::framework::EmptyGradOpMaker);
+                  ops::Conv2DFusionOpInferShape, ops::ConvOpInferVarType,
+                  paddle::framework::EmptyGradOpMaker);

paddle/fluid/operators/conv_fusion_op.cu.cc

@@ -16,8 +16,9 @@ limitations under the License. */
 #include "paddle/fluid/operators/conv_cudnn_op_cache.h"
 #include "paddle/fluid/platform/cudnn_helper.h"
 
-DECLARE_uint64(conv_workspace_size_limit);
-DECLARE_bool(cudnn_exhaustive_search);
+DEFINE_int64(cudnn_exhaustive_search_times, -1,
+             "Exhaustive search times for cuDNN convolution, "
+             "defalut is 1, only search once.");
 
 namespace paddle {
 namespace operators {
@@ -117,20 +118,7 @@ class CUDNNConvFusionOpKernel : public framework::OpKernel<T> {
           workspace_size_limit, &algo));
       VLOG(3) << "cuDNN forward algo " << algo;
     } else {
-      AlgorithmsCache<cudnnConvolutionFwdAlgo_t>* algo_cache = nullptr;
-      if (ctx.scope().FindVar(kCUDNNFwdAlgoCache)) {
-        algo_cache =
-            ctx.scope()
-                .FindVar(kCUDNNFwdAlgoCache)
-                ->GetMutable<AlgorithmsCache<cudnnConvolutionFwdAlgo_t>>();
-      } else {
-        algo_cache =
-            const_cast<framework::Scope&>(ctx.scope())
-                .Var(kCUDNNFwdAlgoCache)
-                ->GetMutable<AlgorithmsCache<cudnnConvolutionFwdAlgo_t>>();
-      }
-      algo = algo_cache->GetAlgorithm(
-          x_dims, f_dims, strides, paddings, dilations, 0, [&]() {
+      auto search_func = [&]() {
         int returned_algo_count;
         std::array<cudnnConvolutionFwdAlgoPerf_t, kNUM_CUDNN_FWD_ALGS>
             fwd_perf_stat;
@@ -138,20 +126,52 @@ class CUDNNConvFusionOpKernel : public framework::OpKernel<T> {
           CUDNN_ENFORCE(
               platform::dynload::cudnnFindConvolutionForwardAlgorithmEx(
                   handle, cudnn_input_desc, input_data, cudnn_filter_desc,
-                      filter_data, cudnn_conv_desc, cudnn_output_desc,
-                      output_data, kNUM_CUDNN_FWD_ALGS, &returned_algo_count,
-                      fwd_perf_stat.data(), cudnn_workspace,
-                      workspace_size_limit));
+                  filter_data, cudnn_conv_desc, cudnn_output_desc, output_data,
+                  kNUM_CUDNN_FWD_ALGS, &returned_algo_count,
+                  fwd_perf_stat.data(), cudnn_workspace, workspace_size_limit));
         };
         workspace_handle.RunFunc(cudnn_find_func, workspace_size_limit);
         VLOG(3) << "Perf result: (algo: stat, time, memory)";
         for (int i = 0; i < returned_algo_count; ++i) {
           const auto& stat = fwd_perf_stat[i];
-              VLOG(3) << stat.algo << ": " << stat.status << " " << stat.time
-                      << " " << stat.memory;
+          VLOG(3) << stat.algo << ": " << stat.status << " " << stat.time << " "
+                  << stat.memory;
         }
         return fwd_perf_stat[0].algo;
-          });
+      };
+      AlgorithmsCache<cudnnConvolutionFwdAlgo_t>* algo_cache = nullptr;
+      int search_times = ctx.Attr<int>("search_times");
+      search_times = std::max(
+          static_cast<int>(FLAGS_cudnn_exhaustive_search_times), search_times);
+      if (search_times > 0) {
+        // The searched algo will be cached by `search_times` times for
+        // different input dimension. For other dimensions, select the algo
+        // of closest area.
+        auto var_name = ctx.Inputs("AlgoCache")[0];
+        algo_cache =
+            ctx.scope()
+                .FindVar(var_name)
+                ->GetMutable<AlgorithmsCache<cudnnConvolutionFwdAlgo_t>>();
+        algo = algo_cache->GetAlgorithm(x_dims[2] * x_dims[3], search_times, 0,
+                                        search_func);
+      } else {
+        // Cache searched algo in Var(kCUDNNFwdAlgoCache).
+        // all conv ops use the same kCUDNNFwdAlgoCache variable.
+        if (ctx.scope().FindVar(kCUDNNFwdAlgoCache)) {
+          algo_cache =
+              ctx.scope()
+                  .FindVar(kCUDNNFwdAlgoCache)
+                  ->GetMutable<AlgorithmsCache<cudnnConvolutionFwdAlgo_t>>();
+        } else {
+          // TODO(qingqing) remove const_cast
+          algo_cache =
+              const_cast<framework::Scope*>(ctx.scope().parent())
+                  ->Var(kCUDNNFwdAlgoCache)
+                  ->GetMutable<AlgorithmsCache<cudnnConvolutionFwdAlgo_t>>();
+        }
+        algo = algo_cache->GetAlgorithm(x_dims, f_dims, strides, paddings,
+                                        dilations, 0, search_func);
+      }
       VLOG(3) << "choose algo " << algo;
     }
 
@@ -195,6 +215,27 @@ class CUDNNConvFusionOpKernel : public framework::OpKernel<T> {
       };
       workspace_handle.RunFunc(cudnn_func, workspace_size_in_bytes);
     }
+    std::vector<int> channels = ctx.Attr<std::vector<int>>("split_channels");
+    if (channels.size()) {
+      auto outs = ctx.MultiOutput<framework::Tensor>("Outputs");
+      if (x_dims[0] == 1) {
+        // share data with Output
+        framework::Tensor t;
+        t.ShareDataWith(*output);
+        auto y_dims = output->dims();
+        t.Resize({y_dims[1], y_dims[2], y_dims[3]});
+        int s = 0;
+        for (size_t i = 0; i < channels.size(); ++i) {
+          int e = s + channels[i];
+          outs[i]->ShareDataWith(t.Slice(s, e));
+          outs[i]->Resize({x_dims[0], channels[i], y_dims[2], y_dims[3]});
+          s = e;
+        }
+      } else {
+        // TODO(qingiqng): do copy when batch size large than 1
+        PADDLE_THROW("Batch size greater than 1 is Unsupported");
+      }
+    }
   }
 };
 #endif

paddle/fluid/operators/distributed/collective_server_test.cc

@@ -52,12 +52,12 @@ std::unique_ptr<framework::Scope> GenerateVars(platform::Place place) {
   framework::Scope* scope = new framework::Scope();
   framework::Variable* var = scope->Var("var1");
   auto* slr = var->GetMutable<framework::SelectedRows>();
-  slr->set_height(1000);
+  slr->set_height(20000);
 
   auto* tensor = slr->mutable_value();
   auto* rows = slr->mutable_rows();
 
-  tensor->Resize(framework::make_ddim({3, 5}));
+  tensor->Resize(framework::make_ddim({20000, 1024}));
   tensor->mutable_data<float>(place);
 
   paddle::operators::math::set_constant(ctx, tensor, 32.7);
@@ -83,6 +83,7 @@ void Gather(const std::vector<distributed::RemoteVar>& vars,
 }
 
 TEST(PREFETCH, GPU) {
+  setenv("FLAGS_max_body_size", "2147483647", 1);
   platform::CUDAPlace place;
   platform::DeviceContextPool& pool = platform::DeviceContextPool::Instance();
   auto& ctx = *pool.Get(place);

paddle/fluid/operators/fused/CMakeLists.txt

 include(operators)
-register_operators(EXCLUDES fusion_transpose_flatten_concat_op)
+register_operators(EXCLUDES fusion_transpose_flatten_concat_op fusion_conv_inception_op)
 if (WITH_GPU)
   op_library(fusion_transpose_flatten_concat_op)
+  op_library(fusion_conv_inception_op)
   file(APPEND ${pybind_file} "USE_CUDA_ONLY_OP(fusion_transpose_flatten_concat);\n")
+  file(APPEND ${pybind_file} "USE_CUDA_ONLY_OP(conv2d_inception_fusion);\n")
 endif()

paddle/fluid/operators/fused/fusion_conv_inception_op.cc

+/* Copyright (c) 2016 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. */
+
+#include <string>
+#include <vector>
+#include "paddle/fluid/framework/op_registry.h"
+#ifdef PADDLE_WITH_CUDA
+#include "paddle/fluid/platform/cudnn_helper.h"
+#endif
+
+namespace paddle {
+namespace operators {
+
+class ConvInceptionFusionOp : public framework::OperatorWithKernel {
+ public:
+  using framework::OperatorWithKernel::OperatorWithKernel;
+  void InferShape(framework::InferShapeContext* ctx) const override {
+    // 1 x
+    auto in_dims = ctx->GetInputDim("Input");
+    // 4 filters
+    auto w_dims = ctx->GetInputsDim("Filter");
+
+    PADDLE_ENFORCE(in_dims.size(), 4, "Conv intput should be 4-D tensor.");
+    PADDLE_ENFORCE_EQ(w_dims.size(), 4, "There should be 4 filters");
+    PADDLE_ENFORCE_EQ(w_dims[0][1], in_dims[1]);
+    PADDLE_ENFORCE_EQ(w_dims[1][1], in_dims[1]);
+
+    int n = in_dims[0];
+    // compute output channel
+    // 1st channel
+    int c = w_dims[0][0];
+    // add 2nd channel
+    c += (w_dims[1][0] - w_dims[2][1] * 2);
+    // add 3rd channel
+    c += (w_dims[2][0] - w_dims[3][1]);
+    // add 4-th channel
+    c += w_dims[3][0];
+
+    int h = in_dims[2];
+    int w = in_dims[3];
+
+    ctx->SetOutputDim("Output", {n, c, h, w});
+  }
+
+ protected:
+  framework::OpKernelType GetExpectedKernelType(
+      const framework::ExecutionContext& ctx) const override {
+    return framework::OpKernelType(
+        ctx.Input<framework::LoDTensor>("Input")->type(), ctx.device_context());
+  }
+};
+
+class ConvInceptionFusionOpMaker : public framework::OpProtoAndCheckerMaker {
+ protected:
+  void Make() override {
+    AddInput("Input", "(Tensor) NCHW layout.");
+    AddInput("Filter", "(vector<Tensor>) 4 aggregated filters").AsDuplicable();
+    AddInput("Bias", "(vector<Tensor>) it's lenght is equal to Filter")
+        .AsDuplicable();
+    AddOutput("Output",
+              "(Tensor) The output tensor of convolution operator. "
+              "The format of output tensor is also NCHW.");
+    AddOutput("TempOutput", "").AsDuplicable();
+    AddAttr<std::string>(
+        "pooling_type",
+        "(string), pooling type, can be \"max\" for max-pooling "
+        "and \"avg\" for average-pooling.")
+        .InEnum({"max", "avg"});
+    AddAttr<bool>(
+        "exclusive",
+        "(bool, default True) When true, will exclude the zero-padding in the "
+        "averaging calculating, otherwise, include the zero-padding. Note, it "
+        "is only used when pooling_type is avg. The defalut is True.")
+        .SetDefault(true);
+    AddAttr<std::string>(
+        "activation",
+        "The activation type can be 'identity', 'sigmoid', 'relu', 'relu6' "
+        "'relux' , 'tanh', 'band_pass'")
+        .SetDefault("relu");
+    AddAttr<int>("workspace_size_MB",
+                 "Only used in cudnn kernel. Need set use_cudnn to true."
+                 "workspace size for cudnn, in MB, "
+                 "workspace is a section of GPU memory which will be "
+                 "allocated/freed each time the operator runs, larger "
+                 "workspace size can increase performance but also requires "
+                 "better hardware. This size should be chosen carefully.")
+        .SetDefault(4096);
+    AddComment(R"DOC(
+)DOC");
+  }
+};
+
+}  // namespace operators
+}  // namespace paddle
+
+namespace ops = paddle::operators;
+REGISTER_OPERATOR(conv2d_inception_fusion, ops::ConvInceptionFusionOp,
+                  ops::ConvInceptionFusionOpMaker,
+                  paddle::framework::EmptyGradOpMaker);

