Merge develop

test=develop

paddle/fluid/API.spec

@@ -15,7 +15,7 @@ paddle.fluid.cpu_places (ArgSpec(args=['device_count'], varargs=None, keywords=N
 paddle.fluid.cuda_pinned_places (ArgSpec(args=['device_count'], varargs=None, keywords=None, defaults=(None,)), ('document', 'd0c3ebd813c39958c92b78e3eef7e912'))
 paddle.fluid.Executor.__init__ (ArgSpec(args=['self', 'place'], varargs=None, keywords=None, defaults=None), ('document', '6adf97f83acf6453d4a6a4b1070f3754'))
 paddle.fluid.Executor.close (ArgSpec(args=['self'], varargs=None, keywords=None, defaults=None), ('document', 'f5369953dd0c443961cf79f7a00e1a03'))
-paddle.fluid.Executor.run (ArgSpec(args=['self', 'program', 'feed', 'fetch_list', 'feed_var_name', 'fetch_var_name', 'scope', 'return_numpy', 'use_program_cache'], varargs=None, keywords=None, defaults=(None, None, None, 'feed', 'fetch', None, True, False)), ('document', 'aba8093edebf2d5c869b735b92811e45'))
+paddle.fluid.Executor.run (ArgSpec(args=['self', 'program', 'feed', 'fetch_list', 'feed_var_name', 'fetch_var_name', 'scope', 'return_numpy', 'use_program_cache'], varargs=None, keywords=None, defaults=(None, None, None, 'feed', 'fetch', None, True, False)), ('document', 'f482e93b38b4018796969a2e1dde479d'))
 paddle.fluid.global_scope (ArgSpec(args=[], varargs=None, keywords=None, defaults=None), ('document', 'e148d3ab1ed8edf3e928212a375959c0'))
 paddle.fluid.scope_guard (ArgSpec(args=['scope'], varargs=None, keywords=None, defaults=None), ('document', 'b94d1f6bcc29c4fb58fc0058561250c2'))
 paddle.fluid.DistributeTranspiler.__init__ (ArgSpec(args=['self', 'config'], varargs=None, keywords=None, defaults=(None,)), ('document', '6adf97f83acf6453d4a6a4b1070f3754'))
@@ -47,7 +47,7 @@ paddle.fluid.AsyncExecutor.run (ArgSpec(args=['self', 'program', 'data_feed', 'f
 paddle.fluid.AsyncExecutor.save_model (ArgSpec(args=['self', 'save_path'], varargs=None, keywords=None, defaults=None), ('document', 'c8ac0dfcb3b187aba25d03af7fea56b2'))
 paddle.fluid.AsyncExecutor.stop (ArgSpec(args=['self'], varargs=None, keywords=None, defaults=None), ('document', '5f23d043607bb5d55e466ec3f578e093'))
 paddle.fluid.CompiledProgram.__init__ (ArgSpec(args=['self', 'program_or_graph'], varargs=None, keywords=None, defaults=None), ('document', '6adf97f83acf6453d4a6a4b1070f3754'))
-paddle.fluid.CompiledProgram.with_data_parallel (ArgSpec(args=['self', 'loss_name', 'build_strategy', 'exec_strategy', 'share_vars_from', 'places'], varargs=None, keywords=None, defaults=(None, None, None, None, None)), ('document', '5e8cca4619a5d7c3280fb3cae7021b14'))
+paddle.fluid.CompiledProgram.with_data_parallel (ArgSpec(args=['self', 'loss_name', 'build_strategy', 'exec_strategy', 'share_vars_from', 'places'], varargs=None, keywords=None, defaults=(None, None, None, None, None)), ('document', 'a8c7793803cf976680d9478e378fa356'))
 paddle.fluid.CompiledProgram.with_inference_optimize (ArgSpec(args=['self', 'config'], varargs=None, keywords=None, defaults=None), ('document', '9e5b009d850191a010e859189c127fd8'))
 paddle.fluid.ExecutionStrategy.__init__ __init__(self: paddle.fluid.core.ParallelExecutor.ExecutionStrategy) -> None
 paddle.fluid.BuildStrategy.GradientScaleStrategy.__init__ __init__(self: paddle.fluid.core.ParallelExecutor.BuildStrategy.GradientScaleStrategy, arg0: int) -> None
@@ -286,7 +286,7 @@ paddle.fluid.layers.DynamicRNN.block (ArgSpec(args=['self'], varargs=None, keywo
 paddle.fluid.layers.DynamicRNN.memory (ArgSpec(args=['self', 'init', 'shape', 'value', 'need_reorder', 'dtype'], varargs=None, keywords=None, defaults=(None, None, 0.0, False, 'float32')), ('document', 'b9174d4e91505b0c8ecc193eb51e248d'))
 paddle.fluid.layers.DynamicRNN.output (ArgSpec(args=['self'], varargs='outputs', keywords=None, defaults=None), ('document', 'b439a176a3328de8a75bdc5c08eece4a'))
 paddle.fluid.layers.DynamicRNN.static_input (ArgSpec(args=['self', 'x'], varargs=None, keywords=None, defaults=None), ('document', 'f29ad2478b6b2ad4f413d2936a331ea0'))
-paddle.fluid.layers.DynamicRNN.step_input (ArgSpec(args=['self', 'x'], varargs=None, keywords=None, defaults=None), ('document', '169d694d2224f62b4f3afdc3dbc19e95'))
+paddle.fluid.layers.DynamicRNN.step_input (ArgSpec(args=['self', 'x', 'level'], varargs=None, keywords=None, defaults=(0,)), ('document', '7568c5ac7622a10288d3307a94134655'))
 paddle.fluid.layers.DynamicRNN.update_memory (ArgSpec(args=['self', 'ex_mem', 'new_mem'], varargs=None, keywords=None, defaults=None), ('document', '5d83987da13b98363d6a807a52d8024f'))
 paddle.fluid.layers.StaticRNN.__init__ (ArgSpec(args=['self', 'name'], varargs=None, keywords=None, defaults=(None,)), ('document', '6adf97f83acf6453d4a6a4b1070f3754'))
 paddle.fluid.layers.StaticRNN.memory (ArgSpec(args=['self', 'init', 'shape', 'batch_ref', 'init_value', 'init_batch_dim_idx', 'ref_batch_dim_idx'], varargs=None, keywords=None, defaults=(None, None, None, 0.0, 0, 1)), ('document', 'c24e368e23afac1ed91a78a639d7a9c7'))
@@ -305,7 +305,7 @@ paddle.fluid.layers.tanh (ArgSpec(args=['x', 'name'], varargs=None, keywords=Non
 paddle.fluid.layers.atan (ArgSpec(args=['x', 'name'], varargs=None, keywords=None, defaults=(None,)), ('document', '3a46e0b5f9ce82348406478e610f14c9'))
 paddle.fluid.layers.tanh_shrink (ArgSpec(args=['x', 'name'], varargs=None, keywords=None, defaults=(None,)), ('document', '1e521554b9fdda9061ec6d306f0709b7'))
 paddle.fluid.layers.softshrink (ArgSpec(args=['x', 'name'], varargs=None, keywords=None, defaults=(None,)), ('document', '9eef31597bbafa2bd49691e072296e13'))
-paddle.fluid.layers.sqrt (ArgSpec(args=['x', 'name'], varargs=None, keywords=None, defaults=(None,)), ('document', '072a8541e0f632366bba10f67cb0db27'))
+paddle.fluid.layers.sqrt (ArgSpec(args=['x', 'name'], varargs=None, keywords=None, defaults=(None,)), ('document', 'e9e27491c39ac74d0b1ffe506aec0ebb'))
 paddle.fluid.layers.abs (ArgSpec(args=['x', 'name'], varargs=None, keywords=None, defaults=(None,)), ('document', '64650ac42cf82e9920cb0b172b1d29fd'))
 paddle.fluid.layers.ceil (ArgSpec(args=['x', 'name'], varargs=None, keywords=None, defaults=(None,)), ('document', 'c75d67dc5fe28f68e4cfffead4f698ad'))
 paddle.fluid.layers.floor (ArgSpec(args=['x', 'name'], varargs=None, keywords=None, defaults=(None,)), ('document', '647b16c5da5ef909649ae02abb434973'))
@@ -503,7 +503,7 @@ paddle.fluid.CUDAPinnedPlace.__init__ __init__(self: paddle.fluid.core.CUDAPinne
 paddle.fluid.ParamAttr.__init__ (ArgSpec(args=['self', 'name', 'initializer', 'learning_rate', 'regularizer', 'trainable', 'gradient_clip', 'do_model_average'], varargs=None, keywords=None, defaults=(None, None, 1.0, None, True, None, False)), ('document', '6adf97f83acf6453d4a6a4b1070f3754'))
 paddle.fluid.WeightNormParamAttr.__init__ (ArgSpec(args=['self', 'dim', 'name', 'initializer', 'learning_rate', 'regularizer', 'trainable', 'gradient_clip', 'do_model_average'], varargs=None, keywords=None, defaults=(None, None, None, 1.0, None, True, None, False)), ('document', '6adf97f83acf6453d4a6a4b1070f3754'))
 paddle.fluid.DataFeeder.__init__ (ArgSpec(args=['self', 'feed_list', 'place', 'program'], varargs=None, keywords=None, defaults=(None,)), ('document', '6adf97f83acf6453d4a6a4b1070f3754'))
-paddle.fluid.DataFeeder.decorate_reader (ArgSpec(args=['self', 'reader', 'multi_devices', 'num_places', 'drop_last'], varargs=None, keywords=None, defaults=(None, True)), ('document', '0eed2f198dc73c08a41b61edbc755753'))
+paddle.fluid.DataFeeder.decorate_reader (ArgSpec(args=['self', 'reader', 'multi_devices', 'num_places', 'drop_last'], varargs=None, keywords=None, defaults=(None, True)), ('document', 'f8f3df23c5633c614db781a91b81fb62'))
 paddle.fluid.DataFeeder.feed (ArgSpec(args=['self', 'iterable'], varargs=None, keywords=None, defaults=None), ('document', '459e316301279dfd82001b46f0b8ffca'))
 paddle.fluid.DataFeeder.feed_parallel (ArgSpec(args=['self', 'iterable', 'num_places'], varargs=None, keywords=None, defaults=(None,)), ('document', '543863d1f9d4853758adb613b8659e85'))
 paddle.fluid.clip.ErrorClipByValue.__init__ (ArgSpec(args=['self', 'max', 'min'], varargs=None, keywords=None, defaults=(None,)), ('document', '6adf97f83acf6453d4a6a4b1070f3754'))
@@ -528,11 +528,11 @@ paddle.reader.compose (ArgSpec(args=[], varargs='readers', keywords='kwargs', de
 paddle.reader.chain (ArgSpec(args=[], varargs='readers', keywords=None, defaults=None), ('document', 'd22c34e379a53901ae67a6bca7f4def4'))
 paddle.reader.shuffle (ArgSpec(args=['reader', 'buf_size'], varargs=None, keywords=None, defaults=None), ('document', 'e42ea6fee23ce26b23cb142cd1d6522d'))
 paddle.reader.firstn (ArgSpec(args=['reader', 'n'], varargs=None, keywords=None, defaults=None), ('document', 'c5bb8f7dd4f917f1569a368aab5b8aad'))
-paddle.reader.xmap_readers (ArgSpec(args=['mapper', 'reader', 'process_num', 'buffer_size', 'order'], varargs=None, keywords=None, defaults=(False,)), ('document', '283bc0b8a0e26ae186b8b9bee4aec560'))
+paddle.reader.xmap_readers (ArgSpec(args=['mapper', 'reader', 'process_num', 'buffer_size', 'order'], varargs=None, keywords=None, defaults=(False,)), ('document', '9c804a42f8a4dbaa76b3c98e0ab7f796'))
 paddle.reader.PipeReader.__init__ (ArgSpec(args=['self', 'command', 'bufsize', 'file_type'], varargs=None, keywords=None, defaults=(8192, 'plain')), ('document', '6adf97f83acf6453d4a6a4b1070f3754'))
-paddle.reader.PipeReader.get_line (ArgSpec(args=['self', 'cut_lines', 'line_break'], varargs=None, keywords=None, defaults=(True, '\n')), ('document', '5f80a7ed70052f01665e4c74acccfa69'))
+paddle.reader.PipeReader.get_line (ArgSpec(args=['self', 'cut_lines', 'line_break'], varargs=None, keywords=None, defaults=(True, '\n')), ('document', '9621ae612e595b6c34eb3bb5f3eb1a45'))
 paddle.reader.multiprocess_reader (ArgSpec(args=['readers', 'use_pipe', 'queue_size'], varargs=None, keywords=None, defaults=(True, 1000)), ('document', '7d8b3a96e592107c893d5d51ce968ba0'))
 paddle.reader.Fake.__init__ (ArgSpec(args=['self'], varargs=None, keywords=None, defaults=None), ('document', '6adf97f83acf6453d4a6a4b1070f3754'))
 paddle.reader.creator.np_array (ArgSpec(args=['x'], varargs=None, keywords=None, defaults=None), ('document', '28d457fbc9a71efa4ac91a3be179cada'))
-paddle.reader.creator.text_file (ArgSpec(args=['path'], varargs=None, keywords=None, defaults=None), ('document', '44fe286ab6175a5464d3a961a68c266a'))
-paddle.reader.creator.recordio (ArgSpec(args=['paths', 'buf_size'], varargs=None, keywords=None, defaults=(100,)), ('document', '11b3704ea42cfd537953387a7e58dae8'))
+paddle.reader.creator.text_file (ArgSpec(args=['path'], varargs=None, keywords=None, defaults=None), ('document', 'f45fcb7add066c8e042c6774fc7c3db2'))
+paddle.reader.creator.recordio (ArgSpec(args=['paths', 'buf_size'], varargs=None, keywords=None, defaults=(100,)), ('document', 'b4a94ee0e2cefb495619275c2f8c61d2'))

paddle/fluid/framework/details/CMakeLists.txt

@@ -9,6 +9,7 @@ cc_library(rpc_op_handle SRCS rpc_op_handle.cc DEPS framework_proto scope place
 cc_library(multi_devices_helper SRCS multi_devices_helper.cc DEPS graph graph_helper)
 cc_library(multi_devices_graph_print_pass SRCS multi_devices_graph_print_pass.cc DEPS multi_devices_helper)
 cc_library(multi_devices_graph_check_pass SRCS multi_devices_graph_check_pass.cc DEPS multi_devices_helper)
+cc_library(alloc_continuous_space_for_grad_pass SRCS alloc_continuous_space_for_grad_pass.cc DEPS graph graph_helper)
 
 cc_library(variable_visitor SRCS variable_visitor.cc DEPS lod_tensor selected_rows)
 
@@ -22,6 +23,8 @@ endif()
 if(WITH_GPU)
     nv_library(all_reduce_op_handle SRCS all_reduce_op_handle.cc DEPS op_handle_base scope lod_tensor ddim memory
             dynload_cuda variable_visitor)
+    nv_library(fused_all_reduce_op_handle SRCS fused_all_reduce_op_handle.cc DEPS op_handle_base scope lod_tensor ddim memory
+            dynload_cuda variable_visitor)
     if(WITH_DISTRIBUTE)
         nv_library(reduce_op_handle SRCS reduce_op_handle.cc DEPS op_handle_base variable_visitor scope
             ddim dynload_cuda selected_rows_functor sendrecvop_rpc)
@@ -35,6 +38,8 @@ if(WITH_GPU)
 else()
     cc_library(all_reduce_op_handle SRCS all_reduce_op_handle.cc DEPS op_handle_base scope lod_tensor ddim memory
              variable_visitor)
+    cc_library(fused_all_reduce_op_handle SRCS fused_all_reduce_op_handle.cc DEPS op_handle_base scope lod_tensor ddim memory
+            variable_visitor)
     if(WITH_DISTRIBUTE)
         cc_library(reduce_op_handle SRCS reduce_op_handle.cc DEPS op_handle_base variable_visitor scope
             ddim selected_rows_functor sendrecvop_rpc)
@@ -71,6 +76,8 @@ cc_library(all_reduce_deps_pass SRCS all_reduce_deps_pass.cc DEPS graph graph_he
 cc_library(multi_devices_graph_pass SRCS multi_devices_graph_pass.cc DEPS multi_devices_helper computation_op_handle
         scale_loss_grad_op_handle rpc_op_handle all_reduce_op_handle reduce_op_handle broadcast_op_handle data_balance_op_handle fused_broadcast_op_handle)
 
+cc_library(fuse_all_reduce_op_pass SRCS fuse_all_reduce_op_pass.cc DEPS graph graph_helper fused_all_reduce_op_handle)
+
 set(SSA_GRAPH_EXECUTOR_DEPS graph framework_proto sequential_execution_pass modify_op_lock_and_record_event_pass all_reduce_deps_pass reference_count_pass eager_deletion_pass memory_optimize_pass inplace_op_pass)
 if (WITH_GPU)
   list(APPEND SSA_GRAPH_EXECUTOR_DEPS reference_count_pass)
@@ -98,5 +105,5 @@ cc_library(build_strategy SRCS build_strategy.cc DEPS
         graph_viz_pass multi_devices_graph_pass
         multi_devices_graph_print_pass multi_devices_graph_check_pass
         fuse_elewise_add_act_pass multi_batch_merge_pass 
-        fuse_relu_depthwise_conv_pass
-        memory_optimize_pass lock_free_optimize_pass)
+        fuse_relu_depthwise_conv_pass 
+        memory_optimize_pass lock_free_optimize_pass alloc_continuous_space_for_grad_pass fuse_all_reduce_op_pass)

paddle/fluid/framework/details/alloc_continuous_space_for_grad_pass.cc

+//   Copyright (c) 2019 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 <algorithm>
+#include <string>
+#include <utility>
+#include <vector>
+
+#include "paddle/fluid/framework/details/build_strategy.h"
+#include "paddle/fluid/framework/details/multi_devices_helper.h"
+#include "paddle/fluid/framework/ir/graph_helper.h"
+#include "paddle/fluid/framework/op_registry.h"
+DEFINE_uint32(fuse_parameter_memory_size, 0,  // 0 KB
+              "fuse_parameter_memory_size is up limited memory size "
+              "of one group parameters' gradient which is the input "
+              "of communication calling(e.g NCCLAllReduce). "
+              "The default value is 0, it means that "
+              "not set group according to memory_size.");
+DEFINE_int32(
+    fuse_parameter_groups_size, 3,
+    "fuse_parameter_groups_size is the size of one group parameters' gradient. "
+    "The default value is a experimental result. If the "
+    "fuse_parameter_groups_size is 1, it means that the groups size is "
+    "the number of parameters' gradient. If the fuse_parameter_groups_size is "
+    "-1, it means that there are only one group. The default value is 3, it is "
+    "an experimental value.");
+
+namespace paddle {
+namespace framework {
+namespace details {
+
+static const char kUnKnow[] = "@UNKNOW@";
+static framework::proto::VarType::Type kDefaultDtype =
+    framework::proto::VarType::Type::VarType_Type_BOOL;
+
+class AllocContinuousSpaceForGradPass : public ir::Pass {
+ protected:
+  std::unique_ptr<ir::Graph> ApplyImpl(
+      std::unique_ptr<ir::Graph> graph) const override {
+    ir::Graph &result = *graph;
+
+    auto &places = Get<const std::vector<platform::Place>>(kPlaces);
+    auto &local_scopes = Get<const std::vector<Scope *>>(kLocalScopes);
+
+    ResetAttribute<ParamsAndGrads>(kParamsAndGrads, &result);
+    ResetAttribute<GroupGradsAndParams>(kGroupGradsAndParams, &result);
+
+    // NOTE: The operator nodes should be in topology order.
+    std::vector<ir::Node *> topo_nodes = ir::TopologySortOperations(result);
+    auto &params_grads = result.Get<ParamsAndGrads>(kParamsAndGrads);
+    for (auto &node : topo_nodes) {
+      RecordParamsAndGrads(node, &params_grads);
+    }
+
+    if (params_grads.size() == 0) {
+      VLOG(10) << "Doesn't find gradients";
+      return std::move(graph);
+    }
+
+    std::unordered_map<std::string, ir::Node *> vars;
+    for (ir::Node *node : result.Nodes()) {
+      if (node->IsVar() && node->Var()) {
+        // Note: The graph may have the same name node. For example, parameter
+        // is the input of operator and it also is the output of optimizer;
+        vars.emplace(node->Var()->Name(), node);
+      }
+    }
+
+    auto &group_grads_params =
+        result.Get<GroupGradsAndParams>(kGroupGradsAndParams);
+
+    // Note: the order of params_grads may be changed by SetGroupGradsAndParams.
+    SetGroupGradsAndParams(vars, params_grads, &group_grads_params);
+
+    params_grads.clear();
+    for (auto &group_p_g : group_grads_params) {
+      params_grads.insert(params_grads.begin(), group_p_g.begin(),
+                          group_p_g.end());
+    }
+    for (auto &p_g : params_grads) {
+      std::swap(p_g.first, p_g.second);
+    }
+
+    // Set Gradients as Persistable to prevent this var becoming reusable.
+    auto dtype = kDefaultDtype;
+    for (auto &p_g : params_grads) {
+      // Get gradient var
+      auto iter = vars.find(p_g.second);
+      PADDLE_ENFORCE(iter != vars.end(), "%s is not found.", p_g.second);
+      iter->second->Var()->SetPersistable(true);
+
+      PADDLE_ENFORCE(IsSupportedVarType(iter->second->Var()->GetType()));
+
+      // Get Dtype
+      auto ele_dtype = iter->second->Var()->GetDataType();
+      if (dtype == kDefaultDtype) {
+        dtype = ele_dtype;
+        PADDLE_ENFORCE_NE(ele_dtype, kDefaultDtype);
+      }
+      PADDLE_ENFORCE_EQ(ele_dtype, dtype);
+    }
+
+    // Create the fused variable name.
+    if (!result.Has(kFusedVars)) {
+      result.Set(kFusedVars, new FusedVars);
+    }
+    const std::string prefix(kFusedVarNamePrefix);
+    // The fused_var_name should be unique.
+    auto fused_var_name = prefix + "GRAD@" + params_grads[0].second;
+    auto &fused_var_set = result.Get<FusedVars>(kFusedVars);
+    PADDLE_ENFORCE_EQ(fused_var_set.count(fused_var_name), 0);
+    fused_var_set.insert(fused_var_name);
+
+    InitFusedVarsAndAllocSpaceForVars(places, local_scopes, vars,
+                                      fused_var_name, params_grads);
+
+    return std::move(graph);
+  }
+
+  template <typename AttrType>
+  void ResetAttribute(const std::string &attr_name, ir::Graph *graph) const {
+    if (graph->Has(attr_name)) {
+      VLOG(10) << attr_name << " is reset.";
+      graph->Erase(attr_name);
+    }
+    graph->Set(attr_name, new AttrType);
+  }
+
+  void SetGroupGradsAndParams(
+      const std::unordered_map<std::string, ir::Node *> &var_nodes,
+      const ParamsAndGrads &params_grads,
+      GroupGradsAndParams *group_grads_params) const {
+    SetGroupAccordingToLayers(var_nodes, params_grads, group_grads_params);
+    SetGroupAccordingToMemorySize(var_nodes, group_grads_params);
+    SetGroupAccordingToGroupSize(var_nodes, group_grads_params);
+  }
+
+  void SetGroupAccordingToLayers(
+      const std::unordered_map<std::string, ir::Node *> &var_nodes,
+      const ParamsAndGrads &params_grads,
+      GroupGradsAndParams *group_grads_params) const {
+    std::unordered_map<std::string, std::vector<int>> layer_params;
+
+    for (size_t i = 0; i < params_grads.size(); ++i) {
+      auto pos = params_grads[i].first.find_first_of(".");
+      if (pos == std::string::npos) {
+        layer_params[std::string(kUnKnow)].emplace_back(i);
+      } else {
+        layer_params[params_grads[i].first.substr(0, pos)].emplace_back(i);
+      }
+    }
+
+    group_grads_params->reserve(layer_params.size());
+    for (size_t i = 0; i < params_grads.size(); ++i) {
+      auto pos = params_grads[i].first.find_first_of(".");
+      std::string key = kUnKnow;
+      if (pos != std::string::npos) {
+        key = params_grads[i].first.substr(0, pos);
+      }
+      auto iter = layer_params.find(key);
+      if (iter == layer_params.end()) continue;
+
+      group_grads_params->emplace_back();
+      auto &local_group_grads_params = group_grads_params->back();
+      for (auto &idx : iter->second) {
+        local_group_grads_params.emplace_back(
+            std::make_pair(params_grads[idx].second, params_grads[idx].first));
+      }
+      layer_params.erase(iter);
+    }
+
+    VLOG(10) << "SetGroupAccordingToLayers: ";
+    for (size_t i = 0; i < group_grads_params->size(); ++i) {
+      VLOG(10) << "group " << i;
+      std::stringstream out;
+      for (auto &p_g : group_grads_params->at(i)) {
+        out << "(" << p_g.second << ", " << p_g.first << "), ";
+      }
+      VLOG(10) << out.str();
+    }
+  }
+
+  void SetGroupAccordingToMemorySize(
+      const std::unordered_map<std::string, ir::Node *> &var_nodes,
+      GroupGradsAndParams *group_grads_params) const {
+    if (FLAGS_fuse_parameter_memory_size == 0) {
+      return;
+    }
+    size_t group_memory_size =
+        static_cast<size_t>(FLAGS_fuse_parameter_memory_size);
+    GroupGradsAndParams local_group_grads_params;
+
+    size_t j = 0;
+    while (j < group_grads_params->size()) {
+      local_group_grads_params.emplace_back();
+      auto &group_p_g = local_group_grads_params.back();
+      size_t local_group_memory_size = 0;
+      while (j < group_grads_params->size()) {
+        std::for_each(
+            group_grads_params->at(j).begin(), group_grads_params->at(j).end(),
+            [&local_group_memory_size,
+             &var_nodes](const std::pair<std::string, std::string> &g_p) {
+              auto iter = var_nodes.find(g_p.second);
+              PADDLE_ENFORCE(iter != var_nodes.end(), "%s is not found.",
+                             g_p.second);
+              auto shape = iter->second->Var()->GetShape();
+              size_t size =
+                  framework::SizeOfType(iter->second->Var()->GetDataType());
+              std::for_each(shape.begin(), shape.end(),
+                            [&size](const int64_t &n) { size *= n; });
+              local_group_memory_size += size;
+            });
+        group_p_g.insert(group_p_g.end(), group_grads_params->at(j).begin(),
+                         group_grads_params->at(j).end());
+        ++j;
+        if (local_group_memory_size >= group_memory_size) {
+          break;
+        }
+      }
+    }
+
+    std::swap(*group_grads_params, local_group_grads_params);
+
+    VLOG(10) << string::Sprintf(
+        "SetGroupAccordingToMemorySize(memory_size: %d):",
+        FLAGS_fuse_parameter_memory_size);
+    for (size_t i = 0; i < group_grads_params->size(); ++i) {
+      VLOG(10) << "group " << i;
+      std::stringstream out;
+      for (auto &g_p : group_grads_params->at(i)) {
+        auto iter = var_nodes.find(g_p.second);
+        PADDLE_ENFORCE(iter != var_nodes.end(), "%s is not found.", g_p.second);
+        auto shape = iter->second->Var()->GetShape();
+        size_t size = framework::SizeOfType(iter->second->Var()->GetDataType());
+        std::for_each(shape.begin(), shape.end(),
+                      [&size](const int64_t &n) { size *= n; });
+        out << string::Sprintf("(%s(%d), %s)", g_p.second, size, g_p.first);
+      }
+      VLOG(10) << out.str();
+    }
+  }
+
+  void SetGroupAccordingToGroupSize(
+      const std::unordered_map<std::string, ir::Node *> &var_nodes,
+      GroupGradsAndParams *group_grads_params) const {
+    if (FLAGS_fuse_parameter_groups_size == 1) {
+      return;
+    }
+    size_t group_size = static_cast<size_t>(FLAGS_fuse_parameter_groups_size);
+    if (FLAGS_fuse_parameter_groups_size == -1) {
+      group_size = group_grads_params->size();
+    }
+    PADDLE_ENFORCE_GT(group_size, 1);
+    size_t groups = (group_grads_params->size() + group_size - 1) / group_size;
+    GroupGradsAndParams local_group_grads_params;
+    local_group_grads_params.reserve(groups);
+
+    size_t j = 0;
+    for (size_t i = 0; i < groups; ++i) {
+      local_group_grads_params.emplace_back();
+      auto &group_p_g = local_group_grads_params.back();
+      group_p_g.reserve(group_size);
+      while (j < group_grads_params->size()) {
+        group_p_g.insert(group_p_g.end(), group_grads_params->at(j).begin(),
+                         group_grads_params->at(j).end());
+        ++j;
+        if (j % group_size == 0) break;
+      }
+    }
+    std::swap(*group_grads_params, local_group_grads_params);
+
+    VLOG(10) << "SetGroupAccordingToGroupSize(group_size: " << group_size
+             << "): ";
+    for (size_t i = 0; i < group_grads_params->size(); ++i) {
+      VLOG(10) << "group " << i;
+      std::stringstream out;
+      for (auto &p_g : group_grads_params->at(i)) {
+        out << "(" << p_g.second << ", " << p_g.first << "), ";
+      }
+      VLOG(10) << out.str();
+    }
+  }
+
+ private:
+  bool IsSupportedVarType(const proto::VarType::Type &type) const {
+    // Current only support LOD_TENSOR.
+    return type == proto::VarType::LOD_TENSOR;
+  }
+
+  void AppendAllocSpaceForVarsOp(const std::vector<std::string> &params_name,
+                                 const std::vector<std::string> &grads_name,
+                                 const std::string &fused_var_name,
+                                 BlockDesc *global_block) const {
+    auto op_desc = global_block->AppendOp();
+    op_desc->SetType("alloc_continuous_space");
+    op_desc->SetInput("Input", params_name);
+    op_desc->SetOutput("Output", grads_name);
+    op_desc->SetOutput("FusedOutput", {fused_var_name});
+  }
+
+  void RecordParamsAndGrads(ir::Node *node,
+                            ParamsAndGrads *params_grads) const {
+    try {
+      bool is_bk_op =
+          static_cast<bool>(boost::get<int>(node->Op()->GetAttr(
+                                OpProtoAndCheckerMaker::OpRoleAttrName())) &
+                            static_cast<int>(OpRole::kBackward));
+      if (!is_bk_op) return;
+
+      // Currently, we assume that once gradient is generated, it can be
+      // broadcast, and each gradient is only broadcast once.
+      auto backward_vars =
+          boost::get<std::vector<std::string>>(node->Op()->GetNullableAttr(
+              OpProtoAndCheckerMaker::OpRoleVarAttrName()));
+      PADDLE_ENFORCE_EQ(backward_vars.size() % 2, static_cast<size_t>(0));
+
+      for (size_t i = 0; i < backward_vars.size(); i += 2) {
+        VLOG(10) << "Trainable parameter: " << backward_vars[i]
+                 << ", gradient: " << backward_vars[i + 1];
+
+        params_grads->emplace_back(std::make_pair(
+            backward_vars[i] /*param*/, backward_vars[i + 1] /*grad*/));
+      }
+    } catch (boost::bad_get e) {
+    }
+  }
+
+  void InitFusedVarsAndAllocSpaceForVars(
+      const std::vector<platform::Place> &places,
+      const std::vector<Scope *> &local_scopes,
+      const std::unordered_map<std::string, ir::Node *> &vars,
+      const std::string &fused_var_name,
+      const ParamsAndGrads &params_grads) const {
+    //  Init Gradients and FusedVars
+    VLOG(10) << "Init FusedVars and Gradients.";
+    for (auto it = local_scopes.rbegin(); it != local_scopes.rend(); ++it) {
+      auto &scope = *it;
+
+      PADDLE_ENFORCE(scope->FindVar(fused_var_name) == nullptr,
+                     "%s has existed in scope.", fused_var_name);
+      scope->Var(fused_var_name)->GetMutable<LoDTensor>();
+
+      for (auto &p_g : params_grads) {
+        auto iter = vars.find(p_g.second);
+        PADDLE_ENFORCE(iter != vars.end());
+        PADDLE_ENFORCE_NOT_NULL(iter->second->Var());
+        PADDLE_ENFORCE_EQ(iter->second->Var()->GetType(),
+                          proto::VarType::LOD_TENSOR);
+        scope->Var(p_g.second)->GetMutable<LoDTensor>();
+      }
+    }
+
+    std::vector<std::string> grads_name;
+    std::vector<std::string> params_name;
+    grads_name.reserve(params_grads.size());
+    params_name.reserve(params_grads.size());
+    for (auto &p_g : params_grads) {
+      params_name.emplace_back(p_g.first);
+      grads_name.emplace_back(p_g.second);
+    }
+    framework::ProgramDesc program_desc;
+    AppendAllocSpaceForVarsOp(params_name, grads_name, fused_var_name,
+                              program_desc.MutableBlock(0));
+
+    // Run Only Once Programs
+    for (size_t i = 0; i < local_scopes.size(); ++i) {
+      for (auto &op_desc : program_desc.Block(0).AllOps()) {
+        auto op = OpRegistry::CreateOp(*op_desc);
+        op->Run(*local_scopes[i], places[i]);
+      }
+    }
+  }
+};
+
+}  // namespace details
+}  // namespace framework
+}  // namespace paddle
+
+REGISTER_PASS(alloc_continuous_space_for_grad_pass,
+              paddle::framework::details::AllocContinuousSpaceForGradPass)
+    .RequirePassAttr(paddle::framework::details::kPlaces)
+    .RequirePassAttr(paddle::framework::details::kLocalScopes);