paddle/fluid/operators/fused/fusion_conv_inception_op.cu

+/* Copyright (c) 2016 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. */
+
+#include "paddle/fluid/framework/op_registry.h"
+#include "paddle/fluid/operators/conv_cudnn_op_cache.h"
+#include "paddle/fluid/platform/cudnn_helper.h"
+
+DECLARE_uint64(conv_workspace_size_limit);
+
+namespace paddle {
+namespace operators {
+
+#if CUDNN_VERSION >= 7001
+using Tensor = framework::Tensor;
+using ScopedTensorDescriptor = platform::ScopedTensorDescriptor;
+using ScopedFilterDescriptor = platform::ScopedFilterDescriptor;
+using ScopedConvolutionDescriptor = platform::ScopedConvolutionDescriptor;
+using ScopedActivationDescriptor = platform::ScopedActivationDescriptor;
+using DataLayout = platform::DataLayout;
+
+using ScopedPoolingDescriptor = platform::ScopedPoolingDescriptor;
+using PoolingMode = platform::PoolingMode;
+template <typename T>
+using ScalingParamType = typename platform::CudnnDataType<T>::ScalingParamType;
+
+template <typename T>
+using CudnnDataType = platform::CudnnDataType<T>;
+
+template <typename T>
+class CUDNNConvInceptionFusionOpKernel : public framework::OpKernel<T> {
+ public:
+  void Compute(const framework::ExecutionContext& ctx) const override {
+    auto& dev_ctx = ctx.template device_context<platform::CUDADeviceContext>();
+    auto* input = ctx.Input<Tensor>("Input");
+    auto filters = ctx.MultiInput<framework::Tensor>("Filter");
+    auto bias = ctx.MultiInput<framework::Tensor>("Bias");
+
+    auto* output = ctx.Output<Tensor>("Output");
+    auto temp_outs = ctx.MultiOutput<framework::Tensor>("TempOutput");
+
+    const std::string pool_type = ctx.Attr<std::string>("pooling_type");
+    const std::string activation = ctx.Attr<std::string>("activation");
+    const bool exclusive = ctx.Attr<bool>("exclusive");
+
+    int64_t user_workspace_size =
+        static_cast<size_t>(ctx.Attr<int>("workspace_size_MB"));
+
+    const T* input_data = input->data<T>();
+    T* output_data = output->mutable_data<T>(ctx.GetPlace());
+    T* temp_data = temp_outs[0]->mutable_data<T>(input->dims(), ctx.GetPlace());
+
+    DataLayout layout = DataLayout::kNCHW;
+    std::vector<int> in_dim = framework::vectorize2int(input->dims());
+
+    // ------------------- cudnn descriptors ---------------------
+    PoolingMode pooling_mode;
+    if (pool_type == "max") {
+      pooling_mode = PoolingMode::kMaximum;
+    } else {
+      pooling_mode = exclusive ? PoolingMode::kAverageExclusive
+                               : (PoolingMode::kAverageInclusive);
+    }
+    std::vector<int> k0x0 = {0, 0};
+    std::vector<int> k1x1 = {1, 1};
+    std::vector<int> k1x1_2 = {1, 1};
+    std::vector<int> k3x3 = {3, 3};
+    ScopedPoolingDescriptor pool_desc;
+    ScopedActivationDescriptor act_desc;
+    ScopedTensorDescriptor out_pool_desc;
+    ScopedTensorDescriptor input_desc;
+    cudnnPoolingDescriptor_t cudnn_pool_desc =
+        pool_desc.descriptor(pooling_mode, k3x3, k1x1, k1x1);
+
+    cudnnTensorDescriptor_t cudnn_input_desc = input_desc.descriptor<T>(
+        layout, framework::vectorize2int(input->dims()));
+    cudnnTensorDescriptor_t pool_out_desc = out_pool_desc.descriptor<T>(
+        layout, framework::vectorize2int(input->dims()));
+
+    cudnnDataType_t cudnn_dtype = CudnnDataType<T>::type;
+    cudnnTensorDescriptor_t* out_desc = new cudnnTensorDescriptor_t[4];
+    cudnnFilterDescriptor_t* filter_desc = new cudnnFilterDescriptor_t[4];
+    cudnnTensorDescriptor_t* bias_desc = new cudnnTensorDescriptor_t[4];
+    cudnnTensorDescriptor_t* in_desc = new cudnnTensorDescriptor_t[4];
+    cudnnConvolutionDescriptor_t* conv_desc =
+        new cudnnConvolutionDescriptor_t[4];
+    for (int i = 0; i < 4; ++i) {
+      CUDNN_ENFORCE(
+          platform::dynload::cudnnCreateFilterDescriptor(&filter_desc[i]));
+      CUDNN_ENFORCE(
+          platform::dynload::cudnnCreateTensorDescriptor(&bias_desc[i]));
+      CUDNN_ENFORCE(
+          platform::dynload::cudnnCreateTensorDescriptor(&in_desc[i]));
+      CUDNN_ENFORCE(
+          platform::dynload::cudnnCreateTensorDescriptor(&out_desc[i]));
+      CUDNN_ENFORCE(
+          platform::dynload::cudnnCreateConvolutionDescriptor(&conv_desc[i]));
+    }
+
+    std::vector<std::vector<int>> filter_dims;
+    std::vector<std::vector<int>> bias_dims;
+    std::vector<std::vector<int>> in_dims;
+    std::vector<std::vector<int>> out_dims;
+    std::vector<std::vector<int>> in_strides;
+    std::vector<std::vector<int>> out_strides;
+    std::vector<std::vector<int>> bias_strides;
+
+    cudnnTensorFormat_t format = CUDNN_TENSOR_NCHW;
+    int n = in_dim[0];
+    int h = in_dim[2];
+    int w = in_dim[3];
+    int oc = output->dims()[1];
+
+    cudnnDataType_t compute_type = (cudnn_dtype == CUDNN_DATA_DOUBLE)
+                                       ? CUDNN_DATA_DOUBLE
+                                       : CUDNN_DATA_FLOAT;
+
+    for (int i = 0; i < 4; ++i) {
+      filter_dims.push_back(framework::vectorize2int(filters[i]->dims()));
+      CUDNN_ENFORCE(platform::dynload::cudnnSetFilterNdDescriptor(
+          filter_desc[i], cudnn_dtype, format, 4, filter_dims[i].data()));
+      bias_dims.push_back({1, filter_dims[i][0], 1, 1});
+      bias_strides.push_back({filter_dims[i][0], 1, 1, 1});
+      CUDNN_ENFORCE(platform::dynload::cudnnSetTensorNdDescriptor(
+          bias_desc[i], cudnn_dtype, 4, bias_dims[i].data(),
+          bias_strides[i].data()));
+      in_dims.push_back({n, filter_dims[i][1], h, w});
+      out_dims.push_back({n, filter_dims[i][0], h, w});
+      in_strides.push_back({filter_dims[i][1] * h * w, h * w, w, 1});
+      out_strides.push_back({oc * h * w, h * w, w, 1});
+
+      if (i < 2) {
+        CUDNN_ENFORCE(platform::dynload::cudnnSetConvolutionNdDescriptor(
+            conv_desc[i], 2, k0x0.data(), k1x1.data(), k1x1.data(),
+            CUDNN_CROSS_CORRELATION, compute_type));
+      } else {
+        CUDNN_ENFORCE(platform::dynload::cudnnSetConvolutionNdDescriptor(
+            conv_desc[i], 2, k1x1.data(), k1x1.data(), k1x1.data(),
+            CUDNN_CROSS_CORRELATION, compute_type));
+      }
+      CUDNN_ENFORCE(platform::dynload::cudnnSetConvolutionMathType(
+          conv_desc[i], CUDNN_DEFAULT_MATH));
+    }
+    in_dims[2][1] *= 2;
+    in_strides[2][0] = oc * h * w;
+    out_strides[2][0] = filter_dims[2][0] * h * w;  // this out is continuous.
+    in_strides[3][0] = filter_dims[2][0] * h * w;
+    CUDNN_ENFORCE(
+        platform::dynload::cudnnSetConvolutionGroupCount(conv_desc[2], 2));
+
+    cudnnConvolutionFwdAlgo_t algo[4];
+    auto handle = dev_ctx.cudnn_handle();
+    size_t workspace_size_in_bytes = 0;  // final workspace to allocate.
+
+    size_t workspace_size_limit = kCONV_CUDNN_WORKSPACE_LIMIT_BYTES;
+    if (FLAGS_conv_workspace_size_limit > 0 || user_workspace_size > 0) {
+      int64_t max_user_size =
+          std::max(static_cast<int64_t>(FLAGS_conv_workspace_size_limit),
+                   user_workspace_size);
+      workspace_size_limit = max_user_size * 1024 * 1024;
+    }
+
+    for (int i = 0; i < 4; ++i) {
+      CUDNN_ENFORCE(platform::dynload::cudnnSetTensorNdDescriptor(
+          in_desc[i], cudnn_dtype, 4, in_dims[i].data(), in_strides[i].data()));
+      CUDNN_ENFORCE(platform::dynload::cudnnSetTensorNdDescriptor(
+          out_desc[i], cudnn_dtype, 4, out_dims[i].data(),
+          out_strides[i].data()));
+      CUDNN_ENFORCE(platform::dynload::cudnnGetConvolutionForwardAlgorithm(
+          handle, in_desc[i], filter_desc[i], conv_desc[i], out_desc[i],
+          CUDNN_CONVOLUTION_FWD_SPECIFY_WORKSPACE_LIMIT, workspace_size_limit,
+          &algo[i]));
+      size_t tmp_size = 0;
+      CUDNN_ENFORCE(platform::dynload::cudnnGetConvolutionForwardWorkspaceSize(
+          handle, in_desc[i], filter_desc[i], conv_desc[i], out_desc[i],
+          algo[i], &tmp_size));
+      workspace_size_in_bytes = std::max(workspace_size_in_bytes, tmp_size);
+    }
+    cudnnActivationDescriptor_t cudnn_act_desc =
+        act_desc.descriptor<T>(activation);
+
+    int oc0 = filter_dims[0][0];
+    int oc1 = filter_dims[1][0] - filter_dims[2][1] * 2;
+    int oc3 = filter_dims[3][0];
+    int oc2 = oc - oc0 - oc1 - oc3;
+
+    // branch1: pool + 1x1 conv
+    ScalingParamType<T> alpha = 1.0f, beta = 0.0f;
+    CUDNN_ENFORCE(platform::dynload::cudnnPoolingForward(
+        handle, cudnn_pool_desc, &alpha, cudnn_input_desc, input_data, &beta,
+        pool_out_desc, temp_data));
+
+    std::vector<const void*> in_datas;
+    in_datas.push_back(static_cast<const void*>(temp_data));
+    in_datas.push_back(static_cast<const void*>(input_data));
+    in_datas.push_back(
+        static_cast<const void*>(output_data + (oc0 + oc1) * h * w));
+    T* temp2_data = temp_outs[1]->mutable_data<T>(
+        framework::make_ddim(out_dims[2]), ctx.GetPlace());
+    in_datas.push_back(static_cast<const void*>(temp2_data + oc2 * h * w));
+
+    std::vector<void*> out_datas;
+    out_datas.push_back(static_cast<void*>(output_data));
+    out_datas.push_back(static_cast<void*>(output_data + oc0 * h * w));
+    out_datas.push_back(static_cast<void*>(temp2_data));
+    out_datas.push_back(
+        static_cast<void*>(output_data + (oc0 + oc1 + oc2) * h * w));
+
+    for (int i = 0; i < 4; ++i) {
+      auto func = [&](void* cudnn_workspace) {
+        CUDNN_ENFORCE(platform::dynload::cudnnConvolutionBiasActivationForward(
+            handle, &alpha, in_desc[i], in_datas[i], filter_desc[i],
+            static_cast<const void*>(filters[i]->data<T>()), conv_desc[i],
+            algo[i], cudnn_workspace, workspace_size_in_bytes, &beta,
+            out_desc[i], out_datas[i], bias_desc[i],
+            static_cast<const void*>(bias[i]->data<T>()), cudnn_act_desc,
+            out_desc[i], out_datas[i]));
+      };
+      auto workspace_handle = dev_ctx.cudnn_workspace_handle();
+      workspace_handle.RunFunc(func, workspace_size_in_bytes);
+    }
+
+    cudnnTensorDescriptor_t x_desc;
+    cudnnTensorDescriptor_t y_desc;
+    CUDNN_ENFORCE(platform::dynload::cudnnCreateTensorDescriptor(&x_desc));
+    CUDNN_ENFORCE(platform::dynload::cudnnCreateTensorDescriptor(&y_desc));
+    CUDNN_ENFORCE(platform::dynload::cudnnSetTensorNdDescriptor(
+        x_desc, cudnn_dtype, 4, out_dims[3].data(), out_strides[2].data()));
+    CUDNN_ENFORCE(platform::dynload::cudnnSetTensorNdDescriptor(
+        y_desc, cudnn_dtype, 4, out_dims[3].data(), out_strides[3].data()));
+    CUDNN_ENFORCE(platform::dynload::cudnnTransformTensor(
+        handle, CudnnDataType<T>::kOne(), x_desc,
+        static_cast<const void*>(out_datas[2]), CudnnDataType<T>::kZero(),
+        y_desc, static_cast<void*>(output_data + (oc0 + oc1) * h * w)));
+
+    for (int i = 0; i < 4; ++i) {
+      CUDNN_ENFORCE(
+          platform::dynload::cudnnDestroyTensorDescriptor(in_desc[i]));
+      CUDNN_ENFORCE(
+          platform::dynload::cudnnDestroyTensorDescriptor(out_desc[i]));
+      CUDNN_ENFORCE(
+          platform::dynload::cudnnDestroyFilterDescriptor(filter_desc[i]));
+      CUDNN_ENFORCE(
+          platform::dynload::cudnnDestroyTensorDescriptor(bias_desc[i]));
+      CUDNN_ENFORCE(
+          platform::dynload::cudnnDestroyConvolutionDescriptor(conv_desc[i]));
+    }
+    CUDNN_ENFORCE(platform::dynload::cudnnDestroyTensorDescriptor(x_desc));
+    CUDNN_ENFORCE(platform::dynload::cudnnDestroyTensorDescriptor(y_desc));
+  }
+};
+#endif
+
+}  // namespace operators
+}  // namespace paddle
+
+#if CUDNN_VERSION >= 7001
+namespace ops = paddle::operators;
+REGISTER_OP_CUDA_KERNEL(conv2d_inception_fusion,
+                        ops::CUDNNConvInceptionFusionOpKernel<float>,
+                        ops::CUDNNConvInceptionFusionOpKernel<double>);
+#endif

paddle/fluid/platform/CMakeLists.txt

@@ -84,6 +84,9 @@ cc_test(init_test SRCS init_test.cc DEPS device_context)
 nv_test(cudnn_helper_test SRCS cudnn_helper_test.cc DEPS dynload_cuda)
 nv_test(transform_test SRCS transform_test.cu DEPS memory place device_context)
 
+cc_library(timer SRCS timer.cc)
+cc_test(timer_test SRCS timer_test.cc DEPS timer)
+
 cc_library(device_tracer SRCS device_tracer.cc DEPS boost profiler_proto framework_proto ${GPU_CTX_DEPS})
 cc_library(profiler SRCS profiler.cc DEPS device_context device_tracer)
 cc_test(profiler_test SRCS profiler_test.cc DEPS profiler)

paddle/fluid/platform/dynload/cudnn.cc

@@ -38,6 +38,10 @@ CUDNN_DNN_ROUTINE_EACH_AFTER_R4(DEFINE_WRAP);
 CUDNN_DNN_ROUTINE_EACH_R5(DEFINE_WRAP);
 #endif
 
+#ifdef CUDNN_DNN_ROUTINE_EACH_R6
+CUDNN_DNN_ROUTINE_EACH_R6(DEFINE_WRAP);
+#endif
+
 #ifdef CUDNN_DNN_ROUTINE_EACH_R7
 CUDNN_DNN_ROUTINE_EACH_R7(DEFINE_WRAP);
 #endif

paddle/fluid/platform/dynload/dynamic_loader.cc

@@ -53,6 +53,12 @@ namespace platform {
 namespace dynload {
 static constexpr char cupti_lib_path[] = CUPTI_LIB_PATH;
 
+#if defined(_WIN32) && defined(PADDLE_WITH_CUDA)
+static constexpr char* win_cublas_lib = "cublas64_" PADDLE_CUDA_BINVER ".dll";
+static constexpr char* win_curand_lib = "curand64_" PADDLE_CUDA_BINVER ".dll";
+static constexpr char* win_cudnn_lib = "cudnn64_" PADDLE_CUDNN_BINVER ".dll";
+#endif
+
 static inline std::string join(const std::string& part1,
                                const std::string& part2) {
   // directory separator
@@ -165,6 +171,8 @@ static inline void* GetDsoHandleFromSearchPath(const std::string& search_root,
 void* GetCublasDsoHandle() {
 #if defined(__APPLE__) || defined(__OSX__)
   return GetDsoHandleFromSearchPath(FLAGS_cuda_dir, "libcublas.dylib");
+#elif defined(_WIN32) && defined(PADDLE_WITH_CUDA)
+  return GetDsoHandleFromSearchPath(FLAGS_cuda_dir, win_cublas_lib);
 #else
   return GetDsoHandleFromSearchPath(FLAGS_cuda_dir, "libcublas.so");
 #endif
@@ -173,6 +181,8 @@ void* GetCublasDsoHandle() {
 void* GetCUDNNDsoHandle() {
 #if defined(__APPLE__) || defined(__OSX__)
   return GetDsoHandleFromSearchPath(FLAGS_cudnn_dir, "libcudnn.dylib", false);
+#elif defined(_WIN32) && defined(PADDLE_WITH_CUDA)
+  return GetDsoHandleFromSearchPath(FLAGS_cudnn_dir, win_cudnn_lib);
 #else
   return GetDsoHandleFromSearchPath(FLAGS_cudnn_dir, "libcudnn.so", false);
 #endif
@@ -193,6 +203,8 @@ void* GetCUPTIDsoHandle() {
 void* GetCurandDsoHandle() {
 #if defined(__APPLE__) || defined(__OSX__)
   return GetDsoHandleFromSearchPath(FLAGS_cuda_dir, "libcurand.dylib");
+#elif defined(_WIN32) && defined(PADDLE_WITH_CUDA)
+  return GetDsoHandleFromSearchPath(FLAGS_cuda_dir, win_curand_lib);
 #else
   return GetDsoHandleFromSearchPath(FLAGS_cuda_dir, "libcurand.so");
 #endif

paddle/fluid/platform/timer.cc

+/* 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. */
+
+#include "paddle/fluid/platform/timer.h"
+
+namespace paddle {
+namespace platform {
+
+void Timer::Reset() {
+  _start.tv_sec = 0;
+  _start.tv_usec = 0;
+
+  _count = 0;
+  _elapsed = 0;
+  _paused = true;
+}
+
+void Timer::Start() {
+  Reset();
+  Resume();
+}
+
+void Timer::Pause() {
+  if (_paused) {
+    return;
+  }
+  _elapsed += Tickus();
+  ++_count;
+  _paused = true;
+}
+
+void Timer::Resume() {
+  gettimeofday(&_start, NULL);
+  _paused = false;
+}
+
+int Timer::Count() { return _count; }
+
+double Timer::ElapsedUS() { return static_cast<double>(_elapsed); }
+
+double Timer::ElapsedMS() { return _elapsed / 1000.0; }
+
+double Timer::ElapsedSec() { return _elapsed / 1000000.0; }
+
+int64_t Timer::Tickus() {
+  gettimeofday(&_now, NULL);
+  return (_now.tv_sec - _start.tv_sec) * 1000 * 1000L +
+         (_now.tv_usec - _start.tv_usec);
+}
+
+}  // namespace platform
+}  // namespace paddle