paddle/fluid/framework/details/build_strategy.cc

@@ -46,7 +46,16 @@ class ParallelExecutorPassBuilder : public ir::PassBuilder {
  public:
   explicit ParallelExecutorPassBuilder(const BuildStrategy &strategy)
       : ir::PassBuilder(), strategy_(strategy) {
+    // Add a graph viz pass to record a graph.
+    if (!strategy_.debug_graphviz_path_.empty()) {
+      auto viz_pass = AppendPass("graph_viz_pass");
+      const std::string graph_path = string::Sprintf(
+          "%s%s", strategy_.debug_graphviz_path_.c_str(), "_original_graph");
+      viz_pass->Set<std::string>("graph_viz_path", new std::string(graph_path));
+    }
+
     if (strategy_.enable_sequential_execution_) {
+      VLOG(10) << "Add sequential_execution_pass";
       AppendPass("sequential_execution_pass");
     }
 
@@ -57,6 +66,7 @@ class ParallelExecutorPassBuilder : public ir::PassBuilder {
 
     // Add op fusion.
     if (strategy.fuse_relu_depthwise_conv_) {
+      VLOG(10) << "Add fuse_relu_depthwise_conv_pass";
       AppendPass("fuse_relu_depthwise_conv_pass");
     }
 
@@ -68,29 +78,30 @@ class ParallelExecutorPassBuilder : public ir::PassBuilder {
 
     // Add automatically inplace.
     if (strategy_.enable_inplace_) {
+      VLOG(10) << "Add inplace_pass";
       AppendPass("inplace_pass");
     }
 
+    if (strategy.fuse_elewise_add_act_ops_) {
+      VLOG(10) << "Add fuse_elewise_add_act_pass";
+      AppendPass("fuse_elewise_add_act_pass");
+    }
+
+    // for single card training, fuse_all_reduce_ops is unnecessary.
+    // alloc_continuous_space_for_grad_pass should be before of MultiDevPass.
+    if (strategy.fuse_all_reduce_ops_) {
+      VLOG(10) << "Add alloc_continuous_space_for_grad_pass";
+      AppendPass("alloc_continuous_space_for_grad_pass");
+    }
+
     // Add a graph viz pass to record a graph.
-    if (!strategy_.debug_graphviz_path_.empty()) {
+    if (!strategy.debug_graphviz_path_.empty()) {
       auto viz_pass = AppendPass("graph_viz_pass");
       const std::string graph_path = string::Sprintf(
-          "%s%s", strategy_.debug_graphviz_path_.c_str(), "_original_graph");
+          "%s%s", strategy.debug_graphviz_path_.c_str(), "_fused_graph");
       viz_pass->Set<std::string>("graph_viz_path", new std::string(graph_path));
     }
 
-    if (strategy.fuse_elewise_add_act_ops_) {
-      auto fuse_elewise_add_act_pass = AppendPass("fuse_elewise_add_act_pass");
-      // Add a graph viz pass to record a graph.
-      if (!strategy.debug_graphviz_path_.empty()) {
-        auto viz_pass = AppendPass("graph_viz_pass");
-        const std::string graph_path = string::Sprintf(
-            "%s%s", strategy.debug_graphviz_path_.c_str(), "_fused_graph");
-        viz_pass->Set<std::string>("graph_viz_path",
-                                   new std::string(graph_path));
-      }
-    }
-
     CollectiveContext *context = CollectiveContext::GetInstance();
     context->endpoints_ = strategy_.trainers_endpoints_;
     context->trainer_id_ = strategy_.trainer_id_;
@@ -108,11 +119,19 @@ class ParallelExecutorPassBuilder : public ir::PassBuilder {
     // A side-effect of that, memory optimize cannot forsee the fetched vars
     // , so fetchlist should be set persistable before call the Run interface.
     if (strategy.memory_optimize_) {
-      auto memory_optimize_pass = AppendPass("memory_optimize_pass");
+      VLOG(10) << "Add memory_optimize_pass";
+      AppendPass("memory_optimize_pass");
     }
 
     AppendMultiDevPass(strategy);
 
+    if (strategy.fuse_all_reduce_ops_) {
+      // NOTE: fuse_all_reduce_ops will count the number of all_reduce operator
+      // first, if the number is zero, fuse_all_reduce_ops will do nothing.
+      VLOG(10) << "Add fuse_all_reduce_op_pass";
+      AppendPass("fuse_all_reduce_op_pass");
+    }
+
     // Add a graph print pass to record a graph with device info.
     if (!strategy_.debug_graphviz_path_.empty()) {
       auto multi_devices_print_pass = AppendPass("multi_devices_print_pass");
@@ -129,27 +148,29 @@ class ParallelExecutorPassBuilder : public ir::PassBuilder {
     AppendPass("multi_devices_check_pass");
 
     if (SeqOnlyAllReduceOps(strategy)) {
+      VLOG(10) << "Add all_reduce_deps_pass";
       AppendPass("all_reduce_deps_pass");
     }
 
     if (strategy_.remove_unnecessary_lock_) {
+      VLOG(10) << "Add modify_op_lock_and_record_event_pass";
       AppendPass("modify_op_lock_and_record_event_pass");
     }
   }
 
   // Convert graph to run on multi-devices.
   void AppendMultiDevPass(const BuildStrategy &strategy) {
-    ir::Pass *multi_devices_pass;
+    ir::Pass *multi_devices_pass = nullptr;
     if (strategy_.is_distribution_) {
-      VLOG(3) << "multi device parameter server mode";
+      VLOG(10) << "Add dist_multi_devices_pass";
       multi_devices_pass = AppendPass("dist_multi_devices_pass").get();
     } else {
       if (strategy.reduce_ == BuildStrategy::ReduceStrategy::kAllReduce) {
-        VLOG(3) << "multi devices collective mode with allreduce";
+        VLOG(10) << "Add all_reduce_mode_multi_devices_pass";
         multi_devices_pass =
-            AppendPass("allreduce_mode_multi_devices_pass").get();
+            AppendPass("all_reduce_mode_multi_devices_pass").get();
       } else if (strategy.reduce_ == BuildStrategy::ReduceStrategy::kReduce) {
-        VLOG(3) << "multi deivces collective mode with reduce";
+        VLOG(10) << "Add reduce_mode_multi_devices_pass";
         multi_devices_pass = AppendPass("reduce_mode_multi_devices_pass").get();
       } else {
         PADDLE_THROW("Unknown reduce strategy.");
@@ -206,9 +227,26 @@ std::unique_ptr<ir::Graph> BuildStrategy::Apply(
 
 #if defined(PADDLE_WITH_CUDA) && !defined(_WIN32)
       platform::NCCLContextMap *nctx = use_cuda ? nccl_ctxs : nullptr;
-      pass->Erase("nccl_ctxs");
-      pass->SetNotOwned<platform::NCCLContextMap>("nccl_ctxs", nctx);
+      pass->Erase(kNCCLCtxs);
+      pass->SetNotOwned<platform::NCCLContextMap>(kNCCLCtxs, nctx);
 #endif
+    } else if (pass->Type() == "fuse_all_reduce_op_pass") {
+      pass->Erase(kPlaces);
+      pass->SetNotOwned<const std::vector<platform::Place>>(kPlaces, &places);
+      pass->Erase(kLocalScopes);
+      pass->SetNotOwned<const std::vector<Scope *>>(kLocalScopes,
+                                                    &local_scopes);
+#if defined(PADDLE_WITH_CUDA) && !defined(_WIN32)
+      platform::NCCLContextMap *nctx = use_cuda ? nccl_ctxs : nullptr;
+      pass->Erase(kNCCLCtxs);
+      pass->SetNotOwned<platform::NCCLContextMap>(kNCCLCtxs, nctx);
+#endif
+    } else if (pass->Type() == "alloc_continuous_space_for_grad_pass") {
+      pass->Erase(kPlaces);
+      pass->SetNotOwned<const std::vector<platform::Place>>(kPlaces, &places);
+      pass->Erase(kLocalScopes);
+      pass->SetNotOwned<const std::vector<Scope *>>(kLocalScopes,
+                                                    &local_scopes);
     } else if (pass->Type() == "sequential_execution_pass") {
       LOG(INFO) << "set enable_sequential_execution:"
                 << enable_sequential_execution_;
@@ -239,7 +277,7 @@ USE_PASS(fuse_elewise_add_act_pass);
 USE_PASS(graph_viz_pass);
 USE_PASS(multi_batch_merge_pass);
 USE_PASS(reduce_mode_multi_devices_pass);
-USE_PASS(allreduce_mode_multi_devices_pass);
+USE_PASS(all_reduce_mode_multi_devices_pass);
 USE_PASS(dist_multi_devices_pass);
 USE_PASS(multi_devices_check_pass);
 USE_PASS(multi_devices_print_pass);
@@ -249,4 +287,6 @@ USE_PASS(all_reduce_deps_pass);
 USE_PASS(modify_op_lock_and_record_event_pass);
 USE_PASS(inplace_pass);
 USE_PASS(lock_free_optimize_pass);
+USE_PASS(alloc_continuous_space_for_grad_pass);
 USE_PASS(graph_to_program_pass);
+USE_PASS(fuse_all_reduce_op_pass);

paddle/fluid/framework/details/build_strategy.h

@@ -16,6 +16,7 @@
 
 #include <memory>
 #include <string>
+#include <utility>
 #include <vector>
 
 #include "paddle/fluid/framework/ir/pass_builder.h"
@@ -75,6 +76,8 @@ struct BuildStrategy {
 
   bool fuse_elewise_add_act_ops_{false};
 
+  bool fuse_all_reduce_ops_{false};
+
   bool fuse_relu_depthwise_conv_{false};
 
   bool sync_batch_norm_{false};

paddle/fluid/framework/details/fuse_all_reduce_op_pass.cc

+//   Copyright (c) 2019 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 <algorithm>
+#include <string>
+#include <vector>
+
+#include "paddle/fluid/framework/details/all_reduce_op_handle.h"
+#include "paddle/fluid/framework/details/container_cast.h"
+#include "paddle/fluid/framework/details/fused_all_reduce_op_handle.h"
+#include "paddle/fluid/framework/details/multi_devices_helper.h"
+#include "paddle/fluid/framework/ir/graph_helper.h"
+
+namespace paddle {
+namespace framework {
+namespace details {
+
+class FuseAllReduceOpPass : public ir::Pass {
+ protected:
+  std::unique_ptr<ir::Graph> ApplyImpl(
+      std::unique_ptr<ir::Graph> graph) const override {
+    ir::Graph &result = *graph;
+
+    auto &places = Get<const std::vector<platform::Place>>(kPlaces);
+    auto &local_scopes = Get<const std::vector<Scope *>>(kLocalScopes);
+#if defined(PADDLE_WITH_CUDA) && !defined(_WIN32)
+    auto *nccl_ctxs = &Get<platform::NCCLContextMap>(kNCCLCtxs);
+#endif
+
+    std::unordered_set<std::string> grads;
+    auto &params_grads = result.Get<ParamsAndGrads>(kParamsAndGrads);
+    size_t num_of_all_reduce = params_grads.size();
+    grads.reserve(num_of_all_reduce);
+    for (auto p_g : params_grads) {
+      grads.insert(p_g.second);
+    }
+
+    size_t num_place = places.size();
+    std::unordered_map<std::string, ir::Node *> all_reduce_ops;
+    all_reduce_ops.reserve(grads.size());
+    for (auto &node : result.Nodes()) {
+      if (node->IsOp()) {
+        PADDLE_ENFORCE(node->IsWrappedBy<OpHandleBase>());
+        auto *all_reduce_op_handle =
+            dynamic_cast<AllReduceOpHandle *>(&node->Wrapper<OpHandleBase>());
+        if (all_reduce_op_handle) {
+          auto inputs = DynamicCast<VarHandle>(all_reduce_op_handle->Inputs());
+          PADDLE_ENFORCE_EQ(inputs.size(), num_place);
+          // The inputs' name should be the same.
+          auto &grad_name = inputs[0]->name();
+          for (size_t i = 1; i < inputs.size(); ++i) {
+            PADDLE_ENFORCE_EQ(inputs[i]->name(), grad_name,
+                              "The input name should be the same.");
+          }
+          PADDLE_ENFORCE_NE(grads.count(grad_name), static_cast<size_t>(0));
+          all_reduce_ops.emplace(grad_name, node);
+        }
+      }
+    }
+
+    VLOG(10) << "Find all_reduce_ops: " << all_reduce_ops.size();
+    if (all_reduce_ops.size() == 0) {
+      return std::move(graph);
+    }
+
+    PADDLE_ENFORCE_EQ(all_reduce_ops.size(), grads.size(),
+                      "The number of all_reduce OpHandle is not equal to the "
+                      "number of grads. Maybe some gradients are sparse type, "
+                      "it is not supported currently.");
+    VLOG(10) << "Insert fused_all_reduce";
+
+    auto &group_grads_params =
+        graph->Get<GroupGradsAndParams>(kGroupGradsAndParams);
+
+    for (auto &group_g_p : group_grads_params) {
+      size_t group_size = group_g_p.size();
+      PADDLE_ENFORCE_GT(group_size, static_cast<size_t>(0));
+      std::vector<ir::Node *> group_all_reduce_ops;
+      group_all_reduce_ops.reserve(group_size);
+      for (auto &g_p : group_g_p) {
+        group_all_reduce_ops.emplace_back(all_reduce_ops.at(g_p.first));
+      }
+#if defined(PADDLE_WITH_CUDA) && !defined(_WIN32)
+      InsertFusedAllReduce(places, local_scopes, group_size,
+                           group_all_reduce_ops, nccl_ctxs, &result);
+#else
+      InsertFusedAllReduce(places, local_scopes, group_size,
+                           group_all_reduce_ops, &result);
+#endif
+    }
+    return std::move(graph);
+  }
+
+  void InsertFusedAllReduce(const std::vector<platform::Place> &places,
+                            const std::vector<Scope *> &local_scopes,
+                            const size_t num_of_all_reduce,
+                            const std::vector<ir::Node *> &all_reduce_ops,
+#if defined(PADDLE_WITH_CUDA) && !defined(_WIN32)
+                            const platform::NCCLContextMap *nccl_ctxs,
+#endif
+                            ir::Graph *result) const {
+    std::vector<VarHandleBase *> inputs;
+    std::vector<VarHandleBase *> outputs;
+    for (auto &op : all_reduce_ops) {
+      auto &op_handle = op->Wrapper<OpHandleBase>();
+      inputs.insert(inputs.end(), op_handle.Inputs().begin(),
+                    op_handle.Inputs().end());
+      // Remove output
+      for_each(op_handle.Inputs().begin(), op_handle.Inputs().end(),
+               [&op_handle](VarHandleBase *var_handle) {
+                 var_handle->RemoveOutput(&op_handle, op_handle.Node());
+               });
+
+      outputs.insert(outputs.end(), op_handle.Outputs().begin(),
+                     op_handle.Outputs().end());
+      // Remove Input
+      for_each(
+          op_handle.Outputs().begin(), op_handle.Outputs().end(),
+          [](VarHandleBase *var_handle) { var_handle->ClearGeneratedOp(); });
+
+      result->RemoveNode(op_handle.Node());
+    }
+
+#if defined(PADDLE_WITH_CUDA) && !defined(_WIN32)
+    CreateFusedAllReduceOp(inputs, outputs, num_of_all_reduce, places,
+                           local_scopes, nccl_ctxs, result);
+#else
+    CreateFusedAllReduceOp(inputs, outputs, num_of_all_reduce, places,
+                           local_scopes, result);
+#endif
+  }
+
+ private:
+  void CreateFusedAllReduceOp(const std::vector<VarHandleBase *> &inputs,
+                              const std::vector<VarHandleBase *> &outputs,
+                              const size_t num_of_all_reduce,
+                              const std::vector<platform::Place> &places,
+                              const std::vector<Scope *> &local_scopes,
+#if defined(PADDLE_WITH_CUDA) && !defined(_WIN32)
+                              const platform::NCCLContextMap *nccl_ctxs,
+#endif
+                              ir::Graph *result) const {
+#if defined(PADDLE_WITH_CUDA) && !defined(_WIN32)
+    auto *op_handle = new FusedAllReduceOpHandle(
+        result->CreateEmptyNode("fused_all_reduce", ir::Node::Type::kOperation),
+        local_scopes, places, num_of_all_reduce, nccl_ctxs);
+#else
+    auto *op_handle = new FusedAllReduceOpHandle(
+        result->CreateEmptyNode("fused_all_reduce", ir::Node::Type::kOperation),
+        local_scopes, places, num_of_all_reduce);
+#endif
+
+    for (auto in : inputs) {
+      op_handle->AddInput(in);
+    }
+
+    for (auto out : outputs) {
+      op_handle->AddOutput(out);
+    }
+
+#if defined(PADDLE_WITH_CUDA) && !defined(_WIN32)
+    if (!nccl_ctxs) {
+      SetCommunicationContext(places, op_handle);
+    }
+#else
+    SetCommunicationContext(places, op_handle);
+#endif
+  }
+
+  void SetCommunicationContext(const std::vector<platform::Place> &places,
+                               FusedAllReduceOpHandle *op_handle) const {
+    for (size_t i = 0; i < places.size(); ++i) {
+      op_handle->SetDeviceContext(
+          places[i], platform::DeviceContextPool::Instance().Get(places[i]));
+    }
+  }
+};
+
+}  // namespace details
+}  // namespace framework
+}  // namespace paddle
+
+REGISTER_PASS(fuse_all_reduce_op_pass,
+              paddle::framework::details::FuseAllReduceOpPass);

paddle/fluid/framework/details/fused_all_reduce_op_handle.cc

+//   Copyright (c) 2019 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/details/fused_all_reduce_op_handle.h"
+#include <algorithm>
+#include <utility>
+#include "paddle/fluid/framework/details/container_cast.h"
+#include "paddle/fluid/framework/details/reduce_and_gather.h"
+#include "paddle/fluid/framework/details/variable_visitor.h"
+#include "paddle/fluid/platform/profiler.h"
+
+DEFINE_bool(skip_fused_all_reduce_check, false, "");
+namespace paddle {
+namespace framework {
+namespace details {
+
+typedef std::vector<std::vector<std::pair<std::string, const LoDTensor *>>>
+    GradientAndLoDTensor;
+
+#if defined(PADDLE_WITH_CUDA) && !defined(_WIN32)
+FusedAllReduceOpHandle::FusedAllReduceOpHandle(
+    ir::Node *node, const std::vector<Scope *> &local_scopes,
+    const std::vector<platform::Place> &places, const size_t num_of_all_reduce,
+    const platform::NCCLContextMap *ctxs)
+    : OpHandleBase(node),
+      local_scopes_(local_scopes),
+      places_(places),
+      num_of_all_reduce_(num_of_all_reduce),
+      nccl_ctxs_(ctxs) {
+  if (nccl_ctxs_) {
+    for (auto &p : places_) {
+      this->SetDeviceContext(p, nccl_ctxs_->DevCtx(p));
+    }
+  }
+  PADDLE_ENFORCE_EQ(places_.size(), local_scopes_.size());
+}
+#else
+
+FusedAllReduceOpHandle::FusedAllReduceOpHandle(
+    ir::Node *node, const std::vector<Scope *> &local_scopes,
+    const std::vector<platform::Place> &places, const size_t num_of_all_reduce)
+    : OpHandleBase(node),
+      local_scopes_(local_scopes),
+      places_(places),
+      num_of_all_reduce_(num_of_all_reduce) {
+  PADDLE_ENFORCE_EQ(places_.size(), local_scopes_.size());
+}
+
+#endif
+
+void FusedAllReduceOpHandle::RunImpl() {
+  platform::RecordEvent record_event(Name());
+
+  VLOG(4) << this->DebugString();
+
+  WaitInputVarGenerated();
+  // The input: grad0(dev0), grad0(dev1), grad1(dev0), grad1(dev1)...
+  // The output: grad0(dev0), grad0(dev1), grad1(dev0), grad1(dev1)...
+  auto in_var_handles = DynamicCast<VarHandle>(this->Inputs());
+  auto out_var_handles = DynamicCast<VarHandle>(this->Outputs());
+
+  size_t place_num = places_.size();
+  PADDLE_ENFORCE_EQ(
+      in_var_handles.size(), place_num * num_of_all_reduce_,
+      "The NoDummyInputSize should be equal to the number of places.");
+  PADDLE_ENFORCE_EQ(
+      in_var_handles.size(), out_var_handles.size(),
+      "The NoDummyInputSize and NoDummyOutputSize should be equal.");
+
+  GradientAndLoDTensor grads_tensor;
+  grads_tensor.resize(place_num);
+
+  int64_t numel = -1;
+  auto dtype = static_cast<framework::proto::VarType::Type>(0);
+  for (size_t scope_idx = 0; scope_idx < local_scopes_.size(); ++scope_idx) {
+    auto &g_tensor = grads_tensor.at(scope_idx);
+    g_tensor.reserve(num_of_all_reduce_);
+
+    GetGradLoDTensor(scope_idx, in_var_handles, out_var_handles, &g_tensor);
+
+    int64_t element_num = 0;
+    framework::proto::VarType::Type ele_dtype =
+        static_cast<framework::proto::VarType::Type>(0);
+    GetDTypeAndNumel(g_tensor, &ele_dtype, &element_num);
+
+    if (numel == -1) {
+      numel = element_num;
+    }
+    if (dtype == static_cast<framework::proto::VarType::Type>(0)) {
+      dtype = ele_dtype;
+      PADDLE_ENFORCE_NE(ele_dtype,
+                        static_cast<framework::proto::VarType::Type>(0));
+    }
+    PADDLE_ENFORCE_EQ(ele_dtype, dtype);
+
+    // Check whether the address space is contiguous.
+    std::sort(
+        g_tensor.begin(), g_tensor.end(),
+        [](const std::pair<std::string, const LoDTensor *> &grad1,
+           const std::pair<std::string, const LoDTensor *> &grad2) -> bool {
+          return grad1.second->data<void>() < grad2.second->data<void>();
+        });
+
+    for (size_t k = 1; k < g_tensor.size(); ++k) {
+      const void *pre_address = g_tensor.at(k - 1).second->data<void>();
+      int64_t len = g_tensor.at(k - 1).second->numel();
+      auto offset = len * framework::SizeOfType(dtype);
+      void *next_address = reinterpret_cast<void *>(
+          reinterpret_cast<uintptr_t>(pre_address) + offset);
+      const void *cur_address = g_tensor.at(k).second->data<void>();
+      VLOG(10) << k << ", "
+               << " pre_address(" << g_tensor.at(k - 1).first
+               << "): " << pre_address << ", cur_address("
+               << g_tensor.at(k).first << "): " << cur_address
+               << ", offset:" << offset << ", " << next_address << ", "
+               << cur_address;
+      PADDLE_ENFORCE_EQ(next_address, cur_address);
+    }
+  }
+
+  if (!FLAGS_skip_fused_all_reduce_check) {
+    for (size_t scope_idx = 0; scope_idx < place_num; ++scope_idx) {
+      for (size_t j = 1; j < num_of_all_reduce_; ++j) {
+        PADDLE_ENFORCE_EQ(grads_tensor.at(0).at(j).first,
+                          grads_tensor.at(scope_idx).at(j).first);
+      }
+    }
+  }
+
+  std::vector<const void *> lod_tensor_data;
+  for (size_t scope_idx = 0; scope_idx < place_num; ++scope_idx) {
+    auto data = grads_tensor.at(scope_idx).at(0).second->data<void>();
+    lod_tensor_data.emplace_back(data);
+  }
+
+  if (platform::is_gpu_place(places_[0])) {
+#if defined(PADDLE_WITH_CUDA) && !defined(_WIN32)
+    PADDLE_ENFORCE(nccl_ctxs_, "nccl_ctxs should not be nullptr.");
+    int nccl_dtype = platform::ToNCCLDataType(dtype);
+    std::vector<std::function<void()>> all_reduce_calls;
+    for (size_t i = 0; i < local_scopes_.size(); ++i) {
+      auto &p = places_[i];
+      void *buffer = const_cast<void *>(lod_tensor_data.at(i));
+
+      int dev_id = boost::get<platform::CUDAPlace>(p).device;
+      auto &nccl_ctx = nccl_ctxs_->at(dev_id);
+      auto stream = nccl_ctx.stream();
+      auto comm = nccl_ctx.comm_;
+      all_reduce_calls.emplace_back([=] {
+        PADDLE_ENFORCE(platform::dynload::ncclAllReduce(
+            buffer, buffer, numel, static_cast<ncclDataType_t>(nccl_dtype),
+            ncclSum, comm, stream));
+      });
+    }
+
+    this->RunAndRecordEvent([&] {
+      if (all_reduce_calls.size() == 1UL) {
+        // Do not use NCCLGroup when manage NCCL by per thread per device
+        all_reduce_calls[0]();
+      } else {
+        platform::NCCLGroupGuard guard;
+        for (auto &call : all_reduce_calls) {
+          call();
+        }
+      }
+    });
+#else
+    PADDLE_THROW("Not compiled with CUDA");
+#endif
+  } else {
+    // Special handle CPU only Operator's gradient. Like CRF
+    auto grad_name = grads_tensor.at(0).at(0).first;
+    auto &trg = *this->local_scopes_[0]
+                     ->FindVar(kLocalExecScopeName)
+                     ->Get<Scope *>()
+                     ->FindVar(grad_name)
+                     ->GetMutable<framework::LoDTensor>();
+
+    // Reduce All data to trg in CPU
+    ReduceBufferData func(lod_tensor_data, trg.data<void>(), numel);
+    VisitDataType(trg.type(), func);
+
+    for (size_t i = 1; i < local_scopes_.size(); ++i) {
+      auto &scope =
+          *local_scopes_[i]->FindVar(kLocalExecScopeName)->Get<Scope *>();
+      auto &p = places_[i];
+      auto *var = scope.FindVar(grad_name);
+      auto *dev_ctx = dev_ctxes_.at(p);
+      size_t size = numel * SizeOfType(trg.type());
+      RunAndRecordEvent(p, [&trg, var, dev_ctx, p, size] {
+        auto dst_ptr = var->GetMutable<framework::LoDTensor>()->data<void>();
+        platform::CPUPlace cpu_place;
+        memory::Copy(cpu_place, dst_ptr, cpu_place, trg.data<void>(), size);
+      });
+    }
+  }
+}
+
+void FusedAllReduceOpHandle::GetGradLoDTensor(
+    const size_t &scope_idx, const std::vector<VarHandle *> &in_var_handles,
+    const std::vector<VarHandle *> &out_var_handles,
+    std::vector<std::pair<std::string, const LoDTensor *>> *grad_tensor) const {
+  auto *local_scope =
+      local_scopes_.at(scope_idx)->FindVar(kLocalExecScopeName)->Get<Scope *>();
+  size_t place_num = places_.size();
+
+  for (size_t j = 0; j < in_var_handles.size(); j += place_num) {
+    auto var_name = in_var_handles[j]->name();
+    PADDLE_ENFORCE_EQ(var_name, out_var_handles[j]->name());
+    auto &lod_tensor = local_scope->FindVar(var_name)->Get<LoDTensor>();
+    PADDLE_ENFORCE_EQ(lod_tensor.place(), places_.at(scope_idx));
+    grad_tensor->emplace_back(std::make_pair(var_name, &lod_tensor));
+  }
+}
+
+void FusedAllReduceOpHandle::GetDTypeAndNumel(
+    const std::vector<std::pair<std::string, const LoDTensor *>> &grad_tensor,
+    proto::VarType::Type *dtype, int64_t *numel) const {
+  *numel = 0;
+  for (size_t i = 0; i < grad_tensor.size(); ++i) {
+    // Get element number
+    int64_t len = grad_tensor.at(i).second->numel();
+    PADDLE_ENFORCE_GT(len, 0);
+    *numel += len;
+
+    // Get dtype
+    auto ele_type = grad_tensor.at(i).second->type();
+    if (i == 0) {
+      *dtype = ele_type;
+    }
+    PADDLE_ENFORCE_EQ(ele_type, *dtype);
+  }
+}
+
+std::string FusedAllReduceOpHandle::Name() const { return "fused_all_reduce"; }
+}  // namespace details
+}  // namespace framework
+}  // namespace paddle