paddle/fluid/platform/timer.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 <stdlib.h>
+#include "paddle/fluid/platform/port.h"
+
+#ifdef _WIN32
+static unsigned sleep(unsigned seconds) {
+  Sleep(seconds * 1000);
+  return 0;
+}
+#endif
+
+namespace paddle {
+namespace platform {
+
+// A Standard Timer implementation for debugging
+class Timer {
+ public:
+  // a timer class for profiling
+  // Reset() will be called during initialization
+  // all timing variables will be set 0 in Reset()
+  Timer() { Reset(); }
+  void Reset();
+  void Start();
+  void Pause();
+  // Resume will get current system time
+  void Resume();
+  int Count();
+  // return elapsed time in us
+  double ElapsedUS();
+  // return elapsed time in ms
+  double ElapsedMS();
+  // return elapsed time in sec
+  double ElapsedSec();
+
+ private:
+  struct timeval _start;
+  struct timeval _now;
+  int _count;
+  int _elapsed;
+  bool _paused;
+
+  // get us difference between start and now
+  int64_t Tickus();
+};
+
+}  // namespace platform
+}  // namespace paddle

paddle/fluid/platform/timer_test.cc

+//  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.
+#include "paddle/fluid/platform/timer.h"
+#include "gtest/gtest.h"
+
+TEST(Timer, Reset) {
+  paddle::platform::Timer timeline;
+  timeline.Start();
+  sleep(3);
+  timeline.Pause();
+  timeline.Reset();
+}
+
+TEST(Timer, Start) {
+  paddle::platform::Timer timeline;
+  timeline.Start();
+  sleep(3);
+  timeline.Pause();
+}
+
+TEST(Timer, Pause) {
+  paddle::platform::Timer timeline;
+  timeline.Start();
+  sleep(3);
+  timeline.Pause();
+}
+
+TEST(Timer, Resume) {
+  paddle::platform::Timer timeline;
+  timeline.Start();
+  sleep(3);
+  timeline.Pause();
+  timeline.Resume();
+}

paddle/fluid/pybind/pybind.cc

@@ -84,11 +84,15 @@ bool IsCompiledWithCUDA() {
 }
 
 bool IsCompiledWithBrpc() {
-#if defined(PADDLE_WITH_BRPC) || defined(PADDLE_WITH_BRPC_RDMA)
-  return true;
-#else
+#ifndef PADDLE_WITH_DISTRIBUTE
   return false;
 #endif
+
+#ifdef PADDLE_WITH_GRPC
+  return false;
+#endif
+
+  return true;
 }
 
 bool IsCompiledWithDIST() {

paddle/testing/paddle_gtest_main.cc

@@ -28,20 +28,53 @@ int main(int argc, char** argv) {
   for (int i = 0; i < argc; ++i) {
     new_argv.push_back(argv[i]);
   }
+
+  std::vector<std::string> envs;
+  std::vector<std::string> undefok;
+#if defined(PADDLE_WITH_DISTRIBUTE) && !defined(PADDLE_WITH_GRPC)
+  envs.push_back("max_body_size");
+#endif
+
 #if defined(PADDLE_WITH_CUDA) || defined(PADDLE_WITH_HIP)
-  new_argv.push_back(
-      strdup("--tryfromenv=fraction_of_gpu_memory_to_use,allocator_strategy"));
+  envs.push_back("fraction_of_gpu_memory_to_use");
+  envs.push_back("allocator_strategy");
 #elif __clang__
-  new_argv.push_back(
-      strdup("--tryfromenv=use_mkldnn,initial_cpu_memory_in_"
-             "mb,allocator_strategy"));
-  new_argv.push_back(strdup("--undefok=use_mkldnn,initial_cpu_memory_in_mb"));
+  envs.push_back("use_mkldnn");
+  envs.push_back("initial_cpu_memory_in_mb");
+  envs.push_back("allocator_strategy");
+
+  undefok.push_back("use_mkldnn");
+  undefok.push_back("initial_cpu_memory_in_mb");
 #else
-  new_argv.push_back(
-      strdup("--tryfromenv=use_pinned_memory,use_mkldnn,initial_cpu_memory_in_"
-             "mb,allocator_strategy"));
-  new_argv.push_back(strdup("--undefok=use_mkldnn,initial_cpu_memory_in_mb"));
+  envs.push_back("use_pinned_memory");
+  envs.push_back("use_mkldnn");
+  envs.push_back("initial_cpu_memory_in_mb");
+  envs.push_back("allocator_strategy");
+
+  undefok.push_back("use_mkldnn");
+  undefok.push_back("initial_cpu_memory_in_mb");
 #endif
+
+  if (envs.size() > 0) {
+    std::string env_string = "--tryfromenv=";
+    for (auto t : envs) {
+      env_string += t + ",";
+    }
+    env_string = env_string.substr(0, env_string.length() - 1);
+    new_argv.push_back(strdup(env_string.c_str()));
+    VLOG(1) << "gtest env_string:" << env_string;
+  }
+
+  if (undefok.size() > 0) {
+    std::string undefok_string = "--undefok=";
+    for (auto t : undefok) {
+      undefok_string += t + ",";
+    }
+    undefok_string = undefok_string.substr(0, undefok_string.length() - 1);
+    new_argv.push_back(strdup(undefok_string.c_str()));
+    VLOG(1) << "gtest undefok_string:" << undefok_string;
+  }
+
   int new_argc = static_cast<int>(new_argv.size());
   char** new_argv_address = new_argv.data();
   google::ParseCommandLineFlags(&new_argc, &new_argv_address, false);

python/paddle/fluid/__init__.py

@@ -151,12 +151,21 @@ def __bootstrap__():
         read_env_flags.append('rpc_get_thread_num')
         read_env_flags.append('rpc_prefetch_thread_num')
         read_env_flags.append('rpc_disable_reuse_port')
+        if core.is_compiled_with_brpc():
+            read_env_flags.append('max_body_size')
+            #set brpc max body size
+            os.environ['FLAGS_max_body_size'] = "2147483647"
 
     if core.is_compiled_with_cuda():
         read_env_flags += [
-            'fraction_of_gpu_memory_to_use', 'cudnn_deterministic',
-            'enable_cublas_tensor_op_math', 'conv_workspace_size_limit',
-            'cudnn_exhaustive_search', 'memory_optimize_debug', 'selected_gpus'
+            'fraction_of_gpu_memory_to_use',
+            'cudnn_deterministic',
+            'enable_cublas_tensor_op_math',
+            'conv_workspace_size_limit',
+            'cudnn_exhaustive_search',
+            'memory_optimize_debug',
+            'selected_gpus',
+            'cudnn_exhaustive_search_times',
         ]
 
     core.init_gflags([sys.argv[0]] +

python/paddle/fluid/data_feeder.py

@@ -272,8 +272,7 @@ class DataFeeder(object):
             dict: the result of conversion.
 
         Raises:
-            ValueError: If drop_last is False and the data batch which cannot
-            fit for devices.
+            ValueError: If drop_last is False and the data batch which cannot fit for devices.
         """
 
         def __reader_creator__():

python/paddle/fluid/framework.py

@@ -647,20 +647,16 @@ class Operator(object):
                     self.desc.set_input(in_proto.name, [])
 
         if outputs is not None:
-            given = set()
-            need = set()
-            for n in outputs:
-                given.add(n)
             for m in proto.outputs:
-                need.add(m.name)
-            if not given == need:
-                raise ValueError(("Incorrect setting for output(s) of "
-                                  "operator \"%s\". Need: [%s] Given: [%s]") %
-                                 (type,
-                                  ", ".join(six.binary_type(e) for e in need),
-                                  ", ".join(six.binary_type(e) for e in given)))
-
+                if (m.name not in outputs) and m.dispensable:
+                    continue
+                if not ((m.name in outputs) or m.dispensable):
+                    raise ValueError(
+                        ("Incorrect setting for output(s) of "
+                         "operator \"%s\", should set: [%s].") % (type, m.name))
             for out_proto in proto.outputs:
+                if out_proto.name not in outputs:
+                    continue
                 out_args = outputs[out_proto.name]
                 if not isinstance(out_args, list):
                     out_args = [out_args]
@@ -1638,8 +1634,8 @@ class Program(object):
                 parameters, e.g., :code:`trainable`, :code:`optimize_attr`, need
                 to print.
 
-        Returns
-            (str): The debug string.
+        Returns:
+            str : The debug string.
 
         Raises:
             ValueError: If any of required fields is not set and throw_on_error is

python/paddle/fluid/layers/control_flow.py

@@ -1452,6 +1452,7 @@ class DynamicRNN(object):
     def step_input(self, x):
         """
         Mark a sequence as a dynamic RNN input.
+
         Args:
             x(Variable): The input sequence.
 
@@ -1505,6 +1506,7 @@ class DynamicRNN(object):
         """
         Mark a variable as a RNN input. The input will not be scattered into
         time steps.
+
         Args:
             x(Variable): The input variable.
 
@@ -1629,13 +1631,11 @@ class DynamicRNN(object):
         Args:
             init(Variable|None): The initialized variable.
 
-            shape(list|tuple): The memory shape. NOTE the shape does not contain
-            batch_size.
+            shape(list|tuple): The memory shape. NOTE the shape does not contain batch_size.
 
             value(float): the initalized value.
 
-            need_reorder(bool): True if the initialized memory depends on the
-            input sample.
+            need_reorder(bool): True if the initialized memory depends on the input sample.
 
             dtype(str|numpy.dtype): The data type of the initialized memory.
 
@@ -1714,6 +1714,7 @@ class DynamicRNN(object):
         """
         Update the memory from ex_mem to new_mem. NOTE that the shape and data
         type of :code:`ex_mem` and :code:`new_mem` must be same.
+        
         Args:
             ex_mem(Variable): the memory variable.
             new_mem(Variable): the plain variable generated in RNN block.

python/paddle/fluid/layers/detection.py

@@ -65,7 +65,7 @@ def rpn_target_assign(bbox_pred,
                       rpn_negative_overlap=0.3,
                       use_random=True):
     """
-    ** Target Assign Layer for region proposal network (RPN) in Faster-RCNN detection. **
+    **Target Assign Layer for region proposal network (RPN) in Faster-RCNN detection.**
 
     This layer can be, for given the  Intersection-over-Union (IoU) overlap
     between anchors and ground truth boxes, to assign classification and
@@ -148,6 +148,7 @@ def rpn_target_assign(bbox_pred,
                                               cls_logits=cls_logits,
                                               anchor_box=anchor_box,
                                               gt_boxes=gt_boxes)
+
     """
 
     helper = LayerHelper('rpn_target_assign', **locals())
@@ -1525,20 +1526,23 @@ def anchor_generator(input,
                                         anchors, e.g. [0.5, 1.0, 2.0].
        variance(list|tuple): The variances to be used in box regression deltas.
                              Default:[0.1, 0.1, 0.2, 0.2].
-       stride(list|turple): The anchors stride across width and height,
-            e.g. [16.0, 16.0]
+       stride(list|turple): The anchors stride across width and height,e.g. [16.0, 16.0]
        offset(float): Prior boxes center offset. Default: 0.5
        name(str): Name of the prior box op. Default: None.
 