paddle/fluid/framework/details/fused_all_reduce_op_handle.h

+//   Copyright (c) 2019 PaddlePaddle Authors. All Rights Reserved.
+//
+// Licensed under the Apache License, Version 2.0 (the "License");
+// you may not use this file except in compliance with the License.
+// You may obtain a copy of the License at
+//
+//     http://www.apache.org/licenses/LICENSE-2.0
+//
+// Unless required by applicable law or agreed to in writing, software
+// distributed under the License is distributed on an "AS IS" BASIS,
+// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+// See the License for the specific language governing permissions and
+// limitations under the License.
+
+#pragma once
+
+#include <string>
+#include <utility>
+#include <vector>
+#include "paddle/fluid/framework/details/op_handle_base.h"
+#include "paddle/fluid/framework/lod_tensor.h"
+#include "paddle/fluid/framework/scope.h"
+#if defined(PADDLE_WITH_CUDA) && !defined(_WIN32)
+#include "paddle/fluid/platform/nccl_helper.h"
+#endif
+
+namespace paddle {
+namespace framework {
+namespace details {
+
+struct FusedAllReduceOpHandle : public OpHandleBase {
+#if defined(PADDLE_WITH_CUDA) && !defined(_WIN32)
+  FusedAllReduceOpHandle(ir::Node *node,
+                         const std::vector<Scope *> &local_scopes,
+                         const std::vector<platform::Place> &places,
+                         const size_t num_of_all_reduce,
+                         const platform::NCCLContextMap *ctxs);
+#else
+  FusedAllReduceOpHandle(ir::Node *node,
+                         const std::vector<Scope *> &local_scopes,
+                         const std::vector<platform::Place> &places,
+                         const size_t num_of_all_reduce);
+#endif
+  std::string Name() const override;
+
+  // Delay and buffer nccl_all_reduce together can significantly increase
+  // performance. Disable this feature by returning false.
+  bool IsMultiDeviceTransfer() override { return true; };
+
+ protected:
+  void RunImpl() override;
+
+ private:
+  std::vector<Scope *> local_scopes_;
+  std::vector<platform::Place> places_;
+  size_t num_of_all_reduce_;
+#if defined(PADDLE_WITH_CUDA) && !defined(_WIN32)
+  const platform::NCCLContextMap *nccl_ctxs_;
+#endif
+
+  // Check the dtype of the input
+  void GetDTypeAndNumel(
+      const std::vector<std::pair<std::string, const LoDTensor *>> &g_tensor,
+      proto::VarType::Type *dtype, int64_t *total_num) const;
+
+  // Get gradient's name and LoDTensor
+  void GetGradLoDTensor(const size_t &scope_idx,
+                        const std::vector<VarHandle *> &in_var_handles,
+                        const std::vector<VarHandle *> &out_var_handles,
+                        std::vector<std::pair<std::string, const LoDTensor *>>
+                            *grad_tensor) const;
+};
+
+}  // namespace details
+}  // namespace framework
+}  // namespace paddle

paddle/fluid/framework/details/graph_test_base.h

@@ -68,11 +68,11 @@ class SplitOpMaker : public OpProtoAndCheckerMaker {
 
 class DummyVarTypeInference : public VarTypeInference {
  public:
-  void operator()(const OpDesc& op_desc, BlockDesc* block) const override {
-    auto& inputs = op_desc.Input("X");
-    auto type = block->Var(inputs.front())->GetType();
-    auto out_var_name = op_desc.Output("Out").front();
-    block->Var(out_var_name)->SetType(type);
+  void operator()(framework::InferVarTypeContext* ctx) const override {
+    auto& inputs = ctx->Input("X");
+    auto type = ctx->GetType(inputs.front());
+    auto out_var_name = ctx->Output("Out").front();
+    ctx->SetType(out_var_name, type);
   }
 };
 

paddle/fluid/framework/details/multi_devices_graph_pass.cc

@@ -11,18 +11,17 @@
 // 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/details/multi_devices_graph_pass.h"
 #include <algorithm>
 #include <fstream>
 #include <string>
 #include <utility>
 #include <vector>
-
 #include "paddle/fluid/framework/details/all_reduce_op_handle.h"
 #include "paddle/fluid/framework/details/broadcast_op_handle.h"
 #include "paddle/fluid/framework/details/computation_op_handle.h"
 #include "paddle/fluid/framework/details/data_balance_op_handle.h"
 #include "paddle/fluid/framework/details/fused_broadcast_op_handle.h"
-#include "paddle/fluid/framework/details/multi_devices_graph_pass.h"
 #include "paddle/fluid/framework/details/reduce_op_handle.h"
 #include "paddle/fluid/framework/details/rpc_op_handle.h"
 #include "paddle/fluid/framework/details/scale_loss_grad_op_handle.h"
@@ -134,21 +133,26 @@ void AddOutputToLeafOps(ir::Graph *graph) {
 }
 }  // namespace
 
+void MultiDevSSAGraphBuilderBase::CheckGraph(const ir::Graph &graph) const {}
+
 void MultiDevSSAGraphBuilderBase::Init() const {
   all_vars_.clear();
 
   loss_var_name_ = Get<const std::string>(kLossVarName);
+  VLOG(10) << "Init MultiDevSSAGraphBuilder, loss name: " << loss_var_name_;
   places_ = Get<const std::vector<platform::Place>>(kPlaces);
   local_scopes_ = Get<const std::vector<Scope *>>(kLocalScopes);
   strategy_ = Get<const BuildStrategy>(kStrategy);
 #if defined(PADDLE_WITH_CUDA) && !defined(_WIN32)
-  nccl_ctxs_ = &Get<platform::NCCLContextMap>("nccl_ctxs");
+  nccl_ctxs_ = &Get<platform::NCCLContextMap>(kNCCLCtxs);
 #endif
+  PADDLE_ENFORCE_EQ(places_.size(), local_scopes_.size());
 }
 
 std::unique_ptr<ir::Graph> MultiDevSSAGraphBuilderBase::ApplyImpl(
     std::unique_ptr<ir::Graph> graph) const {
   Init();
+  CheckGraph(*graph);
   std::vector<ir::Node *> sorted_ops = SortOperations(*graph);
 
   auto nodes = graph->ReleaseNodes();
@@ -166,7 +170,6 @@ std::unique_ptr<ir::Graph> MultiDevSSAGraphBuilderBase::ApplyImpl(
   result.Set(kGraphOps, new GraphOps);
 
   bool is_forwarding = true;
-  bool insert_collection_ops = NeedCollectiveOps();
 
   for (ir::Node *node : sorted_ops) {
     if (DealWithSpecialOp(&result, node)) {
@@ -185,8 +188,8 @@ std::unique_ptr<ir::Graph> MultiDevSSAGraphBuilderBase::ApplyImpl(
         CreateComputationalOps(&result, node, places_.size());
       }
 
-      // Insert collection ops
-      if (!is_forwarding && insert_collection_ops) {
+      // Insert collective ops if nranks > 1
+      if (!is_forwarding && Get<size_t>(kNRanks) > 1) {
         try {
           bool is_bk_op =
               static_cast<bool>(boost::get<int>(node->Op()->GetAttr(
@@ -200,13 +203,13 @@ std::unique_ptr<ir::Graph> MultiDevSSAGraphBuilderBase::ApplyImpl(
               boost::get<std::vector<std::string>>(node->Op()->GetNullableAttr(
                   OpProtoAndCheckerMaker::OpRoleVarAttrName()));
           PADDLE_ENFORCE_EQ(backward_vars.size() % 2, 0);
-
           for (size_t i = 0; i < backward_vars.size(); i += 2) {
             auto &p_name = backward_vars[i];
             auto &g_name = backward_vars[i + 1];
             VLOG(10) << "Bcast " << g_name << " for parameter " << p_name;
-
-            InsertCollectiveOp(&result, p_name, g_name);
+            if (NeedCollectiveForGrad(g_name, sorted_ops)) {
+              InsertCollectiveOp(&result, p_name, g_name);
+            }
           }
         } catch (boost::bad_get e) {
         }
@@ -226,6 +229,7 @@ std::unique_ptr<ir::Graph> MultiDevSSAGraphBuilderBase::ApplyImpl(
    * Only variables should be the leaves of graph.
    */
   AddOutputToLeafOps(&result);
+
   result.Erase(kGraphOps);
   return graph;
 }
@@ -258,6 +262,11 @@ void MultiDevSSAGraphBuilderBase::InsertScaleLossGradOp(
   }
 }
 
+bool MultiDevSSAGraphBuilderBase::DealWithSpecialOp(ir::Graph *result,
+                                                    ir::Node *node) const {
+  return false;
+}
+
 std::vector<ir::Node *> MultiDevSSAGraphBuilderBase::SortOperations(
     const ir::Graph &graph) const {
   return ir::TopologySortOperations(graph);
@@ -271,8 +280,20 @@ bool MultiDevSSAGraphBuilderBase::UseGPU() const {
   return use_gpu;
 }
 
-bool MultiDevSSAGraphBuilderBase::NeedCollectiveOps() const {
-  return Get<size_t>(kNRanks) > 1;
+bool MultiDevSSAGraphBuilderBase::NeedCollectiveForGrad(
+    const std::string &grad_name, std::vector<ir::Node *> ops) const {
+  // if we have allreduce_op for current gradient variable in the graph,
+  // then we don't need to add allreduce_op_handle for this gradient
+  // NOTE: This is for the case that all gradients should add collective ops
+  for (auto *node : ops) {
+    if (node->Op()->Type() != "allreduce") continue;
+    for (auto in_name : node->Op()->InputArgumentNames()) {
+      if (in_name == grad_name) {
+        return false;
+      }
+    }
+  }
+  return true;
 }
 
 void MultiDevSSAGraphBuilderBase::CreateOpHandleIOs(ir::Graph *result,
@@ -496,20 +517,17 @@ VarHandle *MultiDevSSAGraphBuilderBase::CreateReduceOp(ir::Graph *result,
 }
 
 bool MultiDevSSAGraphBuilderBase::IsScaleLossOp(ir::Node *node) const {
-  return boost::get<int>(
+  return !loss_var_name_.empty() && node->Op() &&
+         boost::get<int>(
              node->Op()->GetAttr(OpProtoAndCheckerMaker::OpRoleAttrName())) ==
              (static_cast<int>(OpRole::kBackward) |
-              static_cast<int>(OpRole::kLoss)) &&
-         !loss_var_name_.empty();  // If loss_var is empty. This is test mode
+              static_cast<int>(OpRole::kLoss));
 }
 
 bool MultiDevSSAGraphBuilderBase::IsSparseGradient(
     const std::string &og) const {
   PADDLE_ENFORCE(all_vars_.count(og) != 0);
-  if (all_vars_.at(og)->GetType() == proto::VarType::SELECTED_ROWS) {
-    return true;
-  }
-  return false;
+  return all_vars_.at(og)->GetType() == proto::VarType::SELECTED_ROWS;
 }
 
 void AllReduceSSAGraphBuilder::InsertCollectiveOp(
@@ -995,7 +1013,7 @@ static int MultiDevSSAGraphBuilderRegister(const std::string &builder_mode) {
 REGISTER_MULTI_DEVICES_PASS(reduce_mode_multi_devices_pass,
                             paddle::framework::details::ReduceSSAGraphBuilder);
 REGISTER_MULTI_DEVICES_PASS(
-    allreduce_mode_multi_devices_pass,
+    all_reduce_mode_multi_devices_pass,
     paddle::framework::details::AllReduceSSAGraphBuilder);
 REGISTER_MULTI_DEVICES_PASS(dist_multi_devices_pass,
                             paddle::framework::details::DistSSAGraphBuilder);

paddle/fluid/framework/details/multi_devices_graph_pass.h

@@ -14,7 +14,10 @@
 
 #pragma once
 
+#include <memory>
 #include <string>
+#include <unordered_map>
+#include <unordered_set>
 #include <utility>
 #include <vector>
 
@@ -31,12 +34,6 @@ namespace framework {
 class Scope;
 namespace details {
 
-constexpr char kLossVarName[] = "loss_var_name";
-constexpr char kPlaces[] = "places";
-constexpr char kLocalScopes[] = "local_scopes";
-constexpr char kStrategy[] = "strategy";
-constexpr char kNRanks[] = "nranks";
-
 class MultiDevSSAGraphBuilderBase : public ir::Pass {
  protected:
   std::unique_ptr<ir::Graph> ApplyImpl(
@@ -44,18 +41,21 @@ class MultiDevSSAGraphBuilderBase : public ir::Pass {
 
   virtual void Init() const;
 
+  virtual void CheckGraph(const ir::Graph &graph) const;
+
   virtual std::vector<ir::Node *> SortOperations(const ir::Graph &graph) const;
 
   virtual void InsertCollectiveOp(ir::Graph *result, const std::string &p_name,
                                   const std::string &g_name) const = 0;
 
-  virtual bool DealWithSpecialOp(ir::Graph *result, ir::Node *node) const = 0;
+  virtual bool DealWithSpecialOp(ir::Graph *result, ir::Node *node) const;
 
   virtual void InsertPostprocessOps(ir::Graph *result) const = 0;
 
   bool UseGPU() const;
 
-  bool NeedCollectiveOps() const;
+  bool NeedCollectiveForGrad(const std::string &grad_name,
+                             std::vector<ir::Node *> ops) const;
 
   bool IsScaleLossOp(ir::Node *node) const;
 
@@ -109,10 +109,6 @@ class AllReduceSSAGraphBuilder : public MultiDevSSAGraphBuilderBase {
   virtual void InsertCollectiveOp(ir::Graph *result, const std::string &p_name,
                                   const std::string &g_name) const;
 
-  virtual bool DealWithSpecialOp(ir::Graph *result, ir::Node *node) const {
-    return false;
-  }
-
   virtual void InsertPostprocessOps(ir::Graph *result) const {}
 };
 

paddle/fluid/framework/details/multi_devices_helper.h

@@ -16,6 +16,9 @@
 
 #include <memory>
 #include <string>
+#include <unordered_map>
+#include <unordered_set>
+#include <utility>
 #include <vector>
 
 #include "paddle/fluid/framework/details/op_handle_base.h"
@@ -44,6 +47,26 @@ const char kGraphVars[] = "vars";
 typedef std::unordered_set<VarHandleBase *> GraphDepVars;
 const char kGraphDepVars[] = "dep_vars";
 
+constexpr char kNCCLCtxs[] = "nccl_ctxs";
+
+constexpr char kLossVarName[] = "loss_var_name";
+constexpr char kPlaces[] = "places";
+constexpr char kLocalScopes[] = "local_scopes";
+constexpr char kStrategy[] = "strategy";
+constexpr char kNRanks[] = "nranks";
+
+typedef std::unordered_set<std::string> FusedVars;
+constexpr char kFusedVars[] = "fused_vars";
+
+typedef std::vector<std::pair<std::string, std::string>> ParamsAndGrads;
+constexpr char kParamsAndGrads[] = "params_grads";
+
+typedef std::vector<std::vector<std::pair<std::string, std::string>>>
+    GroupGradsAndParams;
+constexpr char kGroupGradsAndParams[] = "group_grads_params";
+
+constexpr char kFusedVarNamePrefix[] = "@FUSEDVAR@";
+
 }  // namespace details
 }  // namespace framework
 }  // namespace paddle

paddle/fluid/framework/details/op_registry.h

@@ -16,6 +16,8 @@ limitations under the License. */
 
 #include <string>
 #include <tuple>
+#include <unordered_map>
+#include <unordered_set>
 #include <vector>
 #include "paddle/fluid/framework/grad_op_desc_maker.h"
 #include "paddle/fluid/framework/inplace_op_inference.h"
@@ -127,9 +129,9 @@ struct OpInfoFiller<T, kGradOpDescMaker> {
 template <typename T>
 struct OpInfoFiller<T, kVarTypeInference> {
   void operator()(const char* op_type, OpInfo* info) const {
-    info->infer_var_type_ = [](const OpDesc& fwd_op, BlockDesc* block) {
+    info->infer_var_type_ = [](InferVarTypeContext* context) {
       T inference;
-      inference(fwd_op, block);
+      inference(context);
     };
   }
 };

paddle/fluid/framework/details/reduce_and_gather.h

@@ -53,6 +53,31 @@ struct ReduceLoDTensor {
   }
 };
 
+struct ReduceBufferData {
+  const std::vector<const void *> &src_data_;
+  void *dst_data_;
+  int64_t numel_;
+
+  ReduceBufferData(const std::vector<const void *> &src, void *dst,
+                   int64_t numel)
+      : src_data_(src), dst_data_(dst), numel_(numel) {}
+
+  template <typename T>
+  void apply() const {
+    T *dst_data = reinterpret_cast<T *>(dst_data_);
+    for (size_t i = 0; i < src_data_.size(); ++i) {
+      auto srd_data = reinterpret_cast<const T *>(src_data_[i]);
+      VLOG(10) << "dst: " << dst_data_ << ", " << srd_data;
+      if (srd_data == dst_data_) {
+        continue;
+      }
+
+      std::transform(srd_data, srd_data + numel_, dst_data, dst_data,
+                     [](T a, T b) -> T { return a + b; });
+    }
+  }
+};
+
 inline void GatherLocalSelectedRows(
     const std::vector<const SelectedRows *> &src_selecte_rows_,
     const std::vector<platform::Place> &in_places,

paddle/fluid/framework/ir/CMakeLists.txt

@@ -46,6 +46,7 @@ cc_library(fuse_pass_base SRCS fuse_pass_base.cc DEPS pass)
 pass_library(graph_to_program_pass base)
 pass_library(graph_viz_pass base)
 pass_library(lock_free_optimize_pass base)
+pass_library(cpu_quantize_placement_pass base)
 pass_library(cpu_quantize_pass inference)
 pass_library(cpu_quantize_squash_pass inference)
 pass_library(fc_fuse_pass inference)
@@ -103,6 +104,7 @@ cc_test(test_graph_pattern_detector SRCS graph_pattern_detector_tester.cc DEPS g
 cc_test(test_fc_fuse_pass SRCS fc_fuse_pass_tester.cc DEPS fc_fuse_pass framework_proto)
 cc_test(test_seqpool_concat_fuse_pass SRCS seqpool_concat_fuse_pass_tester.cc DEPS seqpool_concat_fuse_pass framework_proto)
 cc_test(test_is_test_pass SRCS is_test_pass_tester.cc DEPS is_test_pass)
+cc_test(test_cpu_quantize_placement_pass SRCS cpu_quantize_placement_pass_tester.cc DEPS cpu_quantize_placement_pass)
 cc_test(test_cpu_quantize_pass SRCS cpu_quantize_pass_tester.cc DEPS cpu_quantize_pass naive_executor)
 cc_test(test_cpu_quantize_squash_pass SRCS cpu_quantize_squash_pass_tester.cc DEPS cpu_quantize_squash_pass naive_executor)
 if(NOT WIN32)

paddle/fluid/framework/ir/cpu_quantize_placement_pass.cc

+/* Copyright (c) 2019 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/ir/cpu_quantize_placement_pass.h"
+#include <string>
+#include <unordered_set>
+
+namespace paddle {
+namespace framework {
+namespace ir {
+
+std::unique_ptr<ir::Graph> CPUQuantizePlacementPass::ApplyImpl(
+    std::unique_ptr<ir::Graph> graph) const {
+  VLOG(3) << "Marks operators which are to be quantized.";
+  const auto& excluded_ids_list =
+      Get<std::unordered_set<int>>("quantize_excluded_op_ids");
+  const auto& op_types_list =
+      Get<std::unordered_set<std::string>>("quantize_enabled_op_types");
+  for (const Node* n : graph->Nodes()) {
+    if (n->IsOp()) {
+      if (std::find(excluded_ids_list.begin(), excluded_ids_list.end(),
+                    n->id()) != excluded_ids_list.end())
+        continue;
+      auto* op = n->Op();
+      if (op->HasAttr("use_quantizer") || op->HasProtoAttr("use_quantizer")) {
+        if (op_types_list.empty()) {
+          op->SetAttr("use_quantizer", true);
+        } else if (std::find(op_types_list.begin(), op_types_list.end(),
+                             n->Name()) != op_types_list.end()) {
+          op->SetAttr("use_quantizer", true);
+        }
+      }
+    }
+  }
+  return graph;
+}
+
+}  // namespace ir
+}  // namespace framework
+}  // namespace paddle
+
+REGISTER_PASS(cpu_quantize_placement_pass,
+              paddle::framework::ir::CPUQuantizePlacementPass)
+    // a vector of operator type names to be quantized ("conv2d" etc.)
+    .RequirePassAttr("quantize_enabled_op_types")
+    // a vector of operator ids that are to be excluded from quantization
+    .RequirePassAttr("quantize_excluded_op_ids");

paddle/fluid/framework/ir/cpu_quantize_placement_pass.h

+/* Copyright (c) 2019 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 <memory>
+#include "paddle/fluid/framework/ir/pass.h"
+
+namespace paddle {
+namespace framework {
+namespace ir {
+/*
+ * Specifies which operators should be quantized.
+ */
+class CPUQuantizePlacementPass : public Pass {
+ protected:
+  std::unique_ptr<ir::Graph> ApplyImpl(
+      std::unique_ptr<ir::Graph> graph) const override;
+};
+
+}  // namespace ir
+}  // namespace framework
+}  // namespace paddle

paddle/fluid/framework/ir/cpu_quantize_placement_pass_tester.cc

+// Copyright (c) 2019 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/ir/cpu_quantize_placement_pass.h"
+
+#include <gtest/gtest.h>
+#include <boost/logic/tribool.hpp>
+
+namespace paddle {
+namespace framework {
+namespace ir {
+
+void SetOp(ProgramDesc* prog, const std::string& type, const std::string& name,
+           const std::vector<std::string>& inputs,
+           const std::vector<std::string>& outputs,
+           boost::tribool use_quantizer) {
+  auto* op = prog->MutableBlock(0)->AppendOp();
+
+  op->SetType(type);
+
+  if (!boost::indeterminate(use_quantizer))
+    op->SetAttr("use_quantizer", use_quantizer);
+
+  if (type == "conv2d") {
+    op->SetAttr("name", name);
+    op->SetInput("Input", {inputs[0]});
+    op->SetInput("Filter", {inputs[1]});
+    op->SetInput("Bias", {inputs[2]});
+  } else if (type == "relu") {
+    op->SetInput("X", inputs);
+  } else if (type == "concat") {
+    op->SetAttr("axis", 1);
+    op->SetInput("X", {inputs[0], inputs[1]});
+  } else if (type == "pool2d") {
+    op->SetInput("X", {inputs[0]});
+  } else {
+    FAIL() << "Unexpected operator type.";
+  }
+  op->SetOutput("Out", {outputs[0]});
+}
+
+// operator                      use_quantizer
+// ---------------------------------------
+// (a,b)->concat->c              none
+// (c,weights,bias)->conv->f     false
+// f->relu->g                    none
+// g->pool->h                    false
+// (h,weights2,bias2)->conv->k   false
+// k->pool->l                    false
+ProgramDesc BuildProgramDesc() {
+  ProgramDesc prog;
+
+  for (auto& v :
+       std::vector<std::string>({"a", "b", "c", "weights", "bias", "f", "g",
+                                 "h", "weights2", "bias2", "k", "l"})) {
+    auto* var = prog.MutableBlock(0)->Var(v);
+    var->SetType(proto::VarType::SELECTED_ROWS);
+    if (v == "weights" || v == "bias") {
+      var->SetPersistable(true);
+    }
+  }
+
+  SetOp(&prog, "concat", "concat1", {"a", "b"}, {"c"}, boost::indeterminate);
+  SetOp(&prog, "conv2d", "conv1", {"c", "weights", "bias"}, {"f"}, false);
+  SetOp(&prog, "relu", "relu1", {"f"}, {"g"}, boost::indeterminate);
+  SetOp(&prog, "pool2d", "pool1", {"g"}, {"h"}, false);
+  SetOp(&prog, "conv2d", "conv2", {"h", "weights2", "bias2"}, {"k"}, false);
+  SetOp(&prog, "pool2d", "pool2", {"k"}, {"l"}, false);
+
+  return prog;
+}
+
+void MainTest(std::initializer_list<std::string> quantize_enabled_op_types,
+              std::initializer_list<int> quantize_excluded_op_ids,
+              unsigned expected_use_quantizer_true_count) {
+  auto prog = BuildProgramDesc();
+
+  std::unique_ptr<ir::Graph> graph(new ir::Graph(prog));
+
+  auto pass = PassRegistry::Instance().Get("cpu_quantize_placement_pass");
+  pass->Set("quantize_enabled_op_types",
+            new std::unordered_set<std::string>(quantize_enabled_op_types));
+  pass->Set("quantize_excluded_op_ids",
+            new std::unordered_set<int>(quantize_excluded_op_ids));
+
+  graph = pass->Apply(std::move(graph));
+
+  unsigned use_quantizer_true_count = 0;
+
+  for (auto* node : graph->Nodes()) {
+    if (node->IsOp()) {
+      auto* op = node->Op();
+      if (op->HasAttr("use_quantizer") &&
+          boost::get<bool>(op->GetAttr("use_quantizer"))) {
+        ++use_quantizer_true_count;
+      }
+    }
+  }
+
+  EXPECT_EQ(use_quantizer_true_count, expected_use_quantizer_true_count);
+}
+
+TEST(QuantizerPlacementPass, enabled_pool) { MainTest({"pool2d"}, {}, 2); }
+
+TEST(QuantizerPlacementPass, enabled_conv_excluded_one) {
+  MainTest({"conv2d"}, {4}, 1);
+}
+
+TEST(QuantizerPlacementPass, excluded_none) {
+  // 2 conv + 2 pool
+  MainTest({}, {}, 4);
+}
+
+}  // namespace ir
+}  // namespace framework
+}  // namespace paddle
+
+USE_PASS(cpu_quantize_placement_pass);

paddle/fluid/framework/ir/graph_test.cc

@@ -43,20 +43,20 @@ class SumOpMaker : public OpProtoAndCheckerMaker {
 
 class SumOpVarTypeInference : public VarTypeInference {
  public:
-  void operator()(const OpDesc &op_desc, BlockDesc *block) const override {
-    auto &inputs = op_desc.Input("X");
+  void operator()(InferVarTypeContext *ctx) const override {
+    auto &inputs = ctx->Input("X");
     auto default_var_type = proto::VarType::SELECTED_ROWS;
 
     bool any_input_is_lod_tensor = std::any_of(
-        inputs.begin(), inputs.end(), [block](const std::string &name) {
-          return block->Var(name)->GetType() == proto::VarType::LOD_TENSOR;
+        inputs.begin(), inputs.end(), [&ctx](const std::string &name) {
+          return ctx->GetType(name) == proto::VarType::LOD_TENSOR;
         });
     if (any_input_is_lod_tensor) {
       default_var_type = proto::VarType::LOD_TENSOR;
     }
 