     Returns:
-        Anchors(Variable):  The output anchors with a layout of [H, W, num_anchors, 4].
-              H is the height of input, W is the width of input,
-              num_anchors is the box count of each position.
+        Anchors(Variable),Variances(Variable):  
+        
+              two variables:
+        
+              - Anchors(Variable): The output anchors with a layout of [H, W, num_anchors, 4]. \
+                H is the height of input, W is the width of input, \
+                num_anchors is the box count of each position.  \
                 Each anchor is in (xmin, ymin, xmax, ymax) format an unnormalized. 
-        Variances(Variable): The expanded variances of anchors
-              with a layout of [H, W, num_priors, 4].
-              H is the height of input, W is the width of input
-              num_anchors is the box count of each position.
+              - Variances(Variable): The expanded variances of anchors \
+                with a layout of [H, W, num_priors, 4]. \
+                H is the height of input, W is the width of input \
+                num_anchors is the box count of each position. \
                 Each variance is in (xcenter, ycenter, w, h) format.
 
 
@@ -1748,7 +1752,7 @@ def generate_proposals(scores,
                        eta=1.0,
                        name=None):
     """
-    ** Generate proposal Faster-RCNN **
+    **Generate proposal Faster-RCNN**
 
     This operation proposes RoIs according to each box with their probability to be a foreground object and 
     the box can be calculated by anchors. Bbox_deltais and scores to be an object are the output of RPN. Final proposals
@@ -1762,7 +1766,6 @@ def generate_proposals(scores,
     4. Remove predicted boxes with small area. 
     5. Apply NMS to get final proposals as output.
 
-      
     Args:
         scores(Variable): A 4-D Tensor with shape [N, A, H, W] represents the probability for each box to be an object.
             N is batch size, A is number of anchors, H and W are height and width of the feature map.
@@ -1777,6 +1780,7 @@ def generate_proposals(scores,
         nms_thresh(float): Threshold in NMS, 0.5 by default.
         min_size(float): Remove predicted boxes with either height or width < min_size. 0.1 by default.
         eta(float): Apply in adaptive NMS, if adaptive threshold > 0.5, adaptive_threshold = adaptive_threshold * eta in each iteration.
+
     """
     helper = LayerHelper('generate_proposals', **locals())
 

python/paddle/fluid/layers/io.py

@@ -949,12 +949,11 @@ def shuffle(reader, buffer_size):
     is determined by argument buf_size.
 
     Args:
-        param reader: the original reader whose output will be shuffled.
-        type reader: callable
-        param buf_size: shuffle buffer size.
-        type buf_size: int
-        return: the new reader whose output is shuffled.
-        rtype: callable
+        reader(callable): the original reader whose output will be shuffled.
+        buf_size(int): shuffle buffer size.
+
+    Returns:
+        callable: the new reader whose output is shuffled.
     """
     return __create_unshared_decorated_reader__(
         'create_shuffle_reader', reader, {'buffer_size': int(buffer_size)})

python/paddle/fluid/layers/nn.py

@@ -233,7 +233,7 @@ def fc(input,
             dimensions will be flatten to form the first dimension of the final matrix (height of
             the matrix), and the rest `rank(X) - num_flatten_dims` dimensions are flattened to
             form the second dimension of the final matrix (width of the matrix). For example, suppose
-            `X` is a 6-dimensional tensor with a shape [2, 3, 4, 5, 6], and `num_flatten_dims` = 3.
+            `X` is a 5-dimensional tensor with a shape [2, 3, 4, 5, 6], and `num_flatten_dims` = 3.
             Then, the flattened matrix will have a shape [2 x 3 x 4, 5 x 6] = [24, 30].
         param_attr (ParamAttr|list of ParamAttr, default None): The parameter attribute for learnable
             parameters/weights of this layer.
@@ -505,31 +505,33 @@ def lstm(input,
     In the forward pass the output ht and cell output ct for a given iteration can be computed from the recurrent input ht-1, 
     the cell input ct-1 and the previous layer input xt given matrices W, R and biases bW, bR from the following equations:
 
-    $$ i_t = \\sigma(W_{ix}x_{t} + W_{ih}h_{t-1} + bx_i + bh_i) $$
+    .. math::
+    
+       i_t &= \sigma(W_{ix}x_{t} + W_{ih}h_{t-1} + bx_i + bh_i) 
        
-    $$ f_t = \\sigma(W_{fx}x_{t} + W_{fh}h_{t-1} + bx_f + bh_f) $$
+       f_t &= \sigma(W_{fx}x_{t} + W_{fh}h_{t-1} + bx_f + bh_f) 
        
-    $$ o_t = \\sigma(W_{ox}x_{t} + W_{oh}h_{t-1} + bx_o + bh_o) $$
+       o_t &= \sigma(W_{ox}x_{t} + W_{oh}h_{t-1} + bx_o + bh_o) 
        
-    $$ \\tilde{c_t} = tanh(W_{cx}x_t + W_{ch}h_{t-1} + bx_c + bh_c) $$
+       \\tilde{c_t} &= tanh(W_{cx}x_t + W_{ch}h_{t-1} + bx_c + bh_c)
        
-    $$ c_t = f_t \\odot c_{t-1} + i_t \\odot \\tilde{c_t} $$
+       c_t &= f_t \odot c_{t-1} + i_t \odot \\tilde{c_t} 
        
-    $$ h_t = o_t \\odot tanh(c_t) $$
+       h_t &= o_t \odot tanh(c_t) 
 
-    - W terms denote weight matrices (e.g. $W_{ix}$ is the matrix
+    - $W$ terms denote weight matrices (e.g. $W_{ix}$ is the matrix
       of weights from the input gate to the input)
     - The b terms denote bias vectors ($bx_i$ and $bh_i$ are the input gate bias vector).
     - sigmoid is the logistic sigmoid function.
     - $i, f, o$ and $c$ are the input gate, forget gate, output gate,
       and cell activation vectors, respectively, all of which have the same size as
       the cell output activation vector $h$.
-    - The $\odot$ is the element-wise product of the vectors.
-    - `tanh` is the activation functions.
-    - $\tilde{c_t}$ is also called candidate hidden state,
+    - The :math:`\odot` is the element-wise product of the vectors.
+    - :math:`tanh` is the activation functions.
+    - :math:`\\tilde{c_t}` is also called candidate hidden state,
       which is computed based on the current input and the previous hidden state.
 
-    Where sigmoid is the sigmoid operator: sigmoid(x) = 1 / (1 + e^-x), * represents a point-wise multiplication,
+    Where sigmoid is the sigmoid operator: :math:`sigmoid(x) = 1 / (1 + e^{-x})` , * represents a point-wise multiplication, 
     X represensts a matrix multiplication
 
 
@@ -556,13 +558,17 @@ def lstm(input,
 
 
     Returns:
-        rnn_out(Tensor): result of LSTM hidden, shape is (seq_len x batch_size x hidden_size)
+        rnn_out(Tensor),last_h(Tensor),last_c(Tensor):  
+                        
+                        Three tensors, rnn_out, last_h, last_c:
+                        
+                        - rnn_out is result of LSTM hidden, shape is (seq_len x batch_size x hidden_size) \
                           if is_bidirec set to True, shape will be ( seq_len x batch_sze x hidden_size*2)
-        last_h(Tensor): the hidden state of the last step of LSTM
-                        shape is ( num_layers x batch_size x hidden_size )
+                        - last_h is the hidden state of the last step of LSTM \
+                          shape is ( num_layers x batch_size x hidden_size ) \
                           if is_bidirec set to True, shape will be ( num_layers*2 x batch_size x hidden_size)                     
-        last_c(Tensor): the cell state of the last step of LSTM
-                        shape is ( num_layers x batch_size x hidden_size )
+                        - last_c(Tensor): the cell state of the last step of LSTM \
+                          shape is ( num_layers x batch_size x hidden_size ) \
                           if is_bidirec set to True, shape will be ( num_layers*2 x batch_size x hidden_size)                     
 
 
@@ -1220,6 +1226,8 @@ def dropout(x,
     probability) the outputs of some units to zero, while others are remain
     unchanged.
 
+    dropout op can be removed from the program to make the program more efficient.
+
     Args:
         x (Variable): The input tensor variable.
         dropout_prob (float): Probability of setting units to zero.
@@ -1230,20 +1238,22 @@ def dropout(x,
                     units will be dropped. DO NOT use a fixed seed in training.
         name (str|None): A name for this layer(optional). If set None, the layer
                          will be named automatically.
-        dropout_implementation(string): ['downgrade_in_infer'(defauld)|'upscale_in_train']
+        dropout_implementation(string): ['downgrade_in_infer'(default)|'upscale_in_train']
+
                                         1. downgrade_in_infer(default), downgrade the outcome at inference
-                                           train: out = input * mask
-                                           inference: out = input * dropout_prob
-                                           (make is a tensor same shape with input, value is 0 or 1
+
+                                           - train: out = input * mask
+                                           - inference: out = input * dropout_prob
+
+                                           (mask is a tensor same shape with input, value is 0 or 1
                                            ratio of 0 is dropout_prob)
                                         2. upscale_in_train, upscale the outcome at training time
-                                           train: out = input * mask / ( 1.0 - dropout_prob )
-                                           inference: out = input
-                                           (make is a tensor same shape with input, value is 0 or 1
-                                            ratio of 0 is dropout_prob)
-                                           dropout op can be removed from the program.
-                                           the program will be efficient
 
+                                           - train: out = input * mask / ( 1.0 - dropout_prob )
+                                           - inference: out = input
+
+                                           (mask is a tensor same shape with input, value is 0 or 1
+                                           ratio of 0 is dropout_prob)
 
                                         
     Returns:
@@ -1333,11 +1343,15 @@ def cross_entropy(input, label, soft_label=False, ignore_index=kIgnoreIndex):
          A 2-D tensor with shape [N x 1], the cross entropy loss.
 
     Raises:
-        `ValueError`: 1) the 1st dimension of `input` and `label` are not equal.
-                      2) when `soft_label == True`, and the 2nd dimension of
-                         `input` and `label` are not equal.
-                      3) when `soft_label == False`, and the 2nd dimension of
-                         `label` is not 1.
+         ValueError:
+
+                      1. the 1st dimension of ``input`` and ``label`` are not equal.
+                      
+                      2. when ``soft_label == True``, and the 2nd dimension of
+                         ``input`` and ``label`` are not equal.
+                         
+                      3. when ``soft_label == False``, and the 2nd dimension of
+                         ``label`` is not 1.
 
     Examples:
         .. code-block:: python
@@ -1458,7 +1472,7 @@ def chunk_eval(input,
     F1-score of chunk detection.
 
     For some basics of chunking, please refer to 
-    'Chunking with Support Vector Machines <https://aclanthology.info/pdf/N/N01/N01-1025.pdf>'.
+    `Chunking with Support Vector Machines <https://aclanthology.info/pdf/N/N01/N01-1025.pdf>`_ .
 
     ChunkEvalOp computes the precision, recall, and F1-score of chunk detection,
     and supports IOB, IOE, IOBES and IO (also known as plain) tagging schemes.
@@ -2292,7 +2306,8 @@ def sequence_slice(input, offset, length, name=None):
                 out.lod = [[2, 1]],
                 out.dims = (3, 2).
 
-    NOTE: The first dimension size of **input**, **offset** and **length**
+    Note: 
+          The first dimension size of **input**, **offset** and **length**
           should be equal. The **offset** should start from 0.
 
     Args:
@@ -2570,12 +2585,7 @@ def adaptive_pool2d(input,
         raise ValueError(
             "invalid setting 'require_index' true when 'pool_type' is 'avg'.")
 
-    def _is_list_or_tuple_(data):
-        return (isinstance(data, list) or isinstance(data, tuple))
-
-    if not _is_list_or_tuple_(pool_size) or len(pool_size) != 2:
-        raise ValueError(
-            "'pool_size' should be a list or tuple with length as 2.")
+    pool_size = utils.convert_to_list(pool_size, 2, 'pool_size')
 
     if pool_type == "max":
         l_type = 'max_pool2d_with_index'
@@ -2671,12 +2681,7 @@ def adaptive_pool3d(input,
         raise ValueError(
             "invalid setting 'require_index' true when 'pool_type' is 'avg'.")
 
-    def _is_list_or_tuple_(data):
-        return (isinstance(data, list) or isinstance(data, tuple))
-
-    if not _is_list_or_tuple_(pool_size) or len(pool_size) != 3:
-        raise ValueError(
-            "'pool_size' should be a list or tuple with length as 3.")
+    pool_size = utils.convert_to_list(pool_size, 3, 'pool_size')
 
     if pool_type == "max":
         l_type = 'max_pool3d_with_index'
@@ -3013,7 +3018,7 @@ def group_norm(input,
     """
     **Group Normalization Layer**
 
-    Refer to `Group Normalization <https://arxiv.org/abs/1803.08494>`
+    Refer to `Group Normalization <https://arxiv.org/abs/1803.08494>`_ .
 
     Args:
         input(Variable): The input tensor variable.
@@ -3140,8 +3145,8 @@ def conv2d_transpose(input,
 
            H^\prime_{out} &= (H_{in} - 1) * strides[0] - 2 * paddings[0] + dilations[0] * (H_f - 1) + 1 \\\\
            W^\prime_{out} &= (W_{in} - 1) * strides[1] - 2 * paddings[1] + dilations[1] * (W_f - 1) + 1 \\\\
-           H_{out} \in [ H^\prime_{out}, H^\prime_{out} + strides[0] ) \\\\
-           W_{out} \in [ W^\prime_{out}, W^\prime_{out} + strides[1] )
+           H_{out} &\in [ H^\prime_{out}, H^\prime_{out} + strides[0] ) \\\\
+           W_{out} &\in [ W^\prime_{out}, W^\prime_{out} + strides[1] )
 
     Args:
         input(Variable): The input image with [N, C, H, W] format.
@@ -4704,9 +4709,9 @@ def ctc_greedy_decoder(input, blank, name=None):
         name (str): The name of this layer. It is optional.
 
     Returns:
-        Variable: CTC greedy decode result which is a 2-D tensor with shape [Lp, 1].
-                  'Lp' is the sum if all output sequences' length. If all the sequences
-                  in result were empty, the result LoDTensor will be [-1] with
+        Variable: CTC greedy decode result which is a 2-D tensor with shape [Lp, 1]. \
+                  'Lp' is the sum if all output sequences' length. If all the sequences \
+                  in result were empty, the result LoDTensor will be [-1] with  \
                   LoD [[]] and dims [1, 1]. 
 
     Examples:
@@ -5072,6 +5077,7 @@ def hsigmoid(input,
     <http://www.iro.umontreal.ca/~lisa/pointeurs/hierarchical-nnlm-aistats05.pdf>`_
 
     And if you want to use the costumed tree by set 'is_custom' as true you may need to do following things first:
+
     1. using your word dict to build a binary tree, each leaf node should be an word of your word dict
     2. build a dict to store word_id -> word's leaf to root path, we call it path_table.
     3. build a dict to store word_id -> code of word's leaf to root path, we call it path_code. Code
@@ -5079,7 +5085,6 @@ def hsigmoid(input,
     4. now, each word should has its path and code along the path, you can pass a batch of path and code 
        related to the same batch of inputs.
 
-
     Args:
         input (Variable): The input tensor variable with shape
             :math:`[N \\times D]`, where :math:`N` is the size of mini-batch,
@@ -5485,11 +5490,11 @@ def softmax_with_cross_entropy(logits,
 
     .. math::
 
-        max_j = \\max_{i=0}^{K}{\\text{logit}_i}
+        max_j &= \\max_{i=0}^{K}{\\text{logit}_i}
 
-        log\\_max\\_sum_j = \\log\\sum_{i=0}^{K}\\exp(logit_i - max_j)
+        log\\_max\\_sum_j &= \\log\\sum_{i=0}^{K}\\exp(logit_i - max_j)
 
-        softmax_j = \\exp(logit_j - max_j - {log\\_max\\_sum}_j)
+        softmax_j &= \\exp(logit_j - max_j - {log\\_max\\_sum}_j)
 
     and then cross entropy loss is calculated by softmax and label.
 
@@ -5515,10 +5520,10 @@ def softmax_with_cross_entropy(logits,
                                along with the cross entropy loss. Default: False
 
     Returns:
-        Variable or Tuple of two Variables: Return the cross entropy loss if
-                              `return_softmax` is False, otherwise the tuple
-                              (loss, softmax), where the cross entropy loss is
-                              a 2-D tensor with shape [N x 1], and softmax is a
+        Variable or Tuple of two Variables: Return the cross entropy loss if \
+                                            `return_softmax` is False, otherwise the tuple \
+                                            (loss, softmax), where the cross entropy loss is \
+                                            a 2-D tensor with shape [N x 1], and softmax is a \
                                             2-D tensor with shape [N x K].
 
     Examples:
@@ -5792,15 +5797,21 @@ def squeeze(input, axes, name=None):
     the single dimensions will be removed from the shape. If an axis is
     selected with shape entry not equal to one, an error is raised.
 
-    Examples:
+    For example:
+
+    .. code-block:: text
+
         Case 1:
+
           Given
             X.shape = (1, 3, 1, 5)
           and
             axes = [0]
           we get:
             Out.shape = (3, 1, 5)
+
         Case 2:
+
           Given
             X.shape = (1, 3, 1, 5)
           and
@@ -5842,6 +5853,9 @@ def unsqueeze(input, axes, name=None):
     Dimension indices in axes are as seen in the output tensor.
 
     For example:
+
+    .. code-block:: text
+
       Given a tensor such that tensor with shape [3, 4, 5],
       then Unsqueezed tensor with axes=[0, 4] has shape [1, 3, 4, 5, 1].
 
@@ -6729,8 +6743,11 @@ def sequence_scatter(input, index, updates, name=None):
     the columns to update in each row of X.
 
     Here is an example:
+
     Given the following input:
+
     .. code-block:: text
+
         input.data = [[1.0, 1.0, 1.0, 1.0, 1.0, 1.0],
                       [1.0, 1.0, 1.0, 1.0, 1.0, 1.0],
                       [1.0, 1.0, 1.0, 1.0, 1.0, 1.0]]
@@ -6743,7 +6760,9 @@ def sequence_scatter(input, index, updates, name=None):
         updates.lod =  [[  0,            3,                                 8,                         12]]
 
     Then we have the output:
+
     .. code-block:: text
+
         out.data = [[1.3, 1.3, 1.4, 1.0, 1.0, 1.0],
                     [1.0, 1.0, 1.4, 1.3, 1.2, 1.1],
                     [1.0, 1.0, 1.3, 1.2, 1.4, 1.1]]
@@ -6759,7 +6778,7 @@ def sequence_scatter(input, index, updates, name=None):
         name (str|None): The output variable name. Default None.
 
     Returns:
-        output (Variable): The output is a tensor with the same shape as input.
+        Variable: The output is a tensor with the same shape as input.
 
     Examples:
 
@@ -6933,7 +6952,7 @@ def mean_iou(input, label, num_classes):
 
     .. math::
 
-        IOU = \\frac{true\_positiv}{(true\_positive + false\_positive + false\_negative)}.
+        IOU = \\frac{true\_positive}{(true\_positive + false\_positive + false\_negative)}.
 
     The predictions are accumulated in a confusion matrix and mean-IOU
     is then calculated from it.
@@ -6946,9 +6965,13 @@ def mean_iou(input, label, num_classes):
         num_classes (int): The possible number of labels.
 
     Returns:
-        mean_iou (Variable): A Tensor representing the mean intersection-over-union with shape [1].
-        out_wrong(Variable): A Tensor with shape [num_classes]. The wrong numbers of each class.
-        out_correct(Variable): A Tensor with shape [num_classes]. The correct numbers of each class.
+        mean_iou (Variable),out_wrong(Variable),out_correct(Variable): 
+        
+                     Three variables:
+                      
+                     - mean_iou : A Tensor representing the mean intersection-over-union with shape [1].
+                     - out_wrong: A Tensor with shape [num_classes]. The wrong numbers of each class.
+                     - out_correct: A Tensor with shape [num_classes]. The correct numbers of each class.
 
     Examples:
 
@@ -7144,7 +7167,7 @@ def affine_grid(theta, out_shape, name=None):
     Args:
         theta (Variable): A batch of affine transform parameters with shape [N, 2, 3].
         out_shape (Variable | list | tuple): The shape of target output with format [N, C, H, W]. 
-        out_shape can be a Variable or a list or tuple.
+                                             ``out_shape`` can be a Variable or a list or tuple.
         name(str|None): A name for this layer(optional). If set None, the layer
                         will be named automatically.
 