-    auto out_var_name = op_desc.Output("Out").front();
-    block->Var(out_var_name)->SetType(default_var_type);
+    auto out_var_name = ctx->Output("Out").front();
+    ctx->SetType(out_var_name, default_var_type);
   }
 };
 
@@ -71,7 +71,7 @@ class DummyOpMaker : public OpProtoAndCheckerMaker {
 
 class DummyOpVarTypeInference : public VarTypeInference {
  public:
-  void operator()(const OpDesc &op_desc, BlockDesc *block) const override {}
+  void operator()(framework::InferVarTypeContext *ctx) const override {}
 };
 }  // namespace framework
 }  // namespace paddle

paddle/fluid/framework/op_desc.cc

@@ -24,6 +24,7 @@ limitations under the License. */
 #include "paddle/fluid/framework/operator.h"
 #include "paddle/fluid/framework/program_desc.h"
 #include "paddle/fluid/framework/shape_inference.h"
+#include "paddle/fluid/framework/var_type_inference.h"
 
 namespace paddle {
 namespace framework {
@@ -677,7 +678,8 @@ void OpDesc::InferVarType(BlockDesc *block) const {
   // var type inference. Hence, we don't do any "default" setting here.
   auto &info = OpInfoMap::Instance().Get(this->Type());
   if (info.infer_var_type_) {
-    info.infer_var_type_(*this, block);
+    InferVarTypeContext context(this, block);
+    info.infer_var_type_(&context);
   }
 }
 

paddle/fluid/framework/parallel_executor.cc

@@ -254,18 +254,29 @@ ParallelExecutor::ParallelExecutor(const std::vector<platform::Place> &places,
         member_->places_, nccl_id, build_strategy.num_trainers_,
         build_strategy.trainer_id_));
 
-    std::unique_ptr<platform::NCCLContextMap> dev_nccl_ctxs;
-    dev_nccl_ctxs.reset(new platform::NCCLContextMap(member_->places_));
-    // Initialize device context's nccl comm
-    // Note, more than one ParallelExecutor with same place, the nccl comm will
+    // Initialize device context's nccl comm, will be used by normal
+    // Operators like sync_batch_norm, and collective ops.
+    // NOTE: more than one ParallelExecutor with same place, the nccl comm will
     // be rewrite and there will be some problem.
+    // NOTE: NCCL group-calls and non-group-calls can not use the same
+    // NCCL communicator, so for ParallelGraph and Multi-Process mode, re-use
+    // same communicators.
+    std::unique_ptr<platform::NCCLContextMap> dev_nccl_ctxs;
+    if (nccl_id == nullptr) {
+      dev_nccl_ctxs.reset(new platform::NCCLContextMap(member_->places_));
+    }
     for (size_t dev_id = 0; dev_id < member_->places_.size(); ++dev_id) {
-      auto &nccl_ctx = dev_nccl_ctxs->at(dev_id);
       platform::DeviceContextPool &pool =
           platform::DeviceContextPool::Instance();
       auto *dev_ctx = static_cast<platform::CUDADeviceContext *>(
           pool.Get(member_->places_[dev_id]));
-      dev_ctx->set_nccl_comm(nccl_ctx.comm());
+      if (nccl_id != nullptr) {
+        auto &nccl_ctx = member_->nccl_ctxs_->at(member_->places_[dev_id]);
+        dev_ctx->set_nccl_comm(nccl_ctx.comm());
+      } else {
+        auto &nccl_ctx = dev_nccl_ctxs->at(member_->places_[dev_id]);
+        dev_ctx->set_nccl_comm(nccl_ctx.comm());
+      }
     }
 #else
     PADDLE_THROW("Not compiled with CUDA");

paddle/fluid/framework/scope.cc

@@ -34,7 +34,7 @@ DEFINE_double(
     "Memory size threshold (GB) when the garbage collector clear tensors."
     "Disabled when this value is less than 0");
 
-DEFINE_bool(fast_eager_deletion_mode, false,
+DEFINE_bool(fast_eager_deletion_mode, true,
             "Fast eager deletion mode. If enabled, memory would release "
             "immediately without waiting GPU kernel ends.");
 

paddle/fluid/framework/tensor_util.cc

@@ -44,6 +44,11 @@ void TensorCopy(const Tensor& src, const platform::Place& dst_place,
               << dst_place;
       return;
     }
+#ifdef PADDLE_WITH_MKLDNN
+    if (src.layout() == DataLayout::kMKLDNN) {
+      dst->set_mkldnn_prim_desc(src.get_mkldnn_prim_desc());
+    }
+#endif
     memory::Copy(boost::get<platform::CPUPlace>(dst_place), dst_ptr,
                  boost::get<platform::CPUPlace>(src_place), src_ptr, size);
   }

paddle/fluid/framework/type_defs.h

@@ -27,6 +27,7 @@ namespace framework {
 class OperatorBase;
 class OpDesc;
 class InferShapeContext;
+class InferVarTypeContext;
 class BlockDesc;
 class Variable;
 
@@ -53,7 +54,7 @@ using GradOpMakerFN = std::function<std::vector<std::unique_ptr<OpDesc>>(
     const std::vector<BlockDesc*>& grad_block)>;
 
 using InferVarTypeFN =
-    std::function<void(const OpDesc& /*op_desc*/, BlockDesc* /*block*/)>;
+    std::function<void(framework::InferVarTypeContext* /*context*/)>;
 
 using InferShapeFN = std::function<void(InferShapeContext*)>;
 

paddle/fluid/framework/var_type_inference.h

@@ -14,6 +14,8 @@ limitations under the License. */
 
 #pragma once
 #include <string>
+#include <unordered_map>
+#include <vector>
 #include "paddle/fluid/framework/block_desc.h"
 #include "paddle/fluid/framework/op_desc.h"
 #include "paddle/fluid/framework/type_defs.h"
@@ -21,26 +23,123 @@ limitations under the License. */
 namespace paddle {
 namespace framework {
 
+class OpDesc;
+class BlockDesc;
+// default infer var type context
+class InferVarTypeContext {
+ public:
+  InferVarTypeContext(const OpDesc* op, BlockDesc* block)
+      : op_(op), block_(block) {}
+
+  virtual ~InferVarTypeContext() {}
+
+  virtual Attribute GetAttr(const std::string& name) const {
+    PADDLE_ENFORCE_NOT_NULL(op_);
+    return op_->GetAttr(name);
+  }
+
+  virtual bool HasVar(const std::string& name) const {
+    PADDLE_ENFORCE_NOT_NULL(block_);
+    return block_->FindVarRecursive(name) != nullptr;
+  }
+
+  virtual bool HasInput(const std::string& name) const {
+    PADDLE_ENFORCE_NOT_NULL(op_);
+    return op_->Inputs().count(name) > 0;
+  }
+
+  virtual bool HasOutput(const std::string& name) const {
+    PADDLE_ENFORCE_NOT_NULL(op_);
+    return op_->Outputs().count(name) > 0;
+  }
+
+  virtual const std::vector<std::string>& Input(const std::string& name) const {
+    PADDLE_ENFORCE_NOT_NULL(op_);
+    return op_->Input(name);
+  }
+
+  virtual const std::vector<std::string>& Output(
+      const std::string& name) const {
+    PADDLE_ENFORCE_NOT_NULL(op_);
+    return op_->Output(name);
+  }
+
+  virtual proto::VarType::Type GetType(const std::string& name) const {
+    PADDLE_ENFORCE_NOT_NULL(block_);
+    return block_->FindRecursiveOrCreateVar(name).GetType();
+  }
+
+  virtual void SetType(const std::string& name, proto::VarType::Type type) {
+    PADDLE_ENFORCE_NOT_NULL(block_);
+    block_->FindRecursiveOrCreateVar(name).SetType(type);
+  }
+
+  virtual proto::VarType::Type GetDataType(const std::string& name) const {
+    PADDLE_ENFORCE_NOT_NULL(block_);
+    return block_->FindRecursiveOrCreateVar(name).GetDataType();
+  }
+
+  virtual void SetDataType(const std::string& name, proto::VarType::Type type) {
+    PADDLE_ENFORCE_NOT_NULL(block_);
+    block_->FindRecursiveOrCreateVar(name).SetDataType(type);
+  }
+
+  virtual std::vector<proto::VarType::Type> GetDataTypes(
+      const std::string& name) const {
+    PADDLE_ENFORCE_NOT_NULL(block_);
+    return block_->FindRecursiveOrCreateVar(name).GetDataTypes();
+  }
+
+  virtual void SetDataTypes(
+      const std::string& name,
+      const std::vector<proto::VarType::Type>& multiple_data_type) {
+    PADDLE_ENFORCE_NOT_NULL(block_);
+    block_->FindRecursiveOrCreateVar(name).SetDataTypes(multiple_data_type);
+  }
+
+  virtual std::vector<int64_t> GetShape(const std::string& name) const {
+    PADDLE_ENFORCE_NOT_NULL(block_);
+    return block_->FindRecursiveOrCreateVar(name).GetShape();
+  }
+
+  virtual void SetShape(const std::string& name,
+                        const std::vector<int64_t>& dims) {
+    PADDLE_ENFORCE_NOT_NULL(block_);
+    block_->FindRecursiveOrCreateVar(name).SetShape(dims);
+  }
+
+  virtual int32_t GetLoDLevel(const std::string& name) const {
+    PADDLE_ENFORCE_NOT_NULL(block_);
+    return block_->FindRecursiveOrCreateVar(name).GetLoDLevel();
+  }
+
+  virtual void SetLoDLevel(const std::string& name, int32_t lod_level) {
+    PADDLE_ENFORCE_NOT_NULL(block_);
+    block_->FindRecursiveOrCreateVar(name).SetLoDLevel(lod_level);
+  }
+
+ protected:
+  const OpDesc* op_;
+  BlockDesc* block_;
+};
+
 class VarTypeInference {
  public:
   virtual ~VarTypeInference() {}
-  virtual void operator()(const OpDesc& op_desc, BlockDesc* block) const = 0;
+  virtual void operator()(InferVarTypeContext* context) const = 0;  // NOLINT
 };
 
 class PassInDtypeAndVarTypeToOutput : public framework::VarTypeInference {
  public:
-  void operator()(const framework::OpDesc& op_desc,
-                  framework::BlockDesc* block) const final {
+  void operator()(framework::InferVarTypeContext* ctx) const final {  // NOLINT
     auto in_out_var_names = this->GetInputOutputWithSameType();
 
     for (auto& i_o_n : in_out_var_names) {
-      auto& x_name = op_desc.Input(i_o_n.first).at(0);
-      auto& out_name = op_desc.Output(i_o_n.second).at(0);
+      auto& x_name = ctx->Input(i_o_n.first).at(0);
+      auto& out_name = ctx->Output(i_o_n.second).at(0);
 
-      auto& x = block->FindRecursiveOrCreateVar(x_name);
-      auto& out = block->FindRecursiveOrCreateVar(out_name);
-      out.SetType(x.GetType());
-      out.SetDataType(x.GetDataType());
+      ctx->SetType(out_name, ctx->GetType(x_name));
+      ctx->SetDataType(out_name, ctx->GetDataType(x_name));
     }
   }
 

paddle/fluid/framework/var_type_inference_test.cc

@@ -44,20 +44,20 @@ class SumOpMaker : public OpProtoAndCheckerMaker {
 
 class SumOpVarTypeInference : public VarTypeInference {
  public:
-  void operator()(const OpDesc &op_desc, BlockDesc *block) const override {
-    auto &inputs = op_desc.Input("X");
+  void operator()(framework::InferVarTypeContext *ctx) const override {
+    auto &inputs = ctx->Input("X");
     auto default_var_type = proto::VarType::SELECTED_ROWS;
 
     bool any_input_is_lod_tensor = std::any_of(
-        inputs.begin(), inputs.end(), [block](const std::string &name) {
-          return block->Var(name)->GetType() == proto::VarType::LOD_TENSOR;
+        inputs.begin(), inputs.end(), [&ctx](const std::string &name) {
+          return ctx->GetType(name) == proto::VarType::LOD_TENSOR;
         });
     if (any_input_is_lod_tensor) {
       default_var_type = proto::VarType::LOD_TENSOR;
     }
 
-    auto out_var_name = op_desc.Output("Out").front();
-    block->Var(out_var_name)->SetType(default_var_type);
+    auto out_var_name = ctx->Output("Out").front();
+    ctx->SetType(out_var_name, default_var_type);
   }
 };
 }  // namespace framework

paddle/fluid/imperative/layer.cc

@@ -218,7 +218,7 @@ std::map<std::string, std::vector<VarBase*>> OpBase::ApplyGrad() {
                  "%s has no backward implementation", Type());
 
   VLOG(3) << "apply op grad: " << Type();
-  std::vector<framework::VariableValueMap> tmp_grad_outputs;
+  std::vector<VarBasePtrMap> tmp_grad_outputs;
   if (backward_id_ > 0) {
     VLOG(3) << "py_layer_grad";
     tmp_grad_outputs.resize(1);
@@ -241,26 +241,62 @@ std::map<std::string, std::vector<VarBase*>> OpBase::ApplyGrad() {
         auto& outputs = tmp_grad_outputs[k][it.first];
         outputs.reserve(it.second.size());
         for (size_t i = 0; i < it.second.size(); ++i) {
+          VarBase* origin_grad_var_base = it.second[i];
+
           // Allocate a new variable
-          Variable* tmp_var = new framework::Variable();
-          tmp_var->GetMutable<framework::LoDTensor>();
-          outputs.emplace_back(tmp_var);
+          VarBase* tmp_grad_var_base = new VarBase(
+              string::Sprintf("%s@IGrad", origin_grad_var_base->Name()),
+              origin_grad_var_base->DataType(), origin_grad_var_base->Dims(),
+              place_, true, false);
+          outputs.emplace_back(tmp_grad_var_base);
         }
       }
 
-      // Run grad op
-      framework::RuntimeContext ctx(grad_input_vars_[k], tmp_grad_outputs[k]);
-
       // No need to do compile time infer shape here.
       // grad_op_desc_->InferShape(*block_);
       // grad_op_desc->InferVarType(block_);
 
       std::unique_ptr<framework::OperatorBase> opbase =
           framework::OpRegistry::CreateOp(*grad_op_desc);
+
+      auto& info = framework::OpInfoMap::Instance().Get(grad_op_desc->Type());
+      if (info.infer_var_type_) {
+        RuntimeInferVarTypeContext infer_var_type_ctx(
+            &grad_input_vars_[k], &tmp_grad_outputs[k], &attrs_);
+        info.infer_var_type_(&infer_var_type_ctx);
+      }
+
       framework::OperatorWithKernel* op_kernel =
           dynamic_cast<framework::OperatorWithKernel*>(opbase.get());
       PADDLE_ENFORCE_NOT_NULL(op_kernel, "only support op with kernel");
 
+      // Run grad op
+      framework::VariableValueMap grad_invars_map;
+      framework::VariableValueMap grad_outvars_map;
+
+      for (const auto& it : grad_input_vars_[k]) {
+        auto& grad_invars = grad_invars_map[it.first];
+        grad_invars.reserve(it.second.size());
+        for (const VarBase* grad_inp : it.second) {
+          PADDLE_ENFORCE_NOT_NULL(grad_inp->var_, "op %s input %s nullptr",
+                                  grad_op_desc->Type(), grad_inp->Name());
+
+          grad_invars.emplace_back(grad_inp->var_);
+        }
+      }
+
+      for (const auto& it : tmp_grad_outputs[k]) {
+        auto& grad_outvars = grad_outvars_map[it.first];
+        grad_outvars.reserve(it.second.size());
+        for (VarBase* grad_out : it.second) {
+          PADDLE_ENFORCE_NOT_NULL(grad_out->var_, "op %s output %s nullptr",
+                                  grad_op_desc->Type(), grad_out->Name());
+
+          grad_outvars.emplace_back(grad_out->var_);
+        }
+      }
+
+      framework::RuntimeContext ctx(grad_invars_map, grad_outvars_map);
       framework::Scope scope;
       PreparedOp p = PreparedOp::Prepare(ctx, *op_kernel, place_);
       p.op.RuntimeInferShape(scope, place_, ctx);
@@ -277,8 +313,8 @@ std::map<std::string, std::vector<VarBase*>> OpBase::ApplyGrad() {
       PADDLE_ENFORCE_EQ(outputs.size(), origin_outputs.size());
 
       for (size_t i = 0; i < outputs.size(); ++i) {
-        framework::Variable* grad = outputs[i];
-        framework::Variable* orig_grad = origin_outputs[i];
+        framework::Variable* grad = outputs[i]->var_;
+        framework::Variable* orig_grad = origin_outputs[i]->var_;
         AddTo(grad, orig_grad, place_);
         delete grad;
       }
@@ -326,28 +362,35 @@ void PyLayer::RegisterFunc(int func_id, const py::object& py_func) {
 
 int PyLayer::NumFuncs() { return py_funcs_.size(); }
 
-std::vector<Variable*> PyLayer::Apply(int func_id,
-                                      const std::vector<VarBase*>& inputs) {
-  std::vector<framework::Variable*> invars;
-  for (const VarBase* in : inputs) {
-    invars.push_back(in->var_);
-  }
+std::vector<framework::Variable*> PyLayer::Apply(
+    int func_id, const std::vector<VarBase*>& inputs) {
   PADDLE_ENFORCE(py_funcs_.find(func_id) != py_funcs_.end());
-  return CallPythonFunc(py_funcs_[func_id], invars);
+  return CallPythonFunc(py_funcs_[func_id], inputs);
 }
 
-std::vector<Variable*> PyLayer::ApplyGrad(
-    int func_id, const std::vector<framework::Variable*>& inputs) {
+std::vector<VarBase*> PyLayer::ApplyGrad(int func_id,
+                                         const std::vector<VarBase*>& inputs) {
   PADDLE_ENFORCE(py_funcs_.find(func_id) != py_funcs_.end());
-  return CallPythonFunc(py_funcs_[func_id], inputs);
+  auto rets = CallPythonFunc(py_funcs_[func_id], inputs);
+
+  std::vector<VarBase*> outs;
+  outs.reserve(rets.size());
+  for (size_t i = 0U; i != rets.size(); ++i) {
+    outs.emplace_back(new VarBase(
+        string::Sprintf("%s_out_%d", framework::GradVarName(PyLayer::kFwdOut),
+                        i),
+        rets[i], nullptr, true));
+  }
+
+  return outs;
 }
 
 std::vector<framework::Variable*> PyLayer::CallPythonFunc(
-    const py::object& callable, const std::vector<framework::Variable*>& ins) {
+    const py::object& callable, const std::vector<VarBase*>& ins) {
   py::gil_scoped_acquire guard;
   py::tuple in_args(ins.size());
   for (size_t i = 0; i < ins.size(); ++i) {
-    const framework::LoDTensor& t = ins[i]->Get<framework::LoDTensor>();
+    const framework::LoDTensor& t = ins[i]->var_->Get<framework::LoDTensor>();
     in_args[i] = t.IsInitialized() ? py::cast(t) : py::cast(nullptr);
   }
   VLOG(3) << "pyfunc in " << py::len(in_args);
@@ -357,6 +400,7 @@ std::vector<framework::Variable*> PyLayer::CallPythonFunc(
   auto ret_tuple = py::cast<py::tuple>(ret);
   size_t ret_num = py::len(ret_tuple);
   std::vector<framework::Variable*> outs;
+  outs.reserve(ret_num);
   VLOG(3) << "pyfunc out " << ret_num;
   for (size_t i = 0; i < ret_num; ++i) {
     try {
@@ -367,7 +411,7 @@ std::vector<framework::Variable*> PyLayer::CallPythonFunc(
       auto* tensor = var->GetMutable<framework::LoDTensor>();
       tensor->ShareDataWith(*py_out_tensor);
       tensor->set_lod(py_out_tensor->lod());
-      outs.push_back(var);
+      outs.emplace_back(var);
     } catch (py::cast_error&) {
       PADDLE_THROW("The %d-th output must be LoDTensor", i);
     }

paddle/fluid/imperative/layer.h

@@ -18,14 +18,16 @@
 #include "paddle/fluid/framework/python_headers.h"
 // clang-format on
 
-#include <map>     // NOLINT
-#include <string>  // NOLINT
-#include <vector>  // NOLINT
-#include <memory>  // NOLINT
+#include <map>            // NOLINT
+#include <string>         // NOLINT
+#include <vector>         // NOLINT
+#include <memory>         // NOLINT
+#include <unordered_map>  // NOLINT
 
 #include "paddle/fluid/framework/op_desc.h"
 #include "paddle/fluid/framework/operator.h"
 #include "paddle/fluid/framework/var_desc.h"
+#include "paddle/fluid/framework/var_type_inference.h"
 #include "paddle/fluid/platform/enforce.h"
 #include "paddle/fluid/platform/device_context.h"
 #include "paddle/fluid/operators/math/math_function.h"
@@ -135,13 +137,13 @@ class VarBase {
                 persistable) {}
 
  private:
+  // TODO(minqiyang): need support SelectedRows
   VarBase(const std::string& name, framework::proto::VarType::Type dtype,
           const framework::DDim& shape, const platform::Place& place,
           framework::Variable* var, VarBase* grad, bool stop_gradient,
           bool persistable)
       : name_(name),
-        dtype_(dtype),
-        place_(place),
+        type_(framework::proto::VarType::LOD_TENSOR),
         var_(var),
         grads_(grad),
         stop_gradient_(stop_gradient),
@@ -151,10 +153,12 @@ class VarBase {
         pre_op_out_idx_(-1) {
     if (!var_) {
       var_ = new framework::Variable();
-      auto tensor = var_->GetMutable<framework::LoDTensor>();
-      tensor->Resize(shape);
-      tensor->mutable_data(place_, dtype_);
     }
+    auto tensor = var_->GetMutable<framework::LoDTensor>();
+    tensor->Resize(shape);
+    tensor->mutable_data(place, dtype);
+    VLOG(10) << "create varbase: " << name_ << " type: " << dtype
+             << " place: " << place;
   }
 
  public:
@@ -184,7 +188,23 @@ class VarBase {
     }
   }
 
-  inline framework::proto::VarType::Type DType() const { return dtype_; }
+  inline framework::DDim Dims() const {
+    return var_->Get<framework::LoDTensor>().dims();
+  }
+
+  // data type. e.g.. FP32
+  inline void SetDataType(framework::proto::VarType::Type type) {
+    auto tensor = var_->GetMutable<framework::LoDTensor>();
+    tensor->mutable_data(tensor->place(), type);
+  }
+  inline framework::proto::VarType::Type DataType() const {
+    auto tensor = var_->Get<framework::LoDTensor>();
+    return tensor.type();
+  }
+
+  // tensor type. e.g.. LoDTensor
+  inline void SetType(framework::proto::VarType::Type type) { type_ = type; }
+  inline framework::proto::VarType::Type Type() const { return type_; }
 
   inline void SetStopGradient(bool stop_gradient) {
     stop_gradient_ = stop_gradient;
@@ -238,7 +258,7 @@ class VarBase {
   }
 
   std::string name_;
-  framework::proto::VarType::Type dtype_;
+  framework::proto::VarType::Type type_;
   platform::Place place_;
 
   framework::Variable* var_;
@@ -334,11 +354,13 @@ class PYBIND11_HIDDEN OpBase {
   std::map<std::string, std::vector<int>> pre_ops_out_idx_;
 
   // Inputs to a vector of bwd ops.
-  std::vector<framework::VariableValueMap> grad_input_vars_;
+  std::vector<VarBasePtrMap> grad_input_vars_;
   // Outputs to a vector of bwd ops.
-  std::vector<framework::VariableValueMap> grad_output_vars_;
+  std::vector<VarBasePtrMap> grad_output_vars_;
 
   std::vector<py::object> backward_hooks_;
+
+  framework::AttributeMap attrs_;
 };
 
 class Layer {
@@ -365,12 +387,131 @@ class PyLayer {
   static std::vector<framework::Variable*> Apply(
       int func_id, const std::vector<VarBase*>& inputs);
 
-  static std::vector<framework::Variable*> ApplyGrad(
-      int func_id, const std::vector<framework::Variable*>& inputs);
+  static std::vector<VarBase*> ApplyGrad(int func_id,
+                                         const std::vector<VarBase*>& inputs);
 
  private:
   static std::vector<framework::Variable*> CallPythonFunc(
-      const py::object& callable, const std::vector<framework::Variable*>& ins);
+      const py::object& callable, const std::vector<VarBase*>& ins);
+};
+
+// infer var type context for imperative mode
+class PYBIND11_HIDDEN RuntimeInferVarTypeContext
+    : public framework::InferVarTypeContext {
+ public:
+  RuntimeInferVarTypeContext(const imperative::VarBasePtrMap* inputs,
+                             imperative::VarBasePtrMap* outputs,
+                             const framework::AttributeMap* attrs_map)
+      : InferVarTypeContext(nullptr, nullptr),
+        inputs_(inputs),
+        outputs_(outputs),
+        attrs_(attrs_map),
+        input_names_(),
+        output_names_(),
+        var_set_() {
+    input_names_.reserve(inputs_->size());
+    for (auto& it : *inputs_) {
+      for (imperative::VarBase* var : it.second) {
+        input_names_[it.first].emplace_back(var->Name());
+        var_set_[var->Name()] = var;
+      }
+    }
+
+    output_names_.reserve(outputs_->size());
+    for (auto& it : *outputs_) {
+      for (imperative::VarBase* var : it.second) {
+        output_names_[it.first].emplace_back(var->Name());
+        var_set_[var->Name()] = var;
+      }
+    }
+  }
+
+  virtual ~RuntimeInferVarTypeContext() {}
+
+  framework::Attribute GetAttr(const std::string& name) const override {
+    PADDLE_ENFORCE_NOT_NULL(attrs_);
+    return attrs_->at(name);
+  }
+
+  bool HasVar(const std::string& name) const override {
+    return var_set_.count(name) > 0;
+  }
+
+  bool HasInput(const std::string& name) const override {
+    PADDLE_ENFORCE_NOT_NULL(inputs_);
+    return inputs_->count(name) > 0;
+  }
+
+  bool HasOutput(const std::string& name) const override {
+    PADDLE_ENFORCE_NOT_NULL(outputs_);
+    return outputs_->count(name) > 0;
+  }
+
+  const std::vector<std::string>& Input(
+      const std::string& name) const override {
+    return input_names_.at(name);
+  }
+
+  const std::vector<std::string>& Output(
+      const std::string& name) const override {
+    return output_names_.at(name);
+  }
+
+  framework::proto::VarType::Type GetType(
+      const std::string& name) const override {
+    return var_set_.at(name)->Type();
+  }
+
+  void SetType(const std::string& name,
+               framework::proto::VarType::Type type) override {
+    var_set_[name]->SetType(type);
+  }
+
+  framework::proto::VarType::Type GetDataType(
+      const std::string& name) const override {
+    return var_set_.at(name)->DataType();
+  }
+
+  void SetDataType(const std::string& name,
+                   framework::proto::VarType::Type type) override {
+    var_set_[name]->SetDataType(type);
+  }
+
+  std::vector<framework::proto::VarType::Type> GetDataTypes(
+      const std::string& name) const override {
+    PADDLE_THROW("GetDataTypes is not supported in runtime InferVarType");
+  }
+
+  void SetDataTypes(const std::string& name,
+                    const std::vector<framework::proto::VarType::Type>&
+                        multiple_data_type) override {
+    PADDLE_THROW("SetDataTypes is not supported in runtime InferVarType");
+  }
+
+  std::vector<int64_t> GetShape(const std::string& name) const override {
+    PADDLE_THROW("Do not handle Shape in runtime InferVarType");
+  }
+
+  void SetShape(const std::string& name,
+                const std::vector<int64_t>& dims) override {
+    PADDLE_THROW("Do not handle Shape in runtime InferVarType");
+  }
+
+  int32_t GetLoDLevel(const std::string& name) const override {
+    PADDLE_THROW("Do not handle LoDLevel in runtime InferVarType");
+  }
+
+  void SetLoDLevel(const std::string& name, int32_t lod_level) override {
+    PADDLE_THROW("Do not handle LoDLevel in runtime InferVarType");
+  }
+
+ private:
+  const imperative::VarBasePtrMap* inputs_;
+  imperative::VarBasePtrMap* outputs_;
+  const framework::AttributeMap* attrs_;
+  std::unordered_map<std::string, std::vector<std::string>> input_names_;
+  std::unordered_map<std::string, std::vector<std::string>> output_names_;
+  std::unordered_map<std::string, imperative::VarBase*> var_set_;
 };
 