@@ -7157,6 +7180,7 @@ def affine_grid(theta, out_shape, name=None):
     Examples:
 
         .. code-block:: python
+
             theta = fluid.layers.data(name="x", shape=[2, 3], dtype="float32")
             out_shape = fluid.layers.data(name="y", shape=[-1], dtype="float32")
             data = fluid.layers.affine_grid(theta, out_shape)
@@ -7192,9 +7216,10 @@ def affine_grid(theta, out_shape, name=None):
 
 def rank_loss(label, left, right, name=None):
     """
+
     **Rank loss layer for RankNet**
 
-    RankNet(http://icml.cc/2015/wp-content/uploads/2015/06/icml_ranking.pdf)
+    `RankNet <http://icml.cc/2015/wp-content/uploads/2015/06/icml_ranking.pdf>`_
     is a pairwise ranking model with a training sample consisting of a pair
     of documents, A and B. Label P indicates whether A is ranked higher than B
     or not:
@@ -7202,16 +7227,19 @@ def rank_loss(label, left, right, name=None):
     P = {0, 1} or {0, 0.5, 1}, where 0.5 means that there is no information
     about the rank of the input pair.
 
-    Rank loss layer takes three inputs: left (o_i), right (o_j) and
-    label (P_{i,j}). The inputs respectively represent RankNet's output scores
+    Rank loss layer takes three inputs: left ( :math:`o_i` ), right ( :math:`o_j` ) and
+    label ( :math:`P_{i,j}` ). The inputs respectively represent RankNet's output scores
     for documents A and B and the value of label P. The following equation
     computes rank loss C_{i,j} from the inputs:
 
-    $$
-      C_{i,j} = -\tilde{P_{ij}} * o_{i,j} + \log(1 + e^{o_{i,j}}) \\
-      o_{i,j} =  o_i - o_j  \\
-      \tilde{P_{i,j}} = \left \{0, 0.5, 1 \right \} \ or \ \left \{0, 1 \right \}
-    $$
+    .. math::
+
+      C_{i,j} &= -\\tilde{P_{ij}} * o_{i,j} + \log(1 + e^{o_{i,j}}) \\\\
+
+      o_{i,j} &=  o_i - o_j  \\\\
+
+      \\tilde{P_{i,j}} &= \\left \{0, 0.5, 1 \\right \} \ or \ \\left \{0, 1 \\right \}
+
 
     Rank loss layer takes batch inputs with size batch_size (batch_size >= 1).
 
@@ -7237,7 +7265,6 @@ def rank_loss(label, left, right, name=None):
             right = fluid.layers.data(name="right", shape=[4, 1], dtype="float32")
             out = fluid.layers.rank_loss(label, left, right)
 
-
     """
     helper = LayerHelper('rank_loss', **locals())
 
@@ -7269,7 +7296,7 @@ def margin_rank_loss(label, left, right, margin=0.1, name=None):
 
     .. math::
 
-        rank\_loss &= max(0, -label * (left - right) + margin)
+        rank\_loss = max(0, -label * (left - right) + margin)
 
     Args:
        label (Variable): Indicates whether the left is ranked higher than the right or not.
@@ -7278,12 +7305,17 @@ def margin_rank_loss(label, left, right, margin=0.1, name=None):
        margin (float): Indicates the given margin.
        name (str|None): A name for this layer (optional). If set None, the layer
                        will be named automatically.
+
     Returns:
        Variable: The ranking loss.
+
     Raises:
        ValueError: Any of label, left, and right is not a Variable.
+
     Examples:
+
         .. code-block:: python
+
            label = fluid.layers.data(name="label", shape=[4, 1], dtype="float32")
            left = fluid.layers.data(name="left", shape=[4, 1], dtype="float32")
            right = fluid.layers.data(name="right", shape=[4, 1], dtype="float32")
@@ -7587,7 +7619,8 @@ def prelu(x, mode, param_attr=None, name=None):
     """
     Equation:
 
-        y = \max(0, x) + alpha * \min(0, x)
+    .. math::
+        y = \max(0, x) + \\alpha * \min(0, x)
 
     Args:
         x (Variable): The input tensor.
@@ -7730,20 +7763,29 @@ def flatten(x, axis=1, name=None):
     **Flatten layer**
     Flattens the input tensor into a 2D matrix.
     
-    Examples:
+    For Example:
+    
+    .. code-block:: text
+
         Case 1:
+
           Given
             X.shape = (3, 100, 100, 4)
+
           and
             axis = 2
+
           We get:
             Out.shape = (3 * 100, 4 * 100)
 
         Case 2:
+
           Given
             X.shape = (3, 100, 100, 4)
+
           and
             axis = 0
+
           We get:
             Out.shape = (1, 3 * 100 * 100 * 4)
 
@@ -7759,9 +7801,9 @@ def flatten(x, axis=1, name=None):
                         will be named automatically.
 
     Returns:
-        Variable: A 2D tensor with the contents of the input tensor, with input
-                  dimensions up to axis flattened to the outer dimension of
-                  the output and remaining input dimensions flattened into the
+        Variable: A 2D tensor with the contents of the input tensor, with input \
+                  dimensions up to axis flattened to the outer dimension of \
+                  the output and remaining input dimensions flattened into the \
                   inner dimension of the output.
 
     Raises:
@@ -7801,15 +7843,19 @@ def sequence_enumerate(input, win_size, pad_value=0, name=None):
     The enumerated sequence has the same 1st dimension with variable `input`, and
     the 2nd dimension is `win_size`, padded by `pad_value` if necessary in generation.
 
-    Examples:
+    .. code-block:: text
+
         Case 1:
+
           Input:
             X.lod = [[0, 3, 5]]
             X.data = [[1], [2], [3], [4], [5]]
             X.dims = [5, 1]
+
           Attrs:
             win_size = 2
             pad_value = 0
+
           Output:
             Out.lod = [[0, 3, 5]]
             Out.data = [[1, 2], [2, 3], [3, 0], [4, 5], [5, 0]]
@@ -8896,6 +8942,7 @@ def similarity_focus(input, axis, indexes, name=None):
     SimilarityFocus Operator
 
     Generate a similarity focus mask with the same shape of input using the following method:
+    
     1. Extract the 3-D tensor(here the first dimension is BatchSize) corresponding
        to the axis according to the indexes. For example, if axis=1 and indexes=[a],
        it will get the matrix T=X[:, a, :, :]. In this case, if the shape of input X
@@ -8969,14 +9016,16 @@ def similarity_focus(input, axis, indexes, name=None):
         indexes(list): Indicating the indexes of the selected dimension.
 
     Returns:
-        Variable: A tensor variable with the same shape and same type
+        Variable: A tensor variable with the same shape and same type \
                   as the input.
 
     Examples:
         .. code-block:: python
+
             data = fluid.layers.data(
               name='data', shape=[2, 3, 2, 2], dtype='float32')
             x = fluid.layers.layer_norm(input=data, axis=1, indexes=[0])
+
     """
     helper = LayerHelper('similarity_focus', **locals())
     # check attrs
@@ -9055,6 +9104,7 @@ def hash(input, hash_size, num_hash=1, name=None):
 
     Examples:
        .. code-block:: python
+
            word_dict = paddle.dataset.imdb.word_dict()
            x = fluid.layers.data(shape[1], dtype='int32', lod_level=1)
            out = fluid.layers.hash(input=x, num_hash=4, hash_size=1000)
@@ -9075,13 +9125,15 @@ def hash(input, hash_size, num_hash=1, name=None):
 def grid_sampler(x, grid, name=None):
     """
     This operation samples input X by using bilinear interpolation based on
-    flow field grid, which is usually gennerated by affine_grid. The grid of
+    flow field grid, which is usually gennerated by :code:`affine_grid` . The grid of
     shape [N, H, W, 2] is the concatenation of (grid_x, grid_y) coordinates
     with shape [N, H, W] each, where grid_x is indexing the 4th dimension
     (in width dimension) of input data x and grid_y is indexng the 3rd
     dimention (in height dimension), finally results is the bilinear
     interpolation value of 4 nearest corner points.
 
+    .. code-block:: text
+
         Step 1:
         Get (x, y) grid coordinates and scale to [0, H-1/W-1].
 
@@ -9126,16 +9178,18 @@ def grid_sampler(x, grid, name=None):
         name (str, default None): The name of this layer.
 
     Returns:
-        out(Variable): Output of shape [N, C, H, W] data samples input X
+        Variable: Output of shape [N, C, H, W] data samples input X
         using bilnear interpolation based on input grid.
 
-    Exmples:
+    Examples:
+
         .. code-block:: python
 
             x = fluid.layers.data(name='x', shape=[3, 10, 32, 32], dtype='float32')
             theta = fluid.layers.data(name='theta', shape=[3, 2, 3], dtype='float32')
             grid = fluid.layers.affine_grid(input=theta, size=[3, 10, 32, 32]})
             out = fluid.layers.grid_sampler(x=x, grid=grid)
+
     """
     helper = LayerHelper("grid_sampler", **locals())
 
@@ -9203,19 +9257,19 @@ def add_position_encoding(input, alpha, beta, name=None):
     """
     **Add Position Encoding Layer**
 
-    This layer accepts an input 3D-Tensor of shape [N x M x P], and return an
+    This layer accepts an input 3D-Tensor of shape [N x M x P], and returns an
     output Tensor of shape [N x M x P] with positional encoding value.
 
-    Refer to `Attention Is All You Need<http://arxiv.org/pdf/1706.03762.pdf>`_ .
+    Refer to `Attention Is All You Need <http://arxiv.org/pdf/1706.03762.pdf>`_ .
 
     .. math::
-        PE(pos, 2i) = \\sin{(pos / 10000^{2i / P})}   \\\\
-        PE(pos, 2i + 1) = \\cos{(pos / 10000^{2i / P})}  \\\\
-        Out(:, pos, i) = \\alpha * input(:, pos, i) + \\beta * PE(pos, i)
+        PE(pos, 2i) &= \\sin{(pos / 10000^{2i / P})}   \\\\
+        PE(pos, 2i + 1) &= \\cos{(pos / 10000^{2i / P})}  \\\\
+        Out(:, pos, i) &= \\alpha * input(:, pos, i) + \\beta * PE(pos, i)
 
     Where:
-    * PE(pos, 2i): the increment for the number at even position
-    * PE(pos, 2i + 1): the increment for the number at odd position
+      - :math:`PE(pos, 2i)` : the increment for the number at even position
+      - :math:`PE(pos, 2i + 1)` : the increment for the number at odd position
 
     Args:
         input (Variable): 3-D input tensor with shape [N x M x P]
@@ -9230,6 +9284,7 @@ def add_position_encoding(input, alpha, beta, name=None):
         .. code-block:: python
 
           position_tensor = fluid.layers.add_position_encoding(input=tensor)
+
     """
     helper = LayerHelper('add_position_encoding', **locals())
     dtype = helper.input_dtype()
@@ -9262,13 +9317,13 @@ def bilinear_tensor_product(x,
     For example:
 
     .. math::
-       out{i} = x * W_{i} * {y^\mathrm{T}}, i=0,1,...,size-1
+       out_{i} = x * W_{i} * {y^\mathrm{T}}, i=0,1,...,size-1
 
     In this formula:
       - :math:`x`: the first input contains M elements, shape is [batch_size, M].
       - :math:`y`: the second input contains N elements, shape is [batch_size, N].
       - :math:`W_{i}`: the i-th learned weight, shape is [M, N]
-      - :math:`out{i}`: the i-th element of out, shape is [batch_size, size].
+      - :math:`out_{i}`: the i-th element of out, shape is [batch_size, size].
       - :math:`y^\mathrm{T}`: the transpose of :math:`y_{2}`.
 
     Args:

python/paddle/fluid/layers/tensor.py

@@ -393,9 +393,6 @@ def fill_constant_batch_size_like(input,
 
     It also sets *stop_gradient* to True.
 
-    >>> data = fluid.layers.fill_constant_batch_size_like(
-    >>>             input=like, shape=[1], value=0, dtype='int64')
-
     Args:
         input(${input_type}): ${input_comment}.
 
@@ -411,6 +408,14 @@ def fill_constant_batch_size_like(input,
 
     Returns:
         ${out_comment}.
+
+    Examples:
+
+        .. code-block:: python
+
+             data = fluid.layers.fill_constant_batch_size_like(
+                         input=like, shape=[1], value=0, dtype='int64')
+
     """
     helper = LayerHelper("fill_constant_batch_size_like", **locals())
     out = helper.create_variable_for_type_inference(dtype=dtype)

python/paddle/fluid/metrics.py

@@ -362,7 +362,7 @@ class ChunkEvaluator(MetricBase):
     compute the precision recall and F1-score using the accumulated counter
     numbers.
     For some basics of chunking, please refer to 
-    'Chunking with Support Vector Machines <https://aclanthology.info/pdf/N/N01/N01-1025.pdf>'.
+    `Chunking with Support Vector Machines <https://aclanthology.info/pdf/N/N01/N01-1025.pdf>`_ .
     ChunkEvalEvaluator computes the precision, recall, and F1-score of chunk detection,
     and supports IOB, IOE, IOBES and IO (also known as plain) tagging schemes.
 
@@ -391,6 +391,7 @@ class ChunkEvaluator(MetricBase):
     def update(self, num_infer_chunks, num_label_chunks, num_correct_chunks):
         """
         Update the states based on the layers.chunk_eval() ouputs.
+
         Args:
             num_infer_chunks(int|numpy.array): The number of chunks in Inference on the given minibatch.
             num_label_chunks(int|numpy.array): The number of chunks in Label on the given mini-batch.
@@ -450,9 +451,9 @@ class EditDistance(MetricBase):
                 distance, instance_error = distance_evaluator.eval()
 
     In the above example:
-        'distance' is the average of the edit distance in a pass.
 
-        'instance_error' is the instance error rate in a pass.
+        - 'distance' is the average of the edit distance in a pass.
+        - 'instance_error' is the instance error rate in a pass.
 
     """
 
@@ -567,12 +568,15 @@ class DetectionMAP(object):
     Calculate the detection mean average precision (mAP).
 
     The general steps are as follows:
+
     1. calculate the true positive and false positive according to the input
        of detection and labels.
     2. calculate mAP value, support two versions: '11 point' and 'integral'.
 
     Please get more information from the following articles:
+
       https://sanchom.wordpress.com/tag/average-precision/
+
       https://arxiv.org/abs/1512.02325
 
     Args:
@@ -615,8 +619,10 @@ class DetectionMAP(object):
 
     In the above example:
 
-            'cur_map_v' is the mAP of current mini-batch.
-            'accum_map_v' is the accumulative mAP of one pass.
+            - 'cur_map_v' is the mAP of current mini-batch.
+            - 'accum_map_v' is the accumulative mAP of one pass.
+
+ 
     """
 
     def __init__(self,

python/paddle/fluid/tests/unittests/test_conv2d_fusion_op.py

@@ -32,6 +32,8 @@ class TestConv2dFusionOp(OpTest):
         self.activation = 'relu'
         self.add_bias = True
         self.add_residual_data = True
+        self.channels = None
+        self.outputs = None
 
         self.init_group()
         self.init_dilation()
@@ -49,7 +51,7 @@ class TestConv2dFusionOp(OpTest):
         input = np.random.random(self.input_size).astype(self.dtype)
         filter = np.random.random(self.filter_size).astype(self.dtype)
 
-        output = conv2d_forward_naive(input, filter, self.groups,
+        self.output = conv2d_forward_naive(input, filter, self.groups,
                                            conv2d_param).astype(self.dtype)
 
         self.inputs = {
@@ -58,19 +60,20 @@ class TestConv2dFusionOp(OpTest):
         }
 
         if self.add_residual_data:
-            residual_data = np.random.random(output.shape).astype(self.dtype)
+            residual_data = np.random.random(self.output.shape).astype(
+                self.dtype)
             self.inputs['ResidualData'] = OpTest.np_dtype_to_fluid_dtype(
                 residual_data)
-            output += residual_data
+            self.output += residual_data
 
         if self.add_bias:
             bias = np.random.random(self.filter_size[0]).astype(self.dtype)
             self.inputs['Bias'] = OpTest.np_dtype_to_fluid_dtype(bias)
-            output = output + bias.reshape((1, bias.size, 1, 1))
+            self.output = self.output + bias.reshape((1, bias.size, 1, 1))
 
         assert self.activation in ['relu', 'identity']
         if self.activation == 'relu':
-            output = np.maximum(output, 0)
+            self.output = np.maximum(self.output, 0)
 
         self.attrs = {
             'strides': self.stride,
@@ -79,9 +82,12 @@ class TestConv2dFusionOp(OpTest):
             'dilations': self.dilations,
             'data_format': self.data_format,
             'exhaustive_search': self.exhaustive_search,
-            'activation': self.activation
+            'activation': self.activation,
+            'split_channels': self.channels
         }
-        self.outputs = {'Output': output}
+        self.outputs = {'Output': self.output}
+
+        self.set_outputs()
 
     def testcuda(self):
         return core.is_compiled_with_cuda()
@@ -117,6 +123,9 @@ class TestConv2dFusionOp(OpTest):
     def set_search_method(self):
         self.exhaustive_search = False
 
+    def set_outputs(self):
+        pass
+
 
 class TestWithoutResidual(TestConv2dFusionOp):
     def init_bias_residual(self):
@@ -160,5 +169,21 @@ class TestCUDNNExhaustiveSearch(TestConv2dFusionOp):
         self.exhaustive_search = True
 
 
+class TestMultipleOutputs(TestConv2dFusionOp):
+    def init_test_case(self):
+        self.pad = [1, 1]
+        self.stride = [1, 1]
+        self.input_size = [1, 32, 17, 17]  # NCHW
+        assert np.mod(self.input_size[1], self.groups) == 0
+        f_c = self.input_size[1] // self.groups
+        self.filter_size = [126, f_c, 3, 3]
+        self.channels = [84, 42]
+
+    def set_outputs(self):
+        out1 = self.output[:, 0:84, :, :]
+        out2 = self.output[:, 84:126, :, :]
+        self.outputs['Outputs'] = [('out1', out1), ('out2', out2)]
+
+
 if __name__ == '__main__':
     unittest.main()

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

@@ -243,6 +243,10 @@ class TestBook(unittest.TestCase):
             pool, mask = layers.adaptive_pool2d(x, [3, 3], require_index=True)
             self.assertIsNotNone(pool)
             self.assertIsNotNone(mask)
+            self.assertIsNotNone(layers.adaptive_pool2d(x, 3, pool_type='avg'))
+            pool, mask = layers.adaptive_pool2d(x, 3, require_index=True)
+            self.assertIsNotNone(pool)
+            self.assertIsNotNone(mask)
 
     def test_adaptive_pool3d(self):
         program = Program()
@@ -255,6 +259,10 @@ class TestBook(unittest.TestCase):
                 x, [3, 3, 3], require_index=True)
             self.assertIsNotNone(pool)
             self.assertIsNotNone(mask)
+            self.assertIsNotNone(layers.adaptive_pool3d(x, 3, pool_type='avg'))
+            pool, mask = layers.adaptive_pool3d(x, 3, require_index=True)
+            self.assertIsNotNone(pool)
+            self.assertIsNotNone(mask)
 
     def test_lstm_unit(self):
         program = Program()

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

@@ -137,9 +137,9 @@ def append_input_output(block, op_proto, np_list, is_input, dtype):
     var_dict = {}
     for var_proto in proto_list:
         var_name = str(var_proto.name)
-        if is_input:
         if (var_name not in np_list) and var_proto.dispensable:
             continue
+        if is_input:
             assert (var_name in np_list) or (var_proto.dispensable), \
                 "Missing {} as input".format(var_name)
         if var_proto.duplicable:

python/paddle/fluid/transpiler/distribute_transpiler.py

@@ -125,14 +125,23 @@ def slice_variable(var_list, slice_count, min_block_size):
 
 class DistributeTranspilerConfig(object):
     """
-    Args:
-        slice_var_up (bool): Do Tensor slice for pservers, default is True.
-        split_method (PSDispatcher): RoundRobin or HashName can be used
-          try to choose the best method to balance loads for pservers.
-        min_block_size (int): Minimum splitted element number in block.
-          According:https://github.com/PaddlePaddle/Paddle/issues/8638#issuecomment-369912156
+    .. py:attribute:: slice_var_up (bool)
+
+          Do Tensor slice for pservers, default is True.
+
+    .. py:attribute:: split_method (PSDispatcher)
+
+          RoundRobin or HashName can be used.
+          Try to choose the best method to balance loads for pservers.
+
+    .. py:attribute:: min_block_size (int)
+
+          Minimum number of splitted elements in block.
+
+          According to : https://github.com/PaddlePaddle/Paddle/issues/8638#issuecomment-369912156
           We can use bandwidth effiently when data size is larger than 2MB.If you
-          want to change it, please be sure you see the slice_variable function.
+          want to change it, please be sure you have read the slice_variable function.
+
     """
 
     slice_var_up = True