 }  // namespace imperative

paddle/fluid/imperative/tracer.cc

@@ -19,6 +19,7 @@
 #include <unordered_map>
 #include <unordered_set>
 
+#include "paddle/fluid/framework/var_type_inference.h"
 #include "paddle/fluid/operators/math/math_function.h"
 #include "paddle/fluid/platform/device_context.h"
 #include "paddle/fluid/platform/enforce.h"
@@ -135,7 +136,7 @@ framework::VariableNameMap CreateOutputVarNameMap(
 Tracer::Tracer(framework::BlockDesc* root_block) : root_block_(root_block) {}
 
 std::set<std::string> Tracer::Trace(OpBase* op, const VarBasePtrMap& inputs,
-                                    const VarBasePtrMap& outputs,
+                                    VarBasePtrMap* outputs,
                                     framework::AttributeMap attrs_map,
                                     const platform::Place expected_place,
                                     const bool stop_gradient) {
@@ -163,7 +164,7 @@ std::set<std::string> Tracer::Trace(OpBase* op, const VarBasePtrMap& inputs,
     op->TrackPreOp(it.first, it.second);
   }
 
-  op->output_vars_ = outputs;
+  op->output_vars_ = *outputs;
   for (auto it : op->output_vars_) {
     auto& outvars = outvars_map[it.first];
     const std::vector<VarBase*>& outputs = it.second;
@@ -186,7 +187,7 @@ std::set<std::string> Tracer::Trace(OpBase* op, const VarBasePtrMap& inputs,
   framework::VariableNameMap invars_name_map =
       CreateInputVarNameMap(op, inputs);
   framework::VariableNameMap outvars_name_map =
-      CreateOutputVarNameMap(op, outputs);
+      CreateOutputVarNameMap(op, *outputs);
 
   auto& info = framework::OpInfoMap::Instance().Get(op->Type());
   if (info.Checker() != nullptr) {
@@ -197,6 +198,11 @@ std::set<std::string> Tracer::Trace(OpBase* op, const VarBasePtrMap& inputs,
       framework::OpRegistry::CreateOp(op->Type(), invars_name_map,
                                       outvars_name_map, attrs_map);
 
+  if (info.infer_var_type_) {
+    RuntimeInferVarTypeContext infer_var_type_ctx(&inputs, outputs, &attrs_map);
+    info.infer_var_type_(&infer_var_type_ctx);
+  }
+
   // TODO(minqiyang): Support infer var type in imperative mode
   // Run forward op
   VLOG(3) << "tracer running " << op->Type();
@@ -221,6 +227,7 @@ std::set<std::string> Tracer::Trace(OpBase* op, const VarBasePtrMap& inputs,
     VLOG(5) << "start construct backward op";
 
     // construct grad op descs
+    op->attrs_ = attrs_map;
     std::unique_ptr<framework::OpDesc> fwd_op_desc(new framework::OpDesc(
         op->Type(), invars_name_map, outvars_name_map, attrs_map));
     std::unique_ptr<std::unordered_map<std::string, std::string>> grad_to_var(
@@ -247,12 +254,12 @@ std::set<std::string> Tracer::Trace(OpBase* op, const VarBasePtrMap& inputs,
             auto fwd_var_it = current_vars_map.find(grad_invar);
             PADDLE_ENFORCE(fwd_var_it != current_vars_map.end());
             // Forward inputs or outputs.
-            grad_in_vars.emplace_back(fwd_var_it->second->var_);
+            grad_in_vars.emplace_back(fwd_var_it->second);
           } else {
             VarBase* var = current_vars_map[var_it->second];
             InitGrad(var, prepared_op.GetDeviceContext());
             // Douts.
-            grad_in_vars.emplace_back(var->grads_->var_);
+            grad_in_vars.emplace_back(var->grads_);
           }
 
           vars_saved_for_backward.insert(it.first);
@@ -269,7 +276,7 @@ std::set<std::string> Tracer::Trace(OpBase* op, const VarBasePtrMap& inputs,
                          op->Type());
           VarBase* var = current_vars_map[var_it->second];
           InitGrad(var, prepared_op.GetDeviceContext());
-          grad_out_vars.push_back(var->grads_->var_);
+          grad_out_vars.push_back(var->grads_);
         }
       }
     }
@@ -309,23 +316,23 @@ std::vector<VarBase*> Tracer::PyTrace(OpBase* op,
     auto& grad_output_vars =
         op->grad_output_vars_[0][framework::GradVarName(PyLayer::kFwdOut)];
 
-    for (const VarBase* inp : inputs) {
-      grad_input_vars.push_back(inp->var_);
+    for (VarBase* inp : inputs) {
+      grad_input_vars.push_back(inp);
     }
     for (VarBase* out : outputs) {
-      grad_input_vars.push_back(out->var_);
+      grad_input_vars.push_back(out);
     }
 
     // TODO(minqiyang): Add GPU support for PyLayer, only support CPU now
     platform::CPUPlace place;
     for (VarBase* out : outputs) {
       InitGrad(out, platform::DeviceContextPool::Instance().Get(place));
-      grad_input_vars.push_back(out->grads_->var_);
+      grad_input_vars.push_back(out->grads_);
     }
 
     for (VarBase* inp : inputs) {
       InitGrad(inp, platform::DeviceContextPool::Instance().Get(place));
-      grad_output_vars.push_back(inp->grads_->var_);
+      grad_output_vars.push_back(inp->grads_);
     }
   }
   return outputs;

paddle/fluid/imperative/tracer.h

@@ -48,7 +48,7 @@ class Tracer {
   virtual ~Tracer() {}
 
   std::set<std::string> Trace(OpBase* op, const VarBasePtrMap& inputs,
-                              const VarBasePtrMap& outputs,
+                              VarBasePtrMap* outputs,  // NOLINT
                               framework::AttributeMap attrs_map,
                               const platform::Place expected_place,
                               const bool stop_gradient = false);

paddle/fluid/imperative/type_defs.h

@@ -25,6 +25,7 @@ class VarBase;
 class OpBase;
 
 typedef std::map<std::string, std::vector<VarBase*>> VarBasePtrMap;
+typedef std::map<std::string, std::vector<const VarBase*>> ConstVarBasePtrMap;
 typedef std::map<std::string, std::vector<OpBase*>> OpBasePtrMap;
 
 }  // namespace imperative

paddle/fluid/inference/tests/api/CMakeLists.txt

@@ -110,7 +110,7 @@ set(TRANSFORMER_INSTALL_DIR "${INFERENCE_DEMO_INSTALL_DIR}/transformer")
 download_model_and_data(${TRANSFORMER_INSTALL_DIR} "temp%2Ftransformer_model.tar.gz" "temp%2Ftransformer_data.txt.tar.gz")
 inference_analysis_test(test_analyzer_transformer SRCS analyzer_transformer_tester.cc 
   EXTRA_DEPS ${INFERENCE_EXTRA_DEPS}
-  ARGS --infer_model=${TRANSFORMER_INSTALL_DIR}/model --infer_data=${TRANSFORMER_INSTALL_DIR}/data.txt --batch_size=8)
+  ARGS --infer_model=${TRANSFORMER_INSTALL_DIR}/model --infer_data=${TRANSFORMER_INSTALL_DIR}/data.txt --batch_size=8 SERIAL)
 
 # ocr
 set(OCR_INSTALL_DIR "${INFERENCE_DEMO_INSTALL_DIR}/ocr")

paddle/fluid/inference/tests/api/analyzer_transformer_tester.cc

@@ -183,10 +183,13 @@ void SetInput(std::vector<std::vector<PaddleTensor>> *inputs) {
 }
 
 // Easy for profiling independently.
-TEST(Analyzer_Transformer, profile) {
+void profile(bool use_mkldnn = false) {
   AnalysisConfig cfg;
   SetConfig(&cfg);
   std::vector<PaddleTensor> outputs;
+  if (use_mkldnn) {
+    cfg.EnableMKLDNN();
+  }
 
   std::vector<std::vector<PaddleTensor>> input_slots_all;
   SetInput(&input_slots_all);
@@ -194,6 +197,11 @@ TEST(Analyzer_Transformer, profile) {
                  input_slots_all, &outputs, FLAGS_num_threads);
 }
 
+TEST(Analyzer_Transformer, profile) { profile(); }
+#ifdef PADDLE_WITH_MKLDNN
+TEST(Analyzer_Transformer, profile_mkldnn) { profile(true); }
+#endif
+
 // Check the fuse status
 TEST(Analyzer_Transformer, fuse_statis) {
   AnalysisConfig cfg;
@@ -206,9 +214,12 @@ TEST(Analyzer_Transformer, fuse_statis) {
 }
 
 // Compare result of NativeConfig and AnalysisConfig
-TEST(Analyzer_Transformer, compare) {
+void compare(bool use_mkldnn = false) {
   AnalysisConfig cfg;
   SetConfig(&cfg);
+  if (use_mkldnn) {
+    cfg.EnableMKLDNN();
+  }
 
   std::vector<std::vector<PaddleTensor>> input_slots_all;
   SetInput(&input_slots_all);
@@ -216,5 +227,10 @@ TEST(Analyzer_Transformer, compare) {
       reinterpret_cast<const PaddlePredictor::Config *>(&cfg), input_slots_all);
 }
 
+TEST(Analyzer_Transformer, compare) { compare(); }
+#ifdef PADDLE_WITH_MKLDNN
+TEST(Analyzer_Transformer, compare_mkldnn) { compare(true /* use_mkldnn */); }
+#endif
+
 }  // namespace inference
 }  // namespace paddle

paddle/fluid/operators/activation_op.cc

@@ -188,6 +188,9 @@ $$out = x - \\frac{e^{x} - e^{-x}}{e^{x} + e^{-x}}$$
 UNUSED constexpr char SqrtDoc[] = R"DOC(
 Sqrt Activation Operator.
 
+Please make sure legal input, when input a negative value closed to zero,
+you should add a small epsilon(1e-12) to avoid negative number caused by numerical errors.
+
 $out = \sqrt{x}$
 
 )DOC";

paddle/fluid/operators/beam_search_decode_op.cc

@@ -178,10 +178,10 @@ Beam Search Decode Operator. This Operator constructs the full hypotheses for
 each source sentence by walking back along the LoDTensorArray Input(ids)
 whose lods can be used to restore the path in the beam search tree.
 
-The Output(SentenceIds) and Output(SentenceScores) separately contain the 
-generated id sequences and the corresponding scores. The shapes and lods of the 
-two LodTensor are same. The lod level is 2 and the two levels separately 
-indicate how many hypotheses each source sentence has and how many ids each 
+The Output(SentenceIds) and Output(SentenceScores) separately contain the
+generated id sequences and the corresponding scores. The shapes and lods of the
+two LodTensor are same. The lod level is 2 and the two levels separately
+indicate how many hypotheses each source sentence has and how many ids each
 hypothesis has.
 )DOC");
   }
@@ -203,15 +203,12 @@ class BeamSearchDecodeInferShape : public framework::InferShapeBase {
 
 class BeamSearchDecodeInferVarType : public framework::VarTypeInference {
  public:
-  void operator()(const framework::OpDesc& op_desc,
-                  framework::BlockDesc* block) const override {
-    for (auto& o : op_desc.Output("SentenceIds")) {
-      auto& sentence_ids = block->FindRecursiveOrCreateVar(o);
-      sentence_ids.SetType(framework::proto::VarType::LOD_TENSOR);
+  void operator()(framework::InferVarTypeContext* ctx) const override {
+    for (auto& o : ctx->Output("SentenceIds")) {
+      ctx->SetType(o, framework::proto::VarType::LOD_TENSOR);
     }
-    for (auto& o : op_desc.Output("SentenceScores")) {
-      auto& sentence_scores = block->FindRecursiveOrCreateVar(o);
-      sentence_scores.SetType(framework::proto::VarType::LOD_TENSOR);
+    for (auto& o : ctx->Output("SentenceScores")) {
+      ctx->SetType(o, framework::proto::VarType::LOD_TENSOR);
     }
   }
 };

paddle/fluid/operators/beam_search_op.cc

@@ -65,7 +65,7 @@ class BeamSearchOpMaker : public framework::OpProtoAndCheckerMaker {
         .SetDefault(true);
 
     AddComment(R"DOC(
-This operator does the search in beams for one time step. 
+This operator does the search in beams for one time step.
 Specifically, it selects the top-K candidate word ids of current step from
 Input(ids) according to their Input(scores) for all source sentences,
 where K is Attr(beam_size) and Input(ids), Input(scores) are predicted results
@@ -120,15 +120,12 @@ class BeamSearchOp : public framework::OperatorWithKernel {
 
 class BeamSearchInferVarType : public framework::VarTypeInference {
  public:
-  void operator()(const framework::OpDesc &op_desc,
-                  framework::BlockDesc *block) const override {
-    for (auto &o : op_desc.Output("selected_ids")) {
-      auto &selected_ids = block->FindRecursiveOrCreateVar(o);
-      selected_ids.SetType(framework::proto::VarType::LOD_TENSOR);
+  void operator()(framework::InferVarTypeContext *ctx) const override {
+    for (auto &o : ctx->Output("selected_ids")) {
+      ctx->SetType(o, framework::proto::VarType::LOD_TENSOR);
     }
-    for (auto &o : op_desc.Output("selected_scores")) {
-      auto &selected_scores = block->FindRecursiveOrCreateVar(o);
-      selected_scores.SetType(framework::proto::VarType::LOD_TENSOR);
+    for (auto &o : ctx->Output("selected_scores")) {
+      ctx->SetType(o, framework::proto::VarType::LOD_TENSOR);
     }
   }
 };

paddle/fluid/operators/concat_op.cc

@@ -50,9 +50,19 @@ class ConcatOp : public framework::OperatorWithKernel {
         if (j == axis) {
           out_dims[axis] += ins[i][j];
         } else {
-          PADDLE_ENFORCE_EQ(out_dims[j], ins[i][j],
-                            "Input tensors should have the same "
-                            "elements except the specify axis.");
+          if (ctx->IsRuntime()) {
+            // check all shape in run time
+            PADDLE_ENFORCE_EQ(out_dims[j], ins[i][j],
+                              "Input tensors should have the same "
+                              "elements except the specify axis.");
+          } else {
+            // not check -1 with other in compile time
+            if (out_dims[j] != -1 && ins[i][j] != -1) {
+              PADDLE_ENFORCE_EQ(out_dims[j], ins[i][j],
+                                "Input tensors should have the same "
+                                "elements except the specify axis.");
+            }
+          }
         }
       }
     }

paddle/fluid/operators/controlflow/get_places_op.cc

@@ -93,11 +93,9 @@ execution.
 
 class GetPlacesInferVarType : public framework::VarTypeInference {
  public:
-  void operator()(const framework::OpDesc &op_desc,
-                  framework::BlockDesc *block) const override {
-    for (auto &o_name : op_desc.Output("Out")) {
-      block->FindRecursiveOrCreateVar(o_name).SetType(
-          framework::proto::VarType::PLACE_LIST);
+  void operator()(framework::InferVarTypeContext *ctx) const override {
+    for (auto &o_name : ctx->Output("Out")) {
+      ctx->SetType(o_name, framework::proto::VarType::PLACE_LIST);
     }
   }
 };

paddle/fluid/operators/controlflow/tensor_array_read_write_op.cc

@@ -100,16 +100,13 @@ class WriteToArrayInferShape : public framework::InferShapeBase {
 
 class WriteToArrayInferVarType : public framework::VarTypeInference {
  public:
-  void operator()(const framework::OpDesc &op_desc,
-                  framework::BlockDesc *block) const override {
-    auto x_name = op_desc.Input("X")[0];
-    auto out_name = op_desc.Output("Out")[0];
+  void operator()(framework::InferVarTypeContext *ctx) const override {
+    auto x_name = ctx->Input("X")[0];
+    auto out_name = ctx->Output("Out")[0];
     VLOG(10) << "Set Variable " << out_name << " as LOD_TENSOR_ARRAY";
-    auto &out = block->FindRecursiveOrCreateVar(out_name);
-    out.SetType(framework::proto::VarType::LOD_TENSOR_ARRAY);
-    auto *x = block->FindVarRecursive(x_name);
-    if (x != nullptr) {
-      out.SetDataType(x->GetDataType());
+    ctx->SetType(out_name, framework::proto::VarType::LOD_TENSOR_ARRAY);
+    if (ctx->HasVar(x_name)) {
+      ctx->SetDataType(out_name, ctx->GetDataType(x_name));
     }
   }
 };

paddle/fluid/operators/controlflow/while_op.cc

@@ -365,19 +365,16 @@ class WhileGradOpDescMaker : public framework::SingleGradOpDescMaker {
 
 class WhileGradOpVarTypeInference : public framework::VarTypeInference {
  public:
-  void operator()(const framework::OpDesc &op_desc,
-                  framework::BlockDesc *block) const override {
-    auto p_names = op_desc.Input(kX);
-    auto pg_ig_names = op_desc.Output(framework::GradVarName(kX));
+  void operator()(framework::InferVarTypeContext *ctx) const override {
+    auto p_names = ctx->Input(kX);
+    auto pg_ig_names = ctx->Output(framework::GradVarName(kX));
 
     for (size_t i = 0; i < p_names.size(); ++i) {
-      auto &p_var = detail::Ref(block->FindVarRecursive(p_names[i]));
-      auto *g_var = block->FindVarRecursive(pg_ig_names[i]);
-      if (g_var != nullptr) {  // Gradient could be @EMPTY@
+      if (ctx->HasVar(pg_ig_names[i])) {
         VLOG(5) << "Setting " << pg_ig_names[i] << " following " << p_names[i]
-                << " type: " << p_var.GetType();
-        g_var->SetType(p_var.GetType());
-        g_var->SetDataType(p_var.GetDataType());
+                << " type: " << ctx->GetType(p_names[i]);
+        ctx->SetType(pg_ig_names[i], ctx->GetType(p_names[i]));
+        ctx->SetDataType(pg_ig_names[i], ctx->GetDataType(p_names[i]));
       }
     }
   }

paddle/fluid/operators/conv_transpose_op.cc

@@ -13,6 +13,7 @@ See the License for the specific language governing permissions and
 limitations under the License. */
 
 #include "paddle/fluid/operators/conv_transpose_op.h"
+#include <memory>
 #include <string>
 #include <vector>
 
@@ -344,6 +345,28 @@ framework::OpKernelType ConvTransposeOpGrad::GetExpectedKernelType(
                                  ctx.GetPlace(), layout_, library_);
 }
 
+class ConvTransposeGradOpDescMaker : public framework::SingleGradOpDescMaker {
+ public:
+  using framework::SingleGradOpDescMaker::SingleGradOpDescMaker;
+
+ protected:
+  std::unique_ptr<framework::OpDesc> Apply() const override {
+    std::unique_ptr<framework::OpDesc> op(new framework::OpDesc());
+    op->SetType(ForwardOp().Type() + "_grad");
+    op->SetInput("Input", Input("Input"));
+    op->SetInput("Filter", Input("Filter"));
+    op->SetOutput(framework::GradVarName("Input"), InputGrad("Input"));
+    op->SetOutput(framework::GradVarName("Filter"), InputGrad("Filter"));
+    if (ForwardOp().Inputs().count("Bias") > 0) {
+      op->SetInput("Bias", Input("Bias"));
+      op->SetOutput(framework::GradVarName("Bias"), InputGrad("Bias"));
+    }
+    op->SetInput(framework::GradVarName("Output"), OutputGrad("Output"));
+    op->SetAttrMap(Attrs());
+    return op;
+  }
+};
+
 }  // namespace operators
 }  // namespace paddle
 
@@ -352,7 +375,7 @@ namespace ops = paddle::operators;
 // conv2d_transpose
 REGISTER_OPERATOR(conv2d_transpose, ops::ConvTransposeOp,
                   ops::Conv2DTransposeOpMaker,
-                  paddle::framework::DefaultGradOpDescMaker<true>);
+                  ops::ConvTransposeGradOpDescMaker);
 REGISTER_OPERATOR(conv2d_transpose_grad, ops::ConvTransposeOpGrad);
 
 REGISTER_OP_CPU_KERNEL(
@@ -368,7 +391,7 @@ REGISTER_OP_CPU_KERNEL(
 // conv3d_transpose
 REGISTER_OPERATOR(conv3d_transpose, ops::ConvTransposeOp,
                   ops::Conv3DTransposeOpMaker,
-                  paddle::framework::DefaultGradOpDescMaker<true>);
+                  ops::ConvTransposeGradOpDescMaker);
 REGISTER_OPERATOR(conv3d_transpose_grad, ops::ConvTransposeOpGrad);
 
 REGISTER_OP_CPU_KERNEL(
@@ -384,7 +407,7 @@ REGISTER_OP_CPU_KERNEL(
 // depthwise conv2d_transpose
 REGISTER_OPERATOR(depthwise_conv2d_transpose, ops::ConvTransposeOp,
                   ops::Conv2DTransposeOpMaker,
-                  paddle::framework::DefaultGradOpDescMaker<true>);
+                  ops::ConvTransposeGradOpDescMaker);
 REGISTER_OPERATOR(depthwise_conv2d_transpose_grad, ops::ConvTransposeOpGrad);
 
 REGISTER_OP_CPU_KERNEL(

paddle/fluid/operators/cos_sim_op.cc

@@ -74,6 +74,9 @@ class CosSimOpMaker : public framework::OpProtoAndCheckerMaker {
               "Norm of the second input, reduced along the 1st "
               "dimension.")
         .AsIntermediate();
+    AddAttr<bool>(framework::kAllKernelsMustComputeRuntimeShape,
+                  "Skip calling InferShape() function in the runtime.")
+        .SetDefault(true);
 
     AddComment(R"DOC(
 **Cosine Similarity Operator**

paddle/fluid/operators/cos_sim_op.h

@@ -28,17 +28,21 @@ class CosSimKernel : public framework::OpKernel<T> {
  public:
   void Compute(const framework::ExecutionContext& context) const override {
     // get Tensor
-    auto* in_x = context.Input<Tensor>("X");
+    auto* in_x = context.Input<framework::LoDTensor>("X");
     auto* in_y = context.Input<Tensor>("Y");
-    auto* out_z = context.Output<Tensor>("Out");
+    auto* out_z = context.Output<framework::LoDTensor>("Out");
     auto* out_x_norm = context.Output<Tensor>("XNorm");
     auto* out_y_norm = context.Output<Tensor>("YNorm");
-    out_z->mutable_data<T>(context.GetPlace());
-    out_x_norm->mutable_data<T>(context.GetPlace());
-    out_y_norm->mutable_data<T>(context.GetPlace());
 
     int rows_x = in_x->dims()[0];
     int rows_y = in_y->dims()[0];
+    out_z->Resize({rows_x, 1});
+    out_x_norm->Resize({rows_x, 1});
+    out_y_norm->Resize({rows_y, 1});
+    out_z->mutable_data<T>(context.GetPlace());
+    out_x_norm->mutable_data<T>(context.GetPlace());
+    out_y_norm->mutable_data<T>(context.GetPlace());
+    out_z->set_lod(in_x->lod());
 
     int cols = framework::product(in_x->dims()) / rows_x;
 
@@ -81,6 +85,7 @@ class CosSimGradKernel : public framework::OpKernel<T> {
 
     if (rows_x == rows_y) {
       if (out_grad_x) {
+        out_grad_x->Resize(in_x->dims());
         math::CosSimGradFunctor<T> functor(
             in_x_norm->data<T>(), in_y_norm->data<T>(), in_x->data<T>(),
             in_y->data<T>(), in_z->data<T>(), in_grad_z->data<T>(),
@@ -91,6 +96,7 @@ class CosSimGradKernel : public framework::OpKernel<T> {
         for_range(functor);
       }
       if (out_grad_y) {
+        out_grad_y->Resize(in_y->dims());
         math::CosSimGradFunctor<T> functor(
             in_y_norm->data<T>(), in_x_norm->data<T>(), in_y->data<T>(),
             in_x->data<T>(), in_z->data<T>(), in_grad_z->data<T>(),
@@ -102,6 +108,7 @@ class CosSimGradKernel : public framework::OpKernel<T> {
       }
     } else {
       if (out_grad_x) {
+        out_grad_x->Resize(in_x->dims());
         math::CosSimDxFunctor<T> functor(
             in_x_norm->data<T>(), in_y_norm->data<T>(), in_x->data<T>(),
             in_y->data<T>(), in_z->data<T>(), in_grad_z->data<T>(),
@@ -112,6 +119,7 @@ class CosSimGradKernel : public framework::OpKernel<T> {
         for_range(functor);
       }
       if (out_grad_y) {
+        out_grad_y->Resize(in_y->dims());
         out_grad_y->mutable_data<T>(context.GetPlace());
         math::SetConstant<DeviceContext, T> set_zero;
         auto& dev_ctx = context.template device_context<DeviceContext>();

paddle/fluid/operators/distributed_ops/allreduce_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 <future>  // NOLINT
+#include <ostream>
+
+#include "paddle/fluid/framework/data_type.h"
+#include "paddle/fluid/framework/lod_tensor.h"
+#include "paddle/fluid/framework/op_registry.h"
+#ifdef PADDLE_WITH_CUDA
+#include "paddle/fluid/platform/nccl_helper.h"
+#endif
+
+namespace paddle {
+namespace operators {
+
+struct MutableDataFunctor {
+  MutableDataFunctor(void** data, framework::LoDTensor* tensor,
+                     const platform::Place& place)
+      : data_(data), tensor_(tensor), place_(place) {}
+
+  template <typename T>
+  void apply() {
+    *data_ = tensor_->mutable_data<T>(place_);
+  }
+
+  void** data_;
+  framework::LoDTensor* tensor_;
+  platform::Place place_;
+};
+
+class AllReduceOp : public framework::OperatorBase {
+  using OperatorBase::OperatorBase;
+
+  void RunImpl(const framework::Scope& scope,
+               const platform::Place& place) const override {
+    PADDLE_ENFORCE(is_gpu_place(place),
+                   "AllReduce op can run on gpu place only for now.");
+#ifdef PADDLE_WITH_CUDA
+    platform::DeviceContextPool& pool = platform::DeviceContextPool::Instance();
+    auto* ctx = pool.Get(place);
+    auto in_names = Inputs("X");
+    auto out_names = Outputs("Out");
+    PADDLE_ENFORCE_EQ(in_names.size(), 1, "Only support one input");
+    PADDLE_ENFORCE_EQ(out_names.size(), 1, "Only support one output");
+
+    auto* in = scope.FindVar(in_names[0]);
+    auto* out = scope.FindVar(out_names[0]);
+
+    PADDLE_ENFORCE(in->IsType<framework::LoDTensor>() ||
+                       out->IsType<framework::LoDTensor>(),
+                   "Only support allreduce LoDTensors");
+
+    int dtype = -1;
+    auto in_tensor = in->Get<framework::LoDTensor>();
+    dtype = platform::ToNCCLDataType(in_tensor.type());
+
+    int64_t numel = in_tensor.numel();
+    auto* sendbuff = in_tensor.data<void>();
+    auto* out_tensor = out->GetMutable<framework::LoDTensor>();
+    out_tensor->Resize(in_tensor.dims());
+    void* recvbuff = nullptr;
+    framework::VisitDataType(in_tensor.type(),
+                             MutableDataFunctor(&recvbuff, out_tensor, place));
+
+    auto cuda_ctx = static_cast<platform::CUDADeviceContext*>(ctx);
+    auto* comm = cuda_ctx->nccl_comm();
+    // FIXME(typhoonzero): should use nccl stream here.
+    auto stream = cuda_ctx->stream();
+
+    int reduce_type = Attr<int>("reduce_type");
+    ncclRedOp_t red_type = ncclSum;
+    switch (reduce_type) {
+      case 0:
+        red_type = ncclSum;
+        break;
+      case 1:
+        red_type = ncclProd;
+        break;
+      case 2:
+        red_type = ncclMax;
+        break;
+      case 3:
+        red_type = ncclMin;
+        break;
+    }
+
+    PADDLE_ENFORCE(platform::dynload::ncclAllReduce(
+        sendbuff, recvbuff, numel, static_cast<ncclDataType_t>(dtype), red_type,
+        comm, stream));
+#endif
+  }
+};
+
+class AllReduceOpMaker : public framework::OpProtoAndCheckerMaker {
+ public:
+  void Make() {
+    AddInput("X", "(Tensor), tensor to be allreduced.");
+    AddOutput("Out", "(Tensor) the result of allreduced.");
+    AddAttr<int>("reduce_type", "(int) determin the reduce type.")
+        .SetDefault(0);
+    AddComment(R"DOC(
+***AllReduce Operator***
+
+Call NCCL AllReduce internally. Note that this op must be used when one
+thread is managing one GPU device.
+
+For speed reasons, reduce_type should be an integer:
+
+0: sum
+1: prod
+2: max
+3: min
+
+If input and output are the same variable, in-place allreduce will be used.
+)DOC");
+  }
+};
+
+class AllReduceOpShapeInference : public framework::InferShapeBase {
+ public:
+  void operator()(framework::InferShapeContext* ctx) const override {}
+};
+
+}  // namespace operators
+}  // namespace paddle
+
+namespace ops = paddle::operators;
+
+REGISTER_OPERATOR(allreduce, ops::AllReduceOp,
+                  paddle::framework::EmptyGradOpMaker, ops::AllReduceOpMaker,
+                  ops::AllReduceOpShapeInference);

paddle/fluid/operators/distributed_ops/fake_init_op.cc

@@ -56,8 +56,7 @@ class FakeInitOp : public framework::OperatorBase {
 
 class FakeInitOpVarTypeInference : public framework::VarTypeInference {
  public:
-  void operator()(const framework::OpDesc &op_desc,
-                  framework::BlockDesc *block) const override {}
+  void operator()(framework::InferVarTypeContext *ctx) const override {}
 };
 
 class FakeInitOpMaker : public framework::OpProtoAndCheckerMaker {

paddle/fluid/operators/distributed_ops/merge_ids_op.cc

@@ -114,11 +114,10 @@ class MergeIdsOp : public framework::OperatorWithKernel {
 
 class MergeIdsOpInferVarType : public framework::VarTypeInference {
  public:
-  void operator()(const framework::OpDesc &op_desc,
-                  framework::BlockDesc *block) const override {
-    auto *input_var = block->Var(op_desc.Input("Ids")[0]);
-    for (auto &out_var : op_desc.Output("Out")) {
-      block->Var(out_var)->SetType(input_var->GetType());
+  void operator()(framework::InferVarTypeContext *ctx) const override {
+    auto input_type = ctx->GetType(ctx->Input("Ids")[0]);
+    for (auto &out_var : ctx->Output("Out")) {
+      ctx->SetType(out_var, input_type);
     }
   }
 };

paddle/fluid/operators/distributed_ops/split_ids_op.cc

@@ -14,6 +14,8 @@ limitations under the License. */
 
 #include "paddle/fluid/operators/distributed_ops/split_ids_op.h"
 
+#include <memory>
+
 namespace paddle {
 namespace operators {
 
@@ -71,11 +73,10 @@ class SplitIdsOp : public framework::OperatorWithKernel {
 
 class SplitIdsOpInferVarType : public framework::VarTypeInference {
  public:
-  void operator()(const framework::OpDesc &op_desc,
-                  framework::BlockDesc *block) const override {
-    auto *input_var = block->Var(op_desc.Input("Ids")[0]);
-    for (auto &out_var : op_desc.Output("Out")) {
-      block->Var(out_var)->SetType(input_var->GetType());
+  void operator()(framework::InferVarTypeContext *ctx) const override {
+    auto input_type = ctx->GetType(ctx->Input("Ids")[0]);
+    for (auto &out_var : ctx->Output("Out")) {
+      ctx->SetType(out_var, input_type);
     }
   }
 };

paddle/fluid/operators/dropout_op.cc

@@ -13,6 +13,7 @@ See the License for the specific language governing permissions and
 limitations under the License. */
 
 #include "paddle/fluid/operators/dropout_op.h"
+#include <memory>
 #include <string>
 
 namespace paddle {
@@ -106,21 +107,31 @@ class DropoutOpGrad : public framework::OperatorWithKernel {
     PADDLE_ENFORCE_EQ(ctx->Attrs().Get<bool>("is_test"), false,
                       "GradOp is only callable when is_test is false");
 
-    PADDLE_ENFORCE(ctx->HasInput("X"), "Input(X) must not be null.");
     PADDLE_ENFORCE(ctx->HasInput("Mask"), "Mask must not be null.");
     PADDLE_ENFORCE(ctx->HasInput(framework::GradVarName("Out")),
                    "Input(Out@GRAD) must not be null.");
 
-    auto x_dims = ctx->GetInputDim("X");
     auto out_dims = ctx->GetInputDim(framework::GradVarName("Out"));
-    PADDLE_ENFORCE_EQ(x_dims, out_dims,
-                      "Dimensions of Input(X) and Out@Grad must be the same.");
-    auto mask_dims = ctx->GetInputDim("Mask");
-    PADDLE_ENFORCE_EQ(x_dims, mask_dims,
-                      "Dimensions of Input(X) and Mask must be the same.");
-
-    ctx->SetOutputDim(framework::GradVarName("X"), x_dims);
-    ctx->ShareLoD("X", /*->*/ framework::GradVarName("X"));
+
+    ctx->SetOutputDim(framework::GradVarName("X"), out_dims);
+    ctx->ShareLoD(framework::GradVarName("Out"),
+                  /*->*/ framework::GradVarName("X"));
+  }
+};
+
+class DropoutGradOpDescMaker : public framework::SingleGradOpDescMaker {
+ public:
+  using framework::SingleGradOpDescMaker::SingleGradOpDescMaker;
+
+ protected:
+  std::unique_ptr<framework::OpDesc> Apply() const override {
+    std::unique_ptr<framework::OpDesc> op(new framework::OpDesc());
+    op->SetType("dropout_grad");
+    op->SetInput(framework::GradVarName("Out"), OutputGrad("Out"));
+    op->SetInput("Mask", Output("Mask"));
+    op->SetOutput(framework::GradVarName("X"), InputGrad("X"));
+    op->SetAttrMap(Attrs());
+    return op;
   }
 };
 
@@ -129,7 +140,7 @@ class DropoutOpGrad : public framework::OperatorWithKernel {
 
 namespace ops = paddle::operators;
 REGISTER_OPERATOR(dropout, ops::DropoutOp, ops::DropoutOpMaker,
-                  paddle::framework::DefaultGradOpDescMaker<true>);
+                  ops::DropoutGradOpDescMaker);
 REGISTER_OPERATOR(dropout_grad, ops::DropoutOpGrad);
 REGISTER_OP_CPU_KERNEL(
     dropout, ops::CPUDropoutKernel<paddle::platform::CPUDeviceContext, float>,

paddle/fluid/operators/fill_constant_op.cc

@@ -39,12 +39,11 @@ class FillConstantOp : public framework::OperatorWithKernel {
 
 class FillConstantOpVarTypeInference : public framework::VarTypeInference {
  public:
-  void operator()(const framework::OpDesc& op_desc,
-                  framework::BlockDesc* block) const override {
+  void operator()(framework::InferVarTypeContext* ctx) const override {
     auto data_type = static_cast<framework::proto::VarType::Type>(
-        boost::get<int>(op_desc.GetAttr("dtype")));
-    auto& out_var_name = op_desc.Output("Out").front();
-    block->Var(out_var_name)->SetDataType(data_type);
+        boost::get<int>(ctx->GetAttr("dtype")));
+    auto& out_var_name = ctx->Output("Out").front();
+    ctx->SetDataType(out_var_name, data_type);
   }
 };
 

paddle/fluid/operators/fused/fused_embedding_seq_pool_op.cc

@@ -138,22 +138,20 @@ class FusedEmbeddingSeqPoolOpGrad : public framework::OperatorWithKernel {
 class FusedEmbeddingSeqPoolOpGradVarTypeInference
     : public framework::VarTypeInference {
  public:
-  void operator()(const framework::OpDesc& op_desc,
-                  framework::BlockDesc* block) const override {
-    auto out_var_name = op_desc.Output(framework::GradVarName("W")).front();
-    auto attr = op_desc.GetAttr("is_sparse");
+  void operator()(framework::InferVarTypeContext* ctx) const override {
+    auto out_var_name = ctx->Output(framework::GradVarName("W")).front();
+    auto attr = ctx->GetAttr("is_sparse");
     bool is_sparse = boost::get<bool>(attr);
     if (is_sparse) {
       VLOG(3) << "fused_embedding_seq_pool_grad op "
               << framework::GradVarName("W") << " is set to SelectedRows";
-      block->Var(out_var_name)
-          ->SetType(framework::proto::VarType::SELECTED_ROWS);
+      ctx->SetType(out_var_name, framework::proto::VarType::SELECTED_ROWS);
     } else {
       VLOG(3) << "fused_embedding_seq_pool_grad op "
               << framework::GradVarName("W") << " is set to LoDTensor";
-      block->Var(out_var_name)->SetType(framework::proto::VarType::LOD_TENSOR);
+      ctx->SetType(out_var_name, framework::proto::VarType::LOD_TENSOR);
     }
-    block->Var(out_var_name)->SetDataType(block->Var("W")->GetDataType());
+    ctx->SetDataType(out_var_name, ctx->GetDataType(ctx->Input("W")[0]));
   }
 };
 

paddle/fluid/operators/get_tensor_from_selected_rows_op.cc

@@ -81,15 +81,12 @@ GetTensorFromSelectedRows is used to get the tensor from SelectedRows.
 class GetTensorFromSelectedRowsOpVarTypeInference
     : public framework::VarTypeInference {
  public:
-  void operator()(const framework::OpDesc &op_desc,
-                  framework::BlockDesc *block) const final {
-    auto out_var_name = op_desc.Output("Out").front();
-    auto in_var_name = op_desc.Input("X").front();
-
-    auto out_var = block->FindRecursiveOrCreateVar(out_var_name);
-    auto in_var = block->FindRecursiveOrCreateVar(in_var_name);
-    out_var.SetType(framework::proto::VarType::LOD_TENSOR);
-    out_var.SetDataType(in_var.GetDataType());
+  void operator()(framework::InferVarTypeContext *ctx) const {  // NOLINT
+    auto out_var_name = ctx->Output("Out").front();
+    auto in_var_name = ctx->Input("X").front();
+
+    ctx->SetType(out_var_name, framework::proto::VarType::LOD_TENSOR);
+    ctx->SetDataType(out_var_name, ctx->GetDataType(in_var_name));
   }
 };
 

paddle/fluid/operators/hierarchical_sigmoid_op.cc

@@ -197,38 +197,32 @@ class HierarchicalSigmoidGradOp : public framework::OperatorWithKernel {
 class HierarchicalSigmoidGradOpGradVarTypeInference
     : public framework::VarTypeInference {
  public:
-  void operator()(const framework::OpDesc& op_desc,
-                  framework::BlockDesc* block) const override {
-    auto w_grad_var_name = op_desc.Output(framework::GradVarName("W")).front();
-    auto bias_grad_var_name_vec =
-        op_desc.Output(framework::GradVarName("Bias"));
+  void operator()(framework::InferVarTypeContext* ctx) const override {
+    auto w_grad_var_name = ctx->Output(framework::GradVarName("W")).front();
+    auto bias_grad_var_name_vec = ctx->Output(framework::GradVarName("Bias"));
     std::string bias_grad_var_name;
     bool hasBias = false;
     if (bias_grad_var_name_vec.size()) {
       hasBias = true;
-      bias_grad_var_name =
-          op_desc.Output(framework::GradVarName("Bias")).front();
+      bias_grad_var_name = ctx->Output(framework::GradVarName("Bias")).front();
     }
-    auto attr = op_desc.GetAttr("is_sparse");
+    auto attr = ctx->GetAttr("is_sparse");
     bool is_sparse = boost::get<bool>(attr);
     if (is_sparse) {
       VLOG(30) << "hierarchical_sigmoid_grad op " << framework::GradVarName("W")
                << " is set to SelectedRows";
-      block->Var(w_grad_var_name)
-          ->SetType(framework::proto::VarType::SELECTED_ROWS);
+      ctx->SetType(w_grad_var_name, framework::proto::VarType::SELECTED_ROWS);
     } else {
       VLOG(30) << "hierarchical_sigmoid_grad op " << framework::GradVarName("W")
                << " is set to LoDTensor";
-      block->Var(w_grad_var_name)
-          ->SetType(framework::proto::VarType::LOD_TENSOR);
+      ctx->SetType(w_grad_var_name, framework::proto::VarType::LOD_TENSOR);
     }
     if (hasBias) {
       VLOG(30) << "hierarchical_sigmoid_grad op "
                << framework::GradVarName("Bias") << " is set to LoDTensor";
-      block->Var(bias_grad_var_name)
-          ->SetType(framework::proto::VarType::LOD_TENSOR);
+      ctx->SetType(bias_grad_var_name, framework::proto::VarType::LOD_TENSOR);
     }
-    block->Var(w_grad_var_name)->SetDataType(block->Var("W")->GetDataType());
+    ctx->SetDataType(w_grad_var_name, ctx->GetDataType(ctx->Input("W")[0]));
   }
 };
 

paddle/fluid/operators/layer_norm_op.cc

@@ -13,6 +13,7 @@ See the License for the specific language governing permissions and
 limitations under the License. */
 
 #include "paddle/fluid/operators/layer_norm_op.h"
+#include <memory>
 
 namespace paddle {
 namespace operators {
@@ -133,7 +134,7 @@ class LayerNormGradOp : public framework::OperatorWithKernel {
     }
     if (ctx->HasOutput(framework::GradVarName("Bias"))) {
       ctx->SetOutputDim(framework::GradVarName("Bias"),
-                        ctx->GetInputDim("Bias"));
+                        ctx->GetInputDim("Scale"));
     }
   }
 
@@ -157,12 +158,39 @@ class LayerNormGradOp : public framework::OperatorWithKernel {
   }
 };
 
+class LayerNormGradOpDescMaker : public framework::SingleGradOpDescMaker {
+ public:
+  using framework::SingleGradOpDescMaker::SingleGradOpDescMaker;
+
+ protected:
+  std::unique_ptr<framework::OpDesc> Apply() const override {
+    std::unique_ptr<framework::OpDesc> op(new framework::OpDesc());
+    op->SetType("layer_norm_grad");
+    op->SetInput("X", Input("X"));
+    op->SetInput("Mean", Output("Mean"));
+    op->SetInput("Variance", Output("Variance"));
+    if (ForwardOp().Inputs().count("Scale") > 0) {
+      op->SetInput("Scale", Input("Scale"));
+      op->SetOutput(framework::GradVarName("Scale"), InputGrad("Scale"));
+    }
+
+    if (ForwardOp().Inputs().count("Bias") > 0) {
+      op->SetOutput(framework::GradVarName("Bias"), InputGrad("Bias"));
+    }
+
+    op->SetInput(framework::GradVarName("Y"), OutputGrad("Y"));
+    op->SetOutput(framework::GradVarName("X"), InputGrad("X"));
+    op->SetAttrMap(Attrs());
+    return op;
+  }
+};
+
 }  // namespace operators
 }  // namespace paddle
 
 namespace ops = paddle::operators;
 REGISTER_OPERATOR(layer_norm, ops::LayerNormOp, ops::LayerNormOpMaker,
-                  paddle::framework::DefaultGradOpDescMaker<true>);
+                  ops::LayerNormGradOpDescMaker);
 REGISTER_OPERATOR(layer_norm_grad, ops::LayerNormGradOp);
 REGISTER_OP_CPU_KERNEL(
     layer_norm, ops::LayerNormKernel<paddle::platform::CPUDeviceContext, float>,

paddle/fluid/operators/layer_norm_op.h

@@ -245,11 +245,9 @@ class LayerNormGradKernel : public framework::OpKernel<T> {
   void Compute(const framework::ExecutionContext& ctx) const override {
     const float epsilon = ctx.Attr<float>("epsilon");
     auto x = *ctx.Input<Tensor>("X");
-    auto* y = ctx.Input<Tensor>("Y");
     auto* mean = ctx.Input<Tensor>("Mean");
     auto* var = ctx.Input<Tensor>("Variance");
     auto* scale = ctx.Input<Tensor>("Scale");
-    auto* bias = ctx.Input<Tensor>("Bias");
     auto d_y = *ctx.Input<Tensor>(framework::GradVarName("Y"));
     const auto begin_norm_axis = ctx.Attr<int>("begin_norm_axis");
 
@@ -275,18 +273,13 @@ class LayerNormGradKernel : public framework::OpKernel<T> {
       x.Resize(matrix_shape);
       temp.mutable_data<T>(matrix_shape, ctx.GetPlace());
 
-      if (!(bias && scale)) {
-        temp_norm.ShareDataWith(*y);
-        temp_norm.Resize(matrix_shape);
-      } else {
-        temp_norm.mutable_data<T>(matrix_shape, ctx.GetPlace());
-        // get x_norm
-        ElementwiseComputeEx<SubFunctor<T>, DeviceContext, T>(
-            ctx, &x, mean, /*axis*/ 0, SubFunctor<T>(), &temp_norm);
-        ElementwiseComputeEx<DivAndSqrtFunctor<T>, DeviceContext, T>(
-            ctx, &temp_norm, var, /*axis*/ 0,
-            DivAndSqrtFunctor<T>(static_cast<T>(epsilon)), &temp_norm);
-      }
+      temp_norm.mutable_data<T>(matrix_shape, ctx.GetPlace());
+      // get x_norm
+      ElementwiseComputeEx<SubFunctor<T>, DeviceContext, T>(
+          ctx, &x, mean, /*axis*/ 0, SubFunctor<T>(), &temp_norm);
+      ElementwiseComputeEx<DivAndSqrtFunctor<T>, DeviceContext, T>(
+          ctx, &temp_norm, var, /*axis*/ 0,
+          DivAndSqrtFunctor<T>(static_cast<T>(epsilon)), &temp_norm);
     }
 
     if (d_bias) {

paddle/fluid/operators/lod_rank_table_op.cc

@@ -64,11 +64,9 @@ class LoDRankTableInferShape : public framework::InferShapeBase {
 
 class LoDRankTableInferVarType : public framework::VarTypeInference {
  public:
-  void operator()(const framework::OpDesc &op_desc,
-                  framework::BlockDesc *block) const override {
-    for (auto &o : op_desc.Output("Out")) {
-      block->FindRecursiveOrCreateVar(o).SetType(
-          framework::proto::VarType::LOD_RANK_TABLE);
+  void operator()(framework::InferVarTypeContext *ctx) const override {
+    for (auto &o : ctx->Output("Out")) {
+      ctx->SetType(o, framework::proto::VarType::LOD_RANK_TABLE);
     }
   }
 };

paddle/fluid/operators/lod_tensor_to_array_op.cc

@@ -201,10 +201,9 @@ class LoDTensorToArrayInferShape : public framework::InferShapeBase {
 
 class LoDTensorToArrayInferVarType : public framework::VarTypeInference {
  public:
-  void operator()(const framework::OpDesc &op_desc,
-                  framework::BlockDesc *block) const override {
-    for (auto &out_var : op_desc.Output("Out")) {
-      block->Var(out_var)->SetType(framework::proto::VarType::LOD_TENSOR_ARRAY);
+  void operator()(framework::InferVarTypeContext *ctx) const override {
+    for (auto &out_var : ctx->Output("Out")) {
+      ctx->SetType(out_var, framework::proto::VarType::LOD_TENSOR_ARRAY);
     }
   }
 };

paddle/fluid/operators/lookup_table_op.cc

@@ -147,22 +147,20 @@ class LookupTableOpGrad : public framework::OperatorWithKernel {
 
 class LookupTableOpGradVarTypeInference : public framework::VarTypeInference {
  public:
-  void operator()(const framework::OpDesc& op_desc,
-                  framework::BlockDesc* block) const override {
-    auto out_var_name = op_desc.Output(framework::GradVarName("W")).front();
-    auto attr = op_desc.GetAttr("is_sparse");
+  void operator()(framework::InferVarTypeContext* ctx) const override {
+    auto out_var_name = ctx->Output(framework::GradVarName("W")).front();
+    auto attr = ctx->GetAttr("is_sparse");
     bool is_sparse = boost::get<bool>(attr);
     if (is_sparse) {
       VLOG(3) << "lookup_table_grad op " << framework::GradVarName("W")
               << " is set to SelectedRows";
-      block->Var(out_var_name)
-          ->SetType(framework::proto::VarType::SELECTED_ROWS);
+      ctx->SetType(out_var_name, framework::proto::VarType::SELECTED_ROWS);
     } else {
       VLOG(3) << "lookup_table_grad op " << framework::GradVarName("W")
               << " is set to LoDTensor";
-      block->Var(out_var_name)->SetType(framework::proto::VarType::LOD_TENSOR);
+      ctx->SetType(out_var_name, framework::proto::VarType::LOD_TENSOR);
     }
-    block->Var(out_var_name)->SetDataType(block->Var("W")->GetDataType());
+    ctx->SetDataType(out_var_name, ctx->GetDataType(ctx->Input("W")[0]));
   }
 };
 

paddle/fluid/operators/nccl/nccl_op.cc

@@ -60,12 +60,9 @@ class NCCLInitOp : public framework::OperatorBase {
 
 class NCCLInitOpVarTypeInference : public framework::VarTypeInference {
  public:
-  void operator()(const framework::OpDesc &op_desc,
-                  framework::BlockDesc *block) const override {
-    auto out_var_name = op_desc.Output("Communicator").front();
-    auto &out_var = block->FindRecursiveOrCreateVar(out_var_name);
-    auto var_type = framework::proto::VarType::RAW;
-    out_var.SetType(var_type);
+  void operator()(framework::InferVarTypeContext *ctx) const override {
+    auto out_var_name = ctx->Output("Communicator").front();
+    ctx->SetType(out_var_name, framework::proto::VarType::RAW);
   }
 };
 

paddle/fluid/operators/nce_op.cc

@@ -237,23 +237,21 @@ class NCEOpGrad : public framework::OperatorWithKernel {
 
 class NCEOpGradVarTypeInference : public framework::VarTypeInference {
  public:
-  void operator()(const framework::OpDesc &op_desc,
-                  framework::BlockDesc *block) const override {
-    auto weight_grad = op_desc.Output(framework::GradVarName("Weight")).front();
+  void operator()(framework::InferVarTypeContext *ctx) const override {
+    auto weight_grad = ctx->Output(framework::GradVarName("Weight")).front();
 
-    auto attr = op_desc.GetAttr("is_sparse");
+    auto attr = ctx->GetAttr("is_sparse");
     bool is_sparse = boost::get<bool>(attr);
     if (is_sparse) {
       VLOG(3) << "nce_op_grad op " << weight_grad << " and "
               << " is set to SelectedRows";
-      block->Var(weight_grad)
-          ->SetType(framework::proto::VarType::SELECTED_ROWS);
+      ctx->SetType(weight_grad, framework::proto::VarType::SELECTED_ROWS);
     } else {
       VLOG(3) << "nce_op_grad op " << weight_grad << " and "
               << " is set to LoDTensor";
-      block->Var(weight_grad)->SetType(framework::proto::VarType::LOD_TENSOR);
+      ctx->SetType(weight_grad, framework::proto::VarType::LOD_TENSOR);
     }
-    block->Var(weight_grad)->SetDataType(block->Var("Input")->GetDataType());
+    ctx->SetDataType(weight_grad, ctx->GetDataType(ctx->Input("Input")[0]));
   }
 };
 

paddle/fluid/operators/ngraph/ngraph_engine.cc

@@ -92,12 +92,10 @@ static std::vector<std::vector<int>> NgraphOpIntervals(
 
   int size = ops->size();
   int left = 0;
-  while (left < size && ops->at(left)->Type() != framework::kFeedOpType) {
+  while (left < size && ops->at(left)->Type() != framework::kFeedOpType &&
+         ops->at(left)->Type() != framework::kFetchOpType) {
     ++left;
   }
-  if (left == size) {
-    return intervals;
-  }
 
   while (left < size && ops->at(left)->Type() == framework::kFeedOpType) {
     for (auto& var_name_item : ops->at(left)->Outputs()) {
@@ -112,10 +110,6 @@ static std::vector<std::vector<int>> NgraphOpIntervals(
   while (right < size && ops->at(right)->Type() != framework::kFetchOpType) {
     ++right;
   }
-  if (right == size) {
-    return intervals;
-  }
-  if (left >= right) return intervals;
 
   int index = right;
   while (index < size && ops->at(index)->Type() == framework::kFetchOpType) {
@@ -127,6 +121,10 @@ static std::vector<std::vector<int>> NgraphOpIntervals(
     ++index;
   }
 
+  if (left == size || ops->at(left)->Type() == framework::kFetchOpType) {
+    left = 0;
+  }
+
   // (left, right - 1) represents indices between feed and fetch
   int pivot = left;
   while (pivot < right) {
@@ -234,6 +232,7 @@ NgraphEngine::NgraphEngine(const framework::Scope& scope,
 }
 
 void NgraphEngine::Prepare(const std::vector<int>& interval) {
+  bool has_fetch = false, is_full = false;
   for (auto& var : p_bdesc->AllVars()) {
     if (!(var->GetType() == framework::proto::VarType::SELECTED_ROWS ||
           var->GetType() == framework::proto::VarType::LOD_TENSOR ||
@@ -264,6 +263,9 @@ void NgraphEngine::Prepare(const std::vector<int>& interval) {
   std::vector<paddle::framework::OpDesc*> ops_desc;
   for (auto op_desc : p_bdesc->AllOps()) {
     ops_desc.emplace_back(op_desc);
+    if (op_desc->Type() == framework::kFetchOpType) {
+      has_fetch = true;
+    }
   }
 
   for (auto op_desc : ops_desc) {
@@ -276,11 +278,11 @@ void NgraphEngine::Prepare(const std::vector<int>& interval) {
   if (interval[0] > 0 &&
       ops_desc.at(interval[0] - 1)->Type() == framework::kFeedOpType &&
       interval[1] < static_cast<int>(ops_desc.size()) &&
-      ops_desc.at(interval.at(1))->Type() == framework::kFetchOpType) {
-    this->op_state_ = OpState::FULL;
+      ops_desc.at(interval[1])->Type() == framework::kFetchOpType) {
+    is_full = true;
   }
 
-  if (this->op_state_ == OpState::FULL) {
+  if (is_full) {
     this->op_state_ = this->is_test_ ? OpState::FULL_TEST : OpState::FULL_TRAIN;
   } else {
     this->op_state_ =
@@ -293,7 +295,8 @@ void NgraphEngine::Prepare(const std::vector<int>& interval) {
         framework::OpRegistry::CreateOp(*(ops_desc[idx])));
     ++idx;
   }
-  while (ops_desc.at(idx)->Type() != framework::kFetchOpType) {
+  while (idx < static_cast<int>(ops_desc.size()) &&
+         ops_desc.at(idx)->Type() != framework::kFetchOpType) {
     auto op_desc = ops_desc.at(idx);
     for (auto& var_name_item : op_desc->Inputs()) {
       for (auto& var_name : var_name_item.second) {
@@ -303,6 +306,10 @@ void NgraphEngine::Prepare(const std::vector<int>& interval) {
     ++idx;
   }
 
+  if (!has_fetch) {
+    op_state_ = OpState::UNKNOWN;
+  }
+
   BuildNgIO(ops_desc, interval);
 }
 
@@ -378,6 +385,7 @@ void NgraphEngine::BuildNgIO(const std::vector<framework::OpDesc*>& ops_desc,
       }
     }
   }
+
   for (size_t i = 0; i < var_in_.size(); ++i) {
     auto var_name = var_in_[i];
     if (persistables_.find(var_name) == persistables_.end()) {

paddle/fluid/operators/ngraph/ngraph_engine.h

@@ -12,8 +12,8 @@ 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. */
 
-#ifndef PADDLE_FLUID_OPERATORS_NGRAPH_NGRAPH_ENGINE_H_
-#define PADDLE_FLUID_OPERATORS_NGRAPH_NGRAPH_ENGINE_H_
+#pragma once
+
 #include <memory>
 #include <set>
 #include <string>
@@ -35,7 +35,6 @@ enum class OpState {                /* nGraph support state on ops          */
                      PARTIAL_TRAIN, /* Support partial ops for train        */
                      FULL_TEST,     /* Support full list of ops for test    */
                      PARTIAL_TEST,  /* Support partial list of ops for test */
-                     FULL,          /* All ops supported from feed to fetch */
                      UNKNOWN        /* Output all for debug purpose         */
 };
 
@@ -119,4 +118,3 @@ class NgraphEngine {
 
 }  // namespace operators
 }  // namespace paddle
-#endif  // PADDLE_FLUID_OPERATORS_NGRAPH_NGRAPH_ENGINE_H_

paddle/fluid/operators/ngraph/ngraph_engine_op.cc

@@ -37,8 +37,7 @@ class NgraphEngineOpMaker : public framework::OpProtoAndCheckerMaker {
 
 class NgraphEngineInferVarType : public framework::VarTypeInference {
  public:
-  void operator()(const framework::OpDesc &op_desc,
-                  framework::BlockDesc *block) const override {}
+  void operator()(framework::InferVarTypeContext *ctx) const override {}
 };
 
 }  // namespace operators

paddle/fluid/operators/optimizers/lars_momentum_op.cc

@@ -56,9 +56,9 @@ This optimizer use LARS (https://arxiv.org/abs/1708.03888) to optimize each
 weight using a local learning rate:
 
 $$
-local\_lr = \eta  * 
+local\_lr = \eta  *
     \frac{\left \| param \right \|}{\left \| grad \right \| + \beta *\left \| param \right \|} \\
-velocity = mu * velocity + 
+velocity = mu * velocity +
     local\_lr * (grad + \beta * param) \\
 param = param - velocity. \\
 $$
@@ -72,8 +72,7 @@ use L2 regularizers in case of using LARS.
 
 class LarsMomentumOpVarTypeInference : public framework::VarTypeInference {
  public:
-  void operator()(const framework::OpDesc &op_desc,
-                  framework::BlockDesc *block) const override {}
+  void operator()(framework::InferVarTypeContext* ctx) const override {}
 };
 }  // namespace operators
 }  // namespace paddle

paddle/fluid/operators/optimizers/momentum_op.cc

@@ -21,18 +21,14 @@ using Tensor = framework::Tensor;
 
 class MomentumOpInferVarType : public framework::VarTypeInference {
  public:
-  void operator()(const framework::OpDesc& op_desc,
-                  framework::BlockDesc* block) const override {
-    auto input_var = op_desc.Input("Param")[0];
-    for (auto& out_var : op_desc.Output("ParamOut")) {
-      if (block->FindRecursiveOrCreateVar(input_var).GetType() ==
-          framework::proto::VarType::SELECTED_ROWS) {
-        block->FindRecursiveOrCreateVar(out_var).SetType(
-            framework::proto::VarType::SELECTED_ROWS);
-      } else if (block->FindRecursiveOrCreateVar(input_var).GetType() ==
+  void operator()(framework::InferVarTypeContext* ctx) const override {
+    auto& input_var = ctx->Input("Param")[0];
+    for (auto& out_var : ctx->Output("ParamOut")) {
+      if (ctx->GetType(input_var) == framework::proto::VarType::SELECTED_ROWS) {
+        ctx->SetType(out_var, framework::proto::VarType::SELECTED_ROWS);
+      } else if (ctx->GetType(input_var) ==
                  framework::proto::VarType::LOD_TENSOR) {
-        block->FindRecursiveOrCreateVar(out_var).SetType(
-            framework::proto::VarType::LOD_TENSOR);
+        ctx->SetType(out_var, framework::proto::VarType::LOD_TENSOR);
       } else {
         PADDLE_THROW(
             "Only support LodTensor and SelectedRows, Unexpected Input Type.");

paddle/fluid/operators/optimizers/sgd_op.cc

@@ -50,20 +50,18 @@ class SGDOp : public framework::OperatorWithKernel {
 
 class SGDOpInferVarType : public framework::VarTypeInference {
  public:
-  void operator()(const framework::OpDesc &op_desc,
-                  framework::BlockDesc *block) const override {
-    auto input_var_n = op_desc.Input("Param")[0];
-    auto in_var_type = block->FindRecursiveOrCreateVar(input_var_n).GetType();
+  void operator()(framework::InferVarTypeContext *ctx) const override {
+    auto &input_var_n = ctx->Input("Param")[0];
+    auto in_var_type = ctx->GetType(input_var_n);
     PADDLE_ENFORCE(in_var_type == framework::proto::VarType::SELECTED_ROWS ||
                        in_var_type == framework::proto::VarType::LOD_TENSOR,
                    "The input Var's type should be LoDtensor or SelectedRows,"
                    " but the received var(%s)'s type is %s",
                    input_var_n, in_var_type);
 
-    for (auto &out_var_n : op_desc.Output("ParamOut")) {
-      auto &out_var = block->FindRecursiveOrCreateVar(out_var_n);
-      if (out_var.GetType() != in_var_type) {
-        out_var.SetType(in_var_type);
+    for (auto &out_var_n : ctx->Output("ParamOut")) {
+      if (ctx->GetType(out_var_n) != in_var_type) {
+        ctx->SetType(out_var_n, in_var_type);
       }
     }
   }

paddle/fluid/operators/py_func_op.cc

@@ -14,8 +14,11 @@
 
 #include "paddle/fluid/operators/py_func_op.h"
 
+#include <memory>
 #include <set>
 #include <string>
+#include <unordered_set>
+#include <utility>
 #include <vector>
 #include "paddle/fluid/framework/op_registry.h"
 
@@ -91,15 +94,12 @@ static void CallPythonFunc(py::object *callable,
   }
 }
 
-class PyFuncOpVarTypInference : public framework::VarTypeInference {
+class PyFuncOpVarTypeInference : public framework::VarTypeInference {
  public:
-  void operator()(const framework::OpDesc &op,
-                  framework::BlockDesc *block) const override {
-    auto &outs = op.Outputs();
-    bool has_out = (outs.count("Out") > 0 && !outs.at("Out").empty());
+  void operator()(framework::InferVarTypeContext *ctx) const override {
+    bool has_out = (ctx->HasOutput("Out") && !ctx->Output("Out").empty());
 
-    auto &ins = op.Inputs();
-    bool has_in = (ins.count("X") > 0 && !ins.at("X").empty());
+    bool has_in = (ctx->HasInput("X") && !ctx->Input("X").empty());
 
     /**
      * X or Out can be empty, so that py_func can be more flexible
@@ -107,8 +107,8 @@ class PyFuncOpVarTypInference : public framework::VarTypeInference {
      */
     PADDLE_ENFORCE(has_in || has_out, "Input(X) or Output(Out) must exist");
 
-    PADDLE_ENFORCE_GE(boost::get<int>(op.GetAttr(kForwardPythonCallableId)), 0,
-                      "Function id cannot be less than 0");
+    PADDLE_ENFORCE_GE(boost::get<int>(ctx->GetAttr(kForwardPythonCallableId)),
+                      0, "Function id cannot be less than 0");
 
     if (!has_out) return;
 
@@ -118,7 +118,7 @@ class PyFuncOpVarTypInference : public framework::VarTypeInference {
      * the corresponding forward variable
      */
     const std::string kGradVarSuffix = framework::kGradVarSuffix;
-    auto &out_var_names = outs.at("Out");
+    auto &out_var_names = ctx->Output("Out");
     for (auto &out_var_name : out_var_names) {
       if (out_var_name == framework::kEmptyVarName ||
           out_var_name.size() < kGradVarSuffix.size()) {
@@ -128,18 +128,17 @@ class PyFuncOpVarTypInference : public framework::VarTypeInference {
       size_t len = out_var_name.size() - kGradVarSuffix.size();
       if (out_var_name.substr(len) == kGradVarSuffix) {
         auto fwd_var_name = out_var_name.substr(0, len);
-        auto *out_var_desc = block->FindVarRecursive(out_var_name);
-        auto *fwd_var_desc = block->FindVarRecursive(fwd_var_name);
-        PADDLE_ENFORCE_NOT_NULL(out_var_desc, "Backward variable %s not found",
-                                out_var_name);
-        PADDLE_ENFORCE_NOT_NULL(fwd_var_desc, "Forward variable %s not found",
-                                fwd_var_name);
+        PADDLE_ENFORCE(ctx->HasVar(out_var_name),
+                       "Backward variable %s not found", out_var_name);
+        PADDLE_ENFORCE(ctx->HasVar(fwd_var_name),
+                       "Backward variable %s not found", fwd_var_name);
         VLOG(10) << "Infer var_desc of Output(" << out_var_name << ") as Input("
                  << fwd_var_name << ")";
-        out_var_desc->SetShape(fwd_var_desc->GetShape());
-        out_var_desc->SetDataType(fwd_var_desc->GetDataType());
-        out_var_desc->SetLoDLevel(fwd_var_desc->GetLoDLevel());
-        out_var_desc->SetType(fwd_var_desc->GetType());
+
+        ctx->SetShape(out_var_name, ctx->GetShape(fwd_var_name));
+        ctx->SetDataType(out_var_name, ctx->GetDataType(fwd_var_name));
+        ctx->SetLoDLevel(out_var_name, ctx->GetLoDLevel(fwd_var_name));
+        ctx->SetType(out_var_name, ctx->GetType(fwd_var_name));
       }
     }
   }
@@ -309,5 +308,5 @@ class PyFuncOp : public framework::OperatorBase {
 namespace ops = paddle::operators;
 
 REGISTER_OPERATOR(py_func, ops::PyFuncOp, ops::PyFuncOpMaker,
-                  ops::PyFuncOpVarTypInference, ops::PyFuncOpShapeInference,
+                  ops::PyFuncOpVarTypeInference, ops::PyFuncOpShapeInference,
                   ops::PyFuncOpGradDescMaker);

paddle/fluid/operators/reader/create_custom_reader_op.cc

@@ -85,10 +85,10 @@ class CreateCustomReaderOpMaker : public DecoratedReaderMakerBase {
     AddComment(R"DOC(
       CreateCustomReader Operator
 
-      A custom reader can be used for input data preprocessing. 
-      A custom reader holds its own sub-block, which will be executed in CPU 
-      in its 'ReadNext()' function. Users can configurate their own 
-      preprocessing pipelines by inserting operators into custom reader's 
+      A custom reader can be used for input data preprocessing.
+      A custom reader holds its own sub-block, which will be executed in CPU
+      in its 'ReadNext()' function. Users can configurate their own
+      preprocessing pipelines by inserting operators into custom reader's
       sub-block.
     )DOC");
   }
@@ -123,23 +123,22 @@ class CustomReaderInferShape : public framework::InferShapeBase {
 
 class CustomReaderInferVarType : public framework::VarTypeInference {
  public:
-  void operator()(const framework::OpDesc& op_desc,
-                  framework::BlockDesc* block) const override {
-    framework::VarDesc* out_reader = block->FindVar(op_desc.Output("Out")[0]);
-    PADDLE_ENFORCE_NOT_NULL(out_reader);
-    out_reader->SetType(framework::proto::VarType::READER);
+  void operator()(framework::InferVarTypeContext* ctx) const override {
+    auto& out_var_name = ctx->Output("Out")[0];
+    PADDLE_ENFORCE(ctx->HasVar(out_var_name));
+    ctx->SetType(out_var_name, framework::proto::VarType::READER);
 
     auto sink_var_names =
-        boost::get<std::vector<std::string>>(op_desc.GetAttr("sink_var_names"));
+        boost::get<std::vector<std::string>>(ctx->GetAttr("sink_var_names"));
     const auto* sub_block =
-        boost::get<framework::BlockDesc*>(op_desc.GetAttr("sub_block"));
+        boost::get<framework::BlockDesc*>(ctx->GetAttr("sub_block"));
     std::vector<framework::proto::VarType::Type> res_data_types;
     for (const std::string& var_name : sink_var_names) {
       framework::VarDesc* var = sub_block->FindVar(var_name);
       PADDLE_ENFORCE_NOT_NULL(var);
       res_data_types.emplace_back(var->GetDataType());
     }
-    out_reader->SetDataTypes(res_data_types);
+    ctx->SetDataTypes(out_var_name, res_data_types);
   }
 };
 

paddle/fluid/operators/reader/read_op.cc

@@ -51,19 +51,16 @@ class ReadInferShape : public framework::InferShapeBase {
 
 class ReadInferVarType : public framework::VarTypeInference {
  public:
-  void operator()(const framework::OpDesc& op_desc,
-                  framework::BlockDesc* block) const override {
-    bool infer_out = boost::get<bool>(op_desc.GetAttr("infer_out"));
+  void operator()(framework::InferVarTypeContext* ctx) const override {
+    bool infer_out = boost::get<bool>(ctx->GetAttr("infer_out"));
     if (infer_out) {
-      std::string reader_name = op_desc.Input("Reader")[0];
-      std::vector<std::string> out_names = op_desc.Output("Out");
-      framework::VarDesc* reader = block->FindVarRecursive(reader_name);
-      auto dtypes = reader->GetDataTypes();
+      std::string reader_name = ctx->Input("Reader")[0];
+      std::vector<std::string> out_names = ctx->Output("Out");
+      auto dtypes = ctx->GetDataTypes(reader_name);
       PADDLE_ENFORCE_EQ(dtypes.size(), out_names.size());
       for (size_t i = 0; i < dtypes.size(); ++i) {
-        framework::VarDesc& out = block->FindRecursiveOrCreateVar(out_names[i]);
-        out.SetType(framework::proto::VarType::LOD_TENSOR);
-        out.SetDataType(dtypes[i]);
+        ctx->SetType(out_names[i], framework::proto::VarType::LOD_TENSOR);
+        ctx->SetDataType(out_names[i], dtypes[i]);
       }
     }
   }

paddle/fluid/operators/reader/reader_op_registry.cc

@@ -98,11 +98,10 @@ void FileReaderInferShape::operator()(framework::InferShapeContext* ctx) const {
   }
 }
 
-void FileReaderInferVarType::operator()(const framework::OpDesc& op_desc,
-                                        framework::BlockDesc* block) const {
-  std::string reader_name = op_desc.Output("Out")[0];
-  framework::VarDesc* reader = block->FindVarRecursive(reader_name);
-  reader->SetType(framework::proto::VarType::READER);
+void FileReaderInferVarType::operator()(
+    framework::InferVarTypeContext* ctx) const {
+  std::string reader_name = ctx->Output("Out")[0];
+  ctx->SetType(reader_name, framework::proto::VarType::READER);
 }
 
 void DecoratedReaderInferShape::operator()(
@@ -125,13 +124,11 @@ void DecoratedReaderInferShape::operator()(
 }
 
 void DecoratedReaderInferVarType::operator()(
-    const framework::OpDesc& op_desc, framework::BlockDesc* block) const {
-  std::string in_reader_name = op_desc.Input("UnderlyingReader")[0];
-  framework::VarDesc* in_reader = block->FindVarRecursive(in_reader_name);
-  std::string out_reader_name = op_desc.Output("Out")[0];
-  framework::VarDesc* out_reader = block->FindVarRecursive(out_reader_name);
-  out_reader->SetType(framework::proto::VarType::READER);
-  out_reader->SetDataTypes(in_reader->GetDataTypes());
+    framework::InferVarTypeContext* ctx) const {
+  const std::string& in_reader_name = ctx->Input("UnderlyingReader")[0];
+  const std::string& out_reader_name = ctx->Output("Out")[0];
+  ctx->SetType(out_reader_name, framework::proto::VarType::READER);
+  ctx->SetDataTypes(out_reader_name, ctx->GetDataTypes(in_reader_name));
 }
 
 void DecoratedReaderMakerBase::Make() {

paddle/fluid/operators/reader/reader_op_registry.h

@@ -14,7 +14,9 @@
 
 #pragma once
 
+#include <memory>
 #include <string>
+#include <unordered_map>
 #include <vector>
 #include "paddle/fluid/framework/op_registry.h"
 #include "paddle/fluid/framework/reader.h"
@@ -59,8 +61,7 @@ class FileReaderInferShape : public framework::InferShapeBase {
 
 class FileReaderInferVarType : public framework::VarTypeInference {
  public:
-  void operator()(const framework::OpDesc& op_desc,
-                  framework::BlockDesc* block) const override;
+  void operator()(framework::InferVarTypeContext* ctx) const override;
 };
 
 // general infershape for decorated reader
@@ -72,8 +73,7 @@ class DecoratedReaderInferShape : public framework::InferShapeBase {
 // general var type inference for decorated reader
 class DecoratedReaderInferVarType : public framework::VarTypeInference {
  public:
-  void operator()(const framework::OpDesc& op_desc,
-                  framework::BlockDesc* block) const override;
+  void operator()(framework::InferVarTypeContext* ctx) const override;
 };
 
 class DecoratedReaderMakerBase : public framework::OpProtoAndCheckerMaker {

paddle/fluid/operators/save_op.cc

@@ -159,12 +159,9 @@ This operator will serialize and write LoDTensor / SelectedRows variable to file
 
 class SaveOpVarTypeInference : public framework::VarTypeInference {
  public:
-  void operator()(const framework::OpDesc &op_desc,
-                  framework::BlockDesc *block) const override {
-    auto out_var_name = op_desc.Output(LOOKUP_TABLE_PATH).front();
-    auto &out_var = block->FindRecursiveOrCreateVar(out_var_name);
-    auto var_type = framework::proto::VarType::RAW;
-    out_var.SetType(var_type);
+  void operator()(framework::InferVarTypeContext *ctx) const override {
+    auto out_var_name = ctx->Output(LOOKUP_TABLE_PATH).front();
+    ctx->SetType(out_var_name, framework::proto::VarType::RAW);
   }
 };
 

paddle/fluid/operators/scale_op.cc

@@ -14,6 +14,7 @@ limitations under the License. */
 
 #include "paddle/fluid/operators/scale_op.h"
 
+#include <memory>
 #include <string>
 
 #include "paddle/fluid/operators/detail/safe_ref.h"
@@ -69,17 +70,13 @@ $$Out = scale*(X + bias)$$
 
 class ScaleOpVarTypeInference : public framework::VarTypeInference {
  public:
-  void operator()(const framework::OpDesc &op_desc,
-                  framework::BlockDesc *block) const override {
-    auto &in_var_name = op_desc.Input("X").front();
-    auto &in_var = detail::Ref(block->FindVarRecursive(in_var_name));
-
-    auto out_var_name = op_desc.Output("Out").front();
-    auto *out_var = block->FindVarRecursive(out_var_name);
+  void operator()(framework::InferVarTypeContext *ctx) const override {
+    auto &in_var_name = ctx->Input("X").front();
+    auto out_var_name = ctx->Output("Out").front();
 
     if (in_var_name != out_var_name) {
-      out_var->SetType(in_var.GetType());
-      out_var->SetDataType(in_var.GetDataType());
+      ctx->SetType(out_var_name, ctx->GetType(in_var_name));
+      ctx->SetDataType(out_var_name, ctx->GetDataType(in_var_name));
     }
   }
 };

paddle/fluid/operators/sequence_ops/sequence_enumerate_op.cc

@@ -30,13 +30,6 @@ class SequenceEnumerateOp : public framework::OperatorWithKernel {
         "Output(X) of SequenceEnumerate operator should not be null.");
 
     const auto x_dims = ctx->GetInputDim("X");
-    PADDLE_ENFORCE_EQ(
-        x_dims.size(), 2,
-        "Input(X) of SequenceEnumerate operator's rank should be 2.");
-    PADDLE_ENFORCE_EQ(x_dims[1], 1,
-                      "Input(X) of SequenceEnumerate operator's 2nd "
-                      "dimension should be 1.");
-
     const auto win_size = ctx->Attrs().Get<int>("win_size");
     ctx->SetOutputDim("Out", {x_dims[0], win_size});
     ctx->ShareLoD("X", "Out");

paddle/fluid/operators/sequence_ops/sequence_enumerate_op.h

@@ -27,30 +27,47 @@ class SequenceEnumerateKernel : public framework::OpKernel<T> {
     auto* in = context.Input<LoDTensor>("X");
     auto* out = context.Output<LoDTensor>("Out");
     int win_size = context.Attr<int>("win_size");
-    int pad_value = context.Attr<int>("pad_value");
+    auto pad_value = static_cast<T>(context.Attr<int>("pad_value"));
 
     auto in_dims = in->dims();
-    auto in_lod = in->lod();
-
+    auto lod0 = in->lod()[0];
     PADDLE_ENFORCE_EQ(
-        static_cast<uint64_t>(in_dims[0]), in_lod[0].back(),
+        static_cast<uint64_t>(in_dims[0]), lod0.back(),
         "The actual input data's size mismatched with LoD information.");
+    PADDLE_ENFORCE_EQ(
+        in_dims.size(), 2UL,
+        "Input(X) of SequenceEnumerate operator's rank should be 2.");
+    PADDLE_ENFORCE_EQ(in_dims[1], 1,
+                      "Input(X) of SequenceEnumerate operator's 2nd "
+                      "dimension should be 1.");
 
     // Generate enumerate sequence set
-    auto lod0 = in_lod[0];
     auto in_data = in->data<T>();
     out->Resize({in_dims[0], win_size});
+    out->set_lod(in->lod());
     auto out_data = out->mutable_data<T>(context.GetPlace());
     for (size_t i = 0; i < lod0.size() - 1; ++i) {
-      for (size_t idx = lod0[i]; idx < lod0[i + 1]; ++idx) {
-        for (int word_idx = 0; word_idx < win_size; ++word_idx) {
-          size_t word_pos = idx + word_idx;
-          out_data[win_size * idx + word_idx] =
-              word_pos < lod0[i + 1] ? in_data[word_pos] : pad_value;
+      int start = lod0[i];
+      int end = lod0[i + 1];
+      int copy_size = win_size < end - start + 1 ? win_size : end - start + 1;
+      int mid = end + 1 - copy_size;
+      int pad_num = win_size - copy_size;
+      copy_size *= sizeof(T);
+      for (int idx = start; idx < mid; ++idx) {
+        std::memcpy(out_data, in_data + idx, copy_size);
+        out_data += win_size;
+      }
+      for (int idx = mid; idx < end; ++idx) {
+        copy_size -= sizeof(T);
+        pad_num++;
+        std::memcpy(out_data, in_data + idx, copy_size);
+        T* pdata = out_data + copy_size / sizeof(T);
+        for (int i = 0; i < pad_num; ++i) {
+          pdata[i] = pad_value;
         }
+        out_data += win_size;
       }
     }
-    out->set_lod(in->lod());
   }
 };
 

paddle/fluid/operators/slice_op.cu

@@ -31,18 +31,18 @@ __global__ void Padding(const paddle::platform::float16* d_out,
                         paddle::platform::float16* d_in) {
   int64_t out_idx = threadIdx.x + blockDim.x * blockIdx.x;
   if (out_idx < n) {
+    int64_t out_idx_tmp = out_idx;
     int coords[D] = {0};
     for (int i = D - 1; i >= 0; --i) {
-      coords[i] = out_idx % out_dims[i];
-      out_idx /= out_dims[i];
+      coords[i] = out_idx_tmp % out_dims[i];
+      out_idx_tmp /= out_dims[i];
       coords[i] += offsets[i];
     }
 
     int64_t in_idx = 0;
-    for (int i = 0; i < D - 1; ++i) {
-      in_idx += coords[i] * in_dims[i + 1];
+    for (int i = 0; i < D; ++i) {
+      in_idx = in_idx * in_dims[i] + coords[i];
     }
-    in_idx += coords[D - 1];
 
     d_in[in_idx] = d_out[out_idx];
   }
@@ -80,8 +80,8 @@ class SliceGradKernel<paddle::platform::CUDADeviceContext,
     set_zero(dev_ctx, d_in, static_cast<paddle::platform::float16>(0));
 
     int64_t numel = d_out->numel();
-    dim3 blocks((numel - 1) / PADDLE_CUDA_NUM_THREADS + 1, 1, 1);
-    dim3 threads(PADDLE_CUDA_NUM_THREADS, 1, 1);
+    dim3 blocks((numel - 1) / PADDLE_CUDA_NUM_THREADS + 1);
+    dim3 threads(PADDLE_CUDA_NUM_THREADS);
     auto stream = ctx.cuda_device_context().stream();
 
     auto out_shape = framework::vectorize2int(out_dims);

paddle/fluid/operators/softmax_with_cross_entropy_op.cc

@@ -13,6 +13,7 @@ See the License for the specific language governing permissions and
 limitations under the License. */
 
 #include "paddle/fluid/operators/softmax_with_cross_entropy_op.h"
+#include <memory>
 
 namespace paddle {
 namespace operators {
@@ -187,7 +188,6 @@ class SoftmaxGradMaker : public framework::SingleGradOpDescMaker {
     grad_op->SetType("softmax_with_cross_entropy_grad");
     grad_op->SetInput("Label", Input("Label"));
     grad_op->SetInput("Softmax", Output("Softmax"));
-    grad_op->SetInput("Loss", Output("Loss"));
     grad_op->SetInput(framework::GradVarName("Softmax"), OutputGrad("Softmax"));
     grad_op->SetInput(framework::GradVarName("Loss"), OutputGrad("Loss"));
     grad_op->SetOutput(framework::GradVarName("Logits"), InputGrad("Logits"));

paddle/fluid/operators/split_selected_rows_op.cc

@@ -14,6 +14,8 @@ limitations under the License. */
 
 #include "paddle/fluid/operators/split_selected_rows_op.h"
 
+#include <memory>
+
 namespace paddle {
 namespace operators {
 
@@ -60,10 +62,9 @@ class SplitSelectedRowsOp : public framework::OperatorWithKernel {
 
 class SplitSelectedRowsOpInferVarType : public framework::VarTypeInference {
  public:
-  void operator()(const framework::OpDesc &op_desc,
-                  framework::BlockDesc *block) const override {
-    for (auto &out_var : op_desc.Output("Out")) {
-      block->Var(out_var)->SetType(framework::proto::VarType::SELECTED_ROWS);
+  void operator()(framework::InferVarTypeContext *ctx) const override {
+    for (auto &out_var : ctx->Output("Out")) {
+      ctx->SetType(out_var, framework::proto::VarType::SELECTED_ROWS);
     }
   }
 };

paddle/fluid/operators/sum_op.cc

@@ -12,6 +12,7 @@ limitations under the License. */
 #include "paddle/fluid/operators/sum_op.h"
 
 #include <algorithm>
+#include <memory>
 #include <string>
 #include <vector>
 
@@ -159,24 +160,20 @@ the LoD information with the first input.
 
 class SumOpVarTypeInference : public framework::VarTypeInference {
  public:
-  void operator()(const framework::OpDesc& op_desc,
-                  framework::BlockDesc* block) const override {
-    auto& inputs = op_desc.Input("X");
+  void operator()(framework::InferVarTypeContext* ctx) const override {
+    auto& inputs = ctx->Input("X");
     auto var_type = framework::proto::VarType::SELECTED_ROWS;
-    for (auto& name : op_desc.Input("X")) {
-      VLOG(10) << name << " "
-               << block->FindRecursiveOrCreateVar(name).GetType();
+    for (auto& name : ctx->Input("X")) {
+      VLOG(10) << name << " " << ctx->GetType(name);
     }
 
     bool any_input_is_lod_tensor = std::any_of(
-        inputs.begin(), inputs.end(), [block](const std::string& name) {
-          return block->FindRecursiveOrCreateVar(name).GetType() ==
-                 framework::proto::VarType::LOD_TENSOR;
+        inputs.begin(), inputs.end(), [ctx](const std::string& name) {
+          return ctx->GetType(name) == framework::proto::VarType::LOD_TENSOR;
         });
 
-    auto is_tensor_array = [block](const std::string& name) {
-      return block->FindRecursiveOrCreateVar(name).GetType() ==
-             framework::proto::VarType::LOD_TENSOR_ARRAY;
+    auto is_tensor_array = [ctx](const std::string& name) {
+      return ctx->GetType(name) == framework::proto::VarType::LOD_TENSOR_ARRAY;
     };
 
     bool any_input_is_tensor_array =
@@ -188,8 +185,7 @@ class SumOpVarTypeInference : public framework::VarTypeInference {
       if (!all_inputs_are_tensor_array) {
         std::ostringstream os;
         for (auto& each : inputs) {
-          os << "    " << each << " type is "
-             << block->FindRecursiveOrCreateVar(each).GetType() << "\n";
+          os << "    " << each << " type is " << ctx->GetType(each) << "\n";
         }
         PADDLE_ENFORCE(all_inputs_are_tensor_array,
                        "Not all inputs are tensor array:\n%s", os.str());
@@ -199,11 +195,9 @@ class SumOpVarTypeInference : public framework::VarTypeInference {
       var_type = framework::proto::VarType::LOD_TENSOR;
     }
 
-    auto out_var_name = op_desc.Output("Out").front();
-    auto& out_var = block->FindRecursiveOrCreateVar(out_var_name);
-    out_var.SetType(var_type);
-    auto& in_var = detail::Ref(block->FindVarRecursive(inputs.front()));
-    out_var.SetDataType(in_var.GetDataType());
+    auto out_var_name = ctx->Output("Out").front();
+    ctx->SetType(out_var_name, var_type);
+    ctx->SetDataType(out_var_name, ctx->GetDataType(inputs.front()));
   }
 };
 

paddle/fluid/operators/tensor_array_to_tensor_op.cc

@@ -177,10 +177,9 @@ class LoDTensorArray2TensorGradInferShape : public framework::InferShapeBase {
 class LoDTensorArray2TensorGradInferVarType
     : public framework::VarTypeInference {
  public:
-  void operator()(const framework::OpDesc &op_desc,
-                  framework::BlockDesc *block) const override {
-    for (auto &out_var : op_desc.Output(framework::GradVarName("X"))) {
-      block->Var(out_var)->SetType(framework::proto::VarType::LOD_TENSOR_ARRAY);
+  void operator()(framework::InferVarTypeContext *ctx) const override {
+    for (auto &out_var : ctx->Output(framework::GradVarName("X"))) {
+      ctx->SetType(out_var, framework::proto::VarType::LOD_TENSOR_ARRAY);
     }
   }
 };

paddle/fluid/operators/tensorrt/tensorrt_engine_op.cc

@@ -46,8 +46,7 @@ class TensorRTEngineOpMaker : public framework::OpProtoAndCheckerMaker {
 
 class TensorRTEngineInferVarType : public framework::VarTypeInference {
  public:
-  void operator()(const framework::OpDesc &op_desc,
-                  framework::BlockDesc *block) const override {}
+  void operator()(framework::InferVarTypeContext *ctx) const override {}
 };
 
 }  // namespace operators

paddle/fluid/operators/uniform_random_op.cc

@@ -112,17 +112,16 @@ uniform distribution. The random result is in set [min, max].
 
 class UniformRandomOpVarTypeInference : public framework::VarTypeInference {
  public:
-  void operator()(const framework::OpDesc &op_desc,
-                  framework::BlockDesc *block) const override {
-    auto out_var_name = op_desc.Output("Out").front();
+  void operator()(framework::InferVarTypeContext *ctx) const override {
+    auto out_var_name = ctx->Output("Out").front();
     auto var_data_type = static_cast<framework::proto::VarType::Type>(
-        boost::get<int>(op_desc.GetAttr("dtype")));
+        boost::get<int>(ctx->GetAttr("dtype")));
 
-    auto out_var = block->FindRecursiveOrCreateVar(out_var_name);
-    if (out_var.GetType() != framework::proto::VarType::SELECTED_ROWS) {
-      out_var.SetType(framework::proto::VarType::LOD_TENSOR);
+    if (ctx->GetType(out_var_name) !=
+        framework::proto::VarType::SELECTED_ROWS) {
+      ctx->SetType(out_var_name, framework::proto::VarType::LOD_TENSOR);
     }
-    out_var.SetDataType(var_data_type);
+    ctx->SetDataType(out_var_name, var_data_type);
   }
 };
 

paddle/fluid/platform/device_context.h

@@ -23,6 +23,7 @@ limitations under the License. */
 #include "paddle/fluid/platform/cuda_helper.h"
 #include "paddle/fluid/platform/dynload/cublas.h"
 #include "paddle/fluid/platform/dynload/cudnn.h"
+#include "paddle/fluid/platform/dynload/nccl.h"
 #include "paddle/fluid/platform/gpu_info.h"
 #endif
 

paddle/fluid/platform/nccl_helper.h

@@ -22,6 +22,7 @@
 #include <typeindex>
 #include <unordered_map>
 #include <vector>
+
 #include "paddle/fluid/framework/data_type.h"
 #include "paddle/fluid/platform/dynload/nccl.h"
 #include "paddle/fluid/platform/enforce.h"
@@ -79,7 +80,6 @@ struct NCCLContext {
       : ctx_(new CUDADeviceContext(CUDAPlace(dev_id))), comm_{nullptr} {}
 
   cudaStream_t stream() const { return ctx_->stream(); }
-
   ncclComm_t comm() const { return comm_; }
 
   int device_id() const {
@@ -105,9 +105,6 @@ struct NCCLContextMap {
         order_.size(), contexts_.size(),
         "NCCL Context Map does not support contain two or more same device");
 
-    if (places.size() <= 1 && num_trainers == 1) {
-      return;
-    }
     std::unique_ptr<ncclComm_t[]> comms(new ncclComm_t[order_.size()]);
     // if num_trainers == 1, should create a new nccl id for local comms.
     if (num_trainers == 1 && nccl_id == nullptr) {
@@ -127,8 +124,8 @@ struct NCCLContextMap {
           } else {
             rank = trainer_id;
           }
-          VLOG(30) << "init nccl rank: " << rank << " nranks: " << nranks
-                   << "gpu id: " << gpu_id;
+          VLOG(3) << "init nccl rank: " << rank << " nranks: " << nranks
+                  << " gpu id: " << gpu_id;
           PADDLE_ENFORCE(cudaSetDevice(gpu_id));
           PADDLE_ENFORCE(platform::dynload::ncclCommInitRank(
               comms.get() + i, nranks, *nccl_id, rank));

paddle/fluid/pybind/imperative.cc

@@ -38,7 +38,7 @@ void BindTracer(pybind11::module* m) {
       .def("trace",
            [](imperative::Tracer& self, imperative::OpBase* op,
               const imperative::VarBasePtrMap& inputs,
-              const imperative::VarBasePtrMap& outputs,
+              imperative::VarBasePtrMap* outputs,
               framework::AttributeMap attrs_map,
               const platform::CPUPlace expected_place,
               const bool stop_gradient = false) {
@@ -49,7 +49,7 @@ void BindTracer(pybind11::module* m) {
       .def("trace",
            [](imperative::Tracer& self, imperative::OpBase* op,
               const imperative::VarBasePtrMap& inputs,
-              const imperative::VarBasePtrMap& outputs,
+              imperative::VarBasePtrMap* outputs,
               framework::AttributeMap attrs_map,
               const platform::CUDAPlace expected_place,
               const bool stop_gradient = false) {

paddle/fluid/pybind/pybind.cc

@@ -206,7 +206,7 @@ PYBIND11_MODULE(core, m) {
       .def_property("name", &imperative::VarBase::Name,
                     &imperative::VarBase::SetName)
       .def_property_readonly("shape", &imperative::VarBase::Shape)
-      .def_property_readonly("dtype", &imperative::VarBase::DType)
+      .def_property_readonly("dtype", &imperative::VarBase::DataType)
       .def_property("persistable", &imperative::VarBase::IsPersistable,
                     &imperative::VarBase::SetPersistable)
       .def_property("stop_gradient", &imperative::VarBase::IsStopGradient,
@@ -1281,6 +1281,10 @@ All parameter, weight, gradient are variables in Paddle.
           "enable_inplace",
           [](const BuildStrategy &self) { return self.enable_inplace_; },
           [](BuildStrategy &self, bool b) { self.enable_inplace_ = b; })
+      .def_property(
+          "fuse_all_reduce_ops",
+          [](const BuildStrategy &self) { return self.fuse_all_reduce_ops_; },
+          [](BuildStrategy &self, bool b) { self.fuse_all_reduce_ops_ = b; })
       .def("_finalize_strategy_and_create_passes",
            [](BuildStrategy &self) -> std::shared_ptr<ir::PassBuilder> {
              return self.CreatePassesFromStrategy(true);

python/paddle/fluid/__init__.py

@@ -132,7 +132,8 @@ def __bootstrap__():
         'allocator_strategy', 'reader_queue_speed_test_mode',
         'print_sub_graph_dir', 'pe_profile_fname', 'warpctc_dir',
         'inner_op_parallelism', 'enable_parallel_graph',
-        'multiple_of_cupti_buffer_size', 'enable_subgraph_optimize',
+        'fuse_parameter_groups_size', 'multiple_of_cupti_buffer_size',
+        'enable_subgraph_optimize', 'fuse_parameter_memory_size',
         'tracer_profile_fname'
     ]
     if 'Darwin' not in sysstr:

python/paddle/fluid/compiler.py

@@ -119,7 +119,7 @@ class CompiledProgram(object):
                 threads are used, how many iterations to clean up the temp
                 variables. For more information, please refer
                 to fluid.ExecutionStrategy. Default None.
-            share_vars_from(CompiledProgram): If provide, this CompiledProgram
+            share_vars_from(CompiledProgram): If provided, this CompiledProgram
                 will share variables from `share_vars_from`. `share_vars_from`
                 must be run by the executor before this CompiledProgram so that
                 vars are ready.

python/paddle/fluid/data_feeder.py

@@ -308,8 +308,8 @@ class DataFeeder(object):
         Args:
             reader(function): the reader is the function which can generate data.
             multi_devices(bool): whether to use multiple devices or not.
-            num_places(int): if the multi_devices is True, you can specify the number
-                of GPU to use, if 'num_places' is None, the function will use all the
+            num_places(int): if multi_devices is True, you can specify the number
+                of GPU to use, if multi_devices is None, the function will use all the
                 GPU of the current machine. Default None.
             drop_last(bool): whether to drop the last batch if the
                 size of the last batch is less than batch_size. Default True.
@@ -318,7 +318,7 @@ 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 cannot fit for devices.
         """
 
         def __reader_creator__():

python/paddle/fluid/executor.py

@@ -470,13 +470,21 @@ class Executor(object):
             program(Program|CompiledProgram): the program that need to run,
                 if not provided, then default_main_program (not compiled) will be used.
             feed(dict): feed variable map, e.g. {"image": ImageData, "label": LabelData}
-            fetch_list(list): a list of variable or variable names that user want to get, run will return them according to this list.
-            feed_var_name(str): the name for the input variable of feed Operator.
-            fetch_var_name(str): the name for the output variable of fetch Operator.
-            scope(Scope): the scope used to run this program, you can switch it to different scope. default is global_scope
+            fetch_list(list): a list of variable or variable names that user 
+                wants to get, this method will return them according to this list.
+            feed_var_name(str): the name for the input variable of 
+                feed Operator.
+            fetch_var_name(str): the name for the output variable of 
+                fetch Operator.
+            scope(Scope): the scope used to run this program, you can switch 
+                it to different scope. default is global_scope
             return_numpy(bool): if convert the fetched tensor to numpy
-            use_program_cache(bool): set use_program_cache to true if program not changed compare to the last step.
-
+            use_program_cache(bool): whether to use the cached program 
+                settings across batches. Setting it be true would be faster 
+                only when (1) the program is not compiled with data parallel, 
+                and (2) program, feed variable names and fetch_list variable 
+                names do not changed compared to the last step. 
+                
         Returns:
 
             list(numpy.array): fetch result according to fetch_list.

python/paddle/fluid/layers/__init__.py

@@ -33,6 +33,7 @@ from .detection import *
 from . import metric_op
 from .metric_op import *
 from .learning_rate_scheduler import *
+from .collective import *
 
 __all__ = []
 __all__ += nn.__all__

python/paddle/fluid/layers/collective.py

+#   Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from __future__ import print_function
+from ..layer_helper import LayerHelper, unique_name
+
+
+def _allreduce(x, out=None, reduce_type="sum"):
+    helper = LayerHelper("allreduce", **locals())
+    # Convert string reduce type to op int type
+    red_typ_int = 0
+    if reduce_type == "sum":
+        red_typ_int = 0
+    elif reduce_type == "prod":
+        red_typ_int = 1
+    elif reduce_type == "max":
+        red_typ_int = 2
+    elif reduce_type == "min":
+        red_typ_int = 3
+    else:
+        raise TypeError("reduce type can only be [sum|prod|max|min]")
+
+    if out is None:
+        out = helper.create_variable(
+            name=unique_name.generate(".".join([x.name, 'tmp'])),
+            shape=x.shape,
+            dtype=x.dtype,
+            type=x.type,
+            persistable=x.persistable,
+            stop_gradient=True)
+    helper.append_op(
+        type='allreduce',
+        inputs={'X': [x]},
+        outputs={'Out': [out]},
+        attrs={"reduce_type": red_typ_int})
+    return out

python/paddle/fluid/layers/control_flow.py

@@ -28,21 +28,9 @@ import six
 from functools import reduce
 
 __all__ = [
-    'While',
-    'Switch',
-    'increment',
-    'array_write',
-    'create_array',
-    'less_than',
-    'equal',
-    'array_read',
-    'array_length',
-    'IfElse',
-    'DynamicRNN',
-    'StaticRNN',
-    'reorder_lod_tensor_by_rank',
-    'Print',
-    'is_empty',
+    'While', 'Switch', 'increment', 'array_write', 'create_array', 'less_than',
+    'equal', 'array_read', 'array_length', 'IfElse', 'DynamicRNN', 'StaticRNN',
+    'reorder_lod_tensor_by_rank', 'Print', 'is_empty'
 ]
 
 
@@ -1448,12 +1436,13 @@ class DynamicRNN(object):
         self.input_array = []
         self.mem_link = []
 
-    def step_input(self, x):
+    def step_input(self, x, level=0):
         """
         Mark a sequence as a dynamic RNN input.
 
         Args:
             x(Variable): The input sequence.
+            level(int): The level of lod used to split steps. Default: 0.
 
         Returns:
             The current timestep in the input sequence.
@@ -1471,7 +1460,8 @@ class DynamicRNN(object):
             parent_block.append_op(
                 type='lod_rank_table',
                 inputs={"X": x},
-                outputs={"Out": self.lod_rank_table})
+                outputs={"Out": self.lod_rank_table},
+                attrs={"level": level})
             self.max_seq_len = parent_block.create_var(
                 name=unique_name.generate('dynamic_rnn_max_seq_len'),
                 dtype='int64')

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

+# Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from __future__ import print_function
+
+import numpy as np
+import argparse
+import time
+import math
+
+import paddle
+import paddle.fluid as fluid
+import paddle.fluid.profiler as profiler
+from paddle.fluid import core
+import unittest
+from multiprocessing import Process
+import os
+import signal
+from functools import reduce
+from test_dist_base import TestDistRunnerBase, runtime_main
+
+DTYPE = "float32"
+paddle.dataset.mnist.fetch()
+
+# Fix seed for test
+fluid.default_startup_program().random_seed = 1
+fluid.default_main_program().random_seed = 1
+
+
+def cnn_model(data):
+    conv_pool_1 = fluid.nets.simple_img_conv_pool(
+        input=data,
+        filter_size=5,
+        num_filters=20,
+        pool_size=2,
+        pool_stride=2,
+        act="relu",
+        param_attr=fluid.ParamAttr(initializer=fluid.initializer.Constant(
+            value=0.01)))
+    conv_pool_2 = fluid.nets.simple_img_conv_pool(
+        input=conv_pool_1,
+        filter_size=5,
+        num_filters=50,
+        pool_size=2,
+        pool_stride=2,
+        act="relu",
+        param_attr=fluid.ParamAttr(initializer=fluid.initializer.Constant(
+            value=0.01)))
+
+    SIZE = 10
+    input_shape = conv_pool_2.shape
+    param_shape = [reduce(lambda a, b: a * b, input_shape[1:], 1)] + [SIZE]
+    scale = (2.0 / (param_shape[0]**2 * SIZE))**0.5
+
+    predict = fluid.layers.fc(
+        input=conv_pool_2,
+        size=SIZE,
+        act="softmax",
+        param_attr=fluid.param_attr.ParamAttr(
+            initializer=fluid.initializer.Constant(value=0.01)))
+    return predict
+
+
+class TestDistMnist2x2(TestDistRunnerBase):
+    def get_model(self, batch_size=2, single_device=False):
+        # Input data
+        images = fluid.layers.data(name='pixel', shape=[1, 28, 28], dtype=DTYPE)
+        label = fluid.layers.data(name='label', shape=[1], dtype='int64')
+
+        # Train program
+        predict = cnn_model(images)
+        cost = fluid.layers.cross_entropy(input=predict, label=label)
+        avg_cost = fluid.layers.mean(x=cost)
+
+        # Evaluator
+        batch_size_tensor = fluid.layers.create_tensor(dtype='int64')
+        batch_acc = fluid.layers.accuracy(
+            input=predict, label=label, total=batch_size_tensor)
+
+        inference_program = fluid.default_main_program().clone()
+
+        # Reader
+        train_reader = paddle.batch(
+            paddle.dataset.mnist.test(), batch_size=batch_size)
+        test_reader = paddle.batch(
+            paddle.dataset.mnist.test(), batch_size=batch_size)
+
+        # Optimization
+        # TODO(typhoonzero): fix distributed adam optimizer
+        # opt = fluid.optimizer.AdamOptimizer(
+        #     learning_rate=0.001, beta1=0.9, beta2=0.999)
+        opt = fluid.optimizer.Momentum(learning_rate=self.lr, momentum=0.9)
+        if single_device:
+            opt.minimize(avg_cost)
+        else:
+            # multi device or distributed multi device
+            params_grads = opt.backward(avg_cost)
+            data_parallel_param_grads = []
+            for p, g in params_grads:
+                # NOTE: scale will be done on loss scale in multi_devices_graph_pass using nranks.
+                grad_reduce = fluid.layers.collective._allreduce(g)
+                data_parallel_param_grads.append([p, grad_reduce])
+            opt.apply_gradients(data_parallel_param_grads)
+
+        return inference_program, avg_cost, train_reader, test_reader, batch_acc, predict
+
+
+if __name__ == "__main__":
+    runtime_main(TestDistMnist2x2)

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

@@ -43,6 +43,7 @@ class TestParallelExecutorBase(unittest.TestCase):
                                   use_ir_memory_optimize=True,
                                   enable_inplace=True,
                                   fuse_elewise_add_act_ops=False,
+                                  fuse_all_reduce_ops=False,
                                   fuse_relu_depthwise_conv=False,
                                   optimizer=fluid.optimizer.Adam,
                                   use_fast_executor=False,
@@ -80,6 +81,7 @@ class TestParallelExecutorBase(unittest.TestCase):
         build_strategy.fuse_elewise_add_act_ops = fuse_elewise_add_act_ops
         build_strategy.fuse_relu_depthwise_conv = fuse_relu_depthwise_conv
         build_strategy.memory_optimize = False if memory_opt else use_ir_memory_optimize
+        build_strategy.fuse_all_reduce_ops = fuse_all_reduce_ops
         # python memory optimization is conflict with inplace pass.
         # Use ir graph memory optimization after inplace pass is the correct way.
         build_strategy.enable_inplace = False if memory_opt else enable_inplace

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

+# Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from __future__ import print_function
+import unittest
+from test_dist_base import TestDistBase
+
+
+class TestDistMnistNCCL2(TestDistBase):
+    def _setup_config(self):
+        self._sync_mode = True
+        self._use_reduce = False
+        self._use_reader_alloc = False
+        self._nccl2_mode = True
+        self._nccl2_reduce_layer = True
+
+    def test_dist_train(self):
+        import paddle.fluid as fluid
+        if fluid.core.is_compiled_with_cuda():
+            self.check_with_place("dist_allreduce_op.py", delta=1e-5)
+
+
+if __name__ == '__main__':
+    unittest.main()

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

@@ -33,7 +33,10 @@ DEFAULT_BATCH_SIZE = 2
 
 
 class TestDistRunnerBase(object):
-    def get_model(self, batch_size=DEFAULT_BATCH_SIZE, lr=0.1):
+    def get_model(self,
+                  batch_size=DEFAULT_BATCH_SIZE,
+                  lr=0.1,
+                  single_device=False):
         raise NotImplementedError(
             "get_model should be implemented by child classes.")
 
@@ -76,8 +79,12 @@ class TestDistRunnerBase(object):
 
     def run_trainer(self, args):
         self.lr = args.lr
-        test_program, avg_cost, train_reader, test_reader, batch_acc, predict = \
-            self.get_model(batch_size=args.batch_size)
+        if args.nccl2_reduce_layer_local_run:
+            test_program, avg_cost, train_reader, test_reader, batch_acc, predict = \
+                self.get_model(batch_size=args.batch_size, single_device=True)
+        else:
+            test_program, avg_cost, train_reader, test_reader, batch_acc, predict = \
+                self.get_model(batch_size=args.batch_size)
 
         if args.mem_opt:
             fluid.memory_optimize(fluid.default_main_program(), skip_grads=True)
@@ -87,7 +94,7 @@ class TestDistRunnerBase(object):
                                     args.endpoints, args.trainers,
                                     args.sync_mode, args.dc_asgd)
             trainer_prog = t.get_trainer_program()
-        elif args.update_method == "nccl2":
+        elif args.update_method == "nccl2" or args.update_method == "nccl2_reduce_layer":
             # transpile for nccl2
             config = fluid.DistributeTranspilerConfig()
             config.mode = "nccl2"
@@ -110,9 +117,9 @@ class TestDistRunnerBase(object):
         exe = fluid.Executor(place)
         exe.run(fluid.default_startup_program())
 
-        strategy = fluid.ExecutionStrategy()
-        strategy.num_threads = 1
-        strategy.allow_op_delay = False
+        exec_strategy = fluid.ExecutionStrategy()
+        exec_strategy.num_threads = 1
+        exec_strategy.allow_op_delay = False
 
         build_stra = fluid.BuildStrategy()
         # FIXME force disable enable_inplace and memory_optimize
@@ -124,23 +131,25 @@ class TestDistRunnerBase(object):
         else:
             build_stra.reduce_strategy = fluid.BuildStrategy.ReduceStrategy.AllReduce
 
+        pass_builder = None
         if args.batch_merge_repeat > 1:
             pass_builder = build_stra._finalize_strategy_and_create_passes()
             mypass = pass_builder.insert_pass(
                 len(pass_builder.all_passes()) - 3, "multi_batch_merge_pass")
             mypass.set("num_repeats", args.batch_merge_repeat)
 
-        if args.update_method == "nccl2":
+        if args.update_method == "nccl2" or args.update_method == "nccl2_reduce_layer":
             build_stra.num_trainers = len(args.endpoints.split(","))
             build_stra.trainer_id = args.trainer_id
         else:
+            # case args.update_method == "nccl2_reduce_layer":
             build_stra.num_trainers = 1
             build_stra.trainer_id = 0
 
         binary = compiler.CompiledProgram(trainer_prog).with_data_parallel(
             loss_name=avg_cost.name,
             build_strategy=build_stra,
-            exec_strategy=strategy)
+            exec_strategy=exec_strategy)
 
         feed_var_list = [
             var for var in trainer_prog.global_block().vars.values()
@@ -182,7 +191,7 @@ def runtime_main(test_class):
         '--update_method',
         type=str,
         default="local",
-        choices=["pserver", "nccl2", "local"])
+        choices=["pserver", "nccl2", "local", "nccl2_reduce_layer"])
     parser.add_argument('--trainer_id', type=int, required=False, default=0)
     parser.add_argument('--trainers', type=int, required=False, default=1)
     parser.add_argument(
@@ -198,6 +207,11 @@ def runtime_main(test_class):
     parser.add_argument('--lr', required=False, type=float, default=0.001)
     parser.add_argument(
         '--batch_merge_repeat', required=False, type=int, default=1)
+    parser.add_argument(
+        '--nccl2_reduce_layer_local_run',
+        required=False,
+        type=bool,
+        default=False)
 
     args = parser.parse_args()
 
@@ -242,6 +256,11 @@ class TestDistBase(unittest.TestCase):
         self._dc_asgd = False  # must use with async mode
         self._use_reader_alloc = True
         self._nccl2_mode = False
+        # FIXME(typhoonzero): I added this stupid argument to enable
+        # testing allreduce layers, which users can call layers.allreduce
+        # to accumulate tensors at anywhere. Find a better way to do this
+        # test, reduce check this argument everywhere.
+        self._nccl2_reduce_layer = False
         self._lr = 0.001
         self._setup_config()
         self._after_setup_config()
@@ -307,10 +326,16 @@ class TestDistBase(unittest.TestCase):
             cmd += " --batch_size %d" % batch_size
         if batch_merge_repeat > 1:
             cmd += " --batch_merge_repeat %d" % batch_merge_repeat
+        if self._nccl2_reduce_layer:
+            cmd += " --nccl2_reduce_layer_local_run 1"
 
         if self.__use_cuda:
             cmd += " --use_cuda"
-            env_local = {"CUDA_VISIBLE_DEVICES": "0"}
+            env_local = {
+                "CUDA_VISIBLE_DEVICES": "0",
+                "PADDLE_TRAINERS_NUM": "1",
+                "PADDLE_TRAINER_ID": "0"
+            }
         else:
             env_local = {'CPU_NUM': '1'}
 
@@ -427,29 +452,30 @@ class TestDistBase(unittest.TestCase):
             sys.stderr.write("ps1 stderr: %s\n" % fn.read())
 
         # print log
-        if stat0 == 0:
-            sys.stderr.write('trainer 0 stdout: %s\n' % pickle.loads(tr0_out))
         with open("/tmp/tr0_err.log", "r") as fn:
             sys.stderr.write('trainer 0 stderr: %s\n' % fn.read())
-        if stat1 == 0:
-            sys.stderr.write('trainer 1 stdout: %s\n' % pickle.loads(tr1_out))
         with open("/tmp/tr1_err.log", "r") as fn:
             sys.stderr.write('trainer 1 stderr: %s\n' % fn.read())
 
         return pickle.loads(tr0_out), pickle.loads(tr1_out)
 
-    def _run_cluster_nccl2(self, model, envs, check_error_log):
+    def _run_cluster_nccl2(self, model, envs, nccl2_reduce_layer,
+                           check_error_log):
         # NOTE: we reuse ps_endpoints as nccl2 worker endpoints
         worker_endpoints = self._ps_endpoints.split(",")
         w0_ep, w1_ep = worker_endpoints
+        if nccl2_reduce_layer:
+            update_method = "nccl2_reduce_layer"
+        else:
+            update_method = "nccl2"
 
-        tr_cmd = "%s %s --role trainer --endpoints %s --trainer_id %d --current_endpoint %s --update_method nccl2 --lr %f"
+        tr_cmd = "%s %s --role trainer --endpoints %s --trainer_id %d --current_endpoint %s --update_method %s --lr %f"
         tr0_cmd = tr_cmd % \
                   (self._python_interp, model, self._ps_endpoints,
-                   0, w0_ep, self._lr)
+                   0, w0_ep, update_method, self._lr)
         tr1_cmd = tr_cmd % \
                   (self._python_interp, model, self._ps_endpoints,
-                   1, w1_ep, self._lr)
+                   1, w1_ep, update_method, self._lr)
 
         if self._mem_opt:
             tr0_cmd += " --mem_opt"
@@ -463,8 +489,17 @@ class TestDistBase(unittest.TestCase):
         if self.__use_cuda:
             tr0_cmd += " --use_cuda"
             tr1_cmd += " --use_cuda"
-            env0 = {"CUDA_VISIBLE_DEVICES": "0"}
-            env1 = {"CUDA_VISIBLE_DEVICES": "1"}
+            env0 = {
+                "CUDA_VISIBLE_DEVICES": "0",
+                # for test nccl2 layer
+                "PADDLE_TRAINERS_NUM": "2",
+                "PADDLE_TRAINER_ID": "0"
+            }
+            env1 = {
+                "CUDA_VISIBLE_DEVICES": "1",
+                "PADDLE_TRAINERS_NUM": "2",
+                "PADDLE_TRAINER_ID": "1"
+            }
         else:
             env0 = {'CPU_NUM': '1'}
             env1 = {'CPU_NUM': '1'}
@@ -498,8 +533,6 @@ class TestDistBase(unittest.TestCase):
         # print log
         sys.stderr.write('trainer 0 stderr: %s\n' % tr0_err)
         sys.stderr.write('trainer 1 stderr: %s\n' % tr1_err)
-        sys.stderr.write('trainer 0 stdout: %s\n' % tr0_out)
-        sys.stderr.write('trainer 1 stdout: %s\n' % tr1_out)
 
         return pickle.loads(tr0_out), pickle.loads(tr1_out)
 
@@ -528,10 +561,14 @@ class TestDistBase(unittest.TestCase):
 
         local_losses\
             = self._run_local(model_file, required_envs,
-                                       check_error_log)
+                                check_error_log)
         if self._nccl2_mode:
-            tr0_losses, tr1_losses = self._run_cluster_nccl2(
-                model_file, required_envs, check_error_log)
+            if self._nccl2_reduce_layer:
+                tr0_losses, tr1_losses = self._run_cluster_nccl2(
+                    model_file, required_envs, True, check_error_log)
+            else:
+                tr0_losses, tr1_losses = self._run_cluster_nccl2(
+                    model_file, required_envs, False, check_error_log)
         else:
             tr0_losses, tr1_losses = self._run_cluster(
                 model_file, required_envs, check_error_log)

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

-#   Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserved.
+# 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.

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

+# Copyright (c) 2018 PaddlePaddle Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from parallel_executor_test_base import TestParallelExecutorBase
+import paddle.fluid as fluid
+import paddle.fluid.core as core
+import numpy as np
+import paddle
+import paddle.dataset.mnist as mnist
+import unittest
+import os
+
+
+def simple_fc_net(use_feed):
+    img = fluid.layers.data(name='image', shape=[784], dtype='float32')
+    label = fluid.layers.data(name='label', shape=[1], dtype='int64')
+
+    hidden = img
+    for _ in range(4):
+        hidden = fluid.layers.fc(
+            hidden,
+            size=200,
+            act='relu',
+            bias_attr=fluid.ParamAttr(
+                initializer=fluid.initializer.Constant(value=1.0)))
+    prediction = fluid.layers.fc(hidden, size=10, act='softmax')
+    loss = fluid.layers.cross_entropy(input=prediction, label=label)
+    loss = fluid.layers.mean(loss)
+    return loss
+
+
+def fc_with_batchnorm(use_feed):
+    img = fluid.layers.data(name='image', shape=[784], dtype='float32')
+    label = fluid.layers.data(name='label', shape=[1], dtype='int64')
+
+    hidden = img
+    for _ in range(2):
+        hidden = fluid.layers.fc(
+            hidden,
+            size=200,
+            act='relu',
+            bias_attr=fluid.ParamAttr(
+                initializer=fluid.initializer.Constant(value=1.0)))
+
+        hidden = fluid.layers.batch_norm(input=hidden)
+
+    prediction = fluid.layers.fc(hidden, size=10, act='softmax')
+    loss = fluid.layers.cross_entropy(input=prediction, label=label)
+    loss = fluid.layers.mean(loss)
+    return loss
+
+
+class TestMNIST(TestParallelExecutorBase):
+    @classmethod
+    def setUpClass(cls):
+        os.environ['CPU_NUM'] = str(4)
+
+    def _init_data(self, random=True):
+        np.random.seed(5)
+        if random:
+            img = np.random.random(size=[32, 784]).astype(np.float32)
+        else:
+            img = np.ones(shape=[32, 784], dtype='float32')
+        label = np.ones(shape=[32, 1], dtype='int64')
+        return img, label
+
+    def _compare_fuse_all_reduce_ops(self, model, use_cuda, random_data=True):
+        if use_cuda and not core.is_compiled_with_cuda():
+            return
+        img, label = self._init_data(random_data)
+
+        def _optimizer(learning_rate=1e-6):
+            optimizer = fluid.optimizer.SGD(
+                learning_rate=learning_rate,
+                regularization=fluid.regularizer.L2Decay(1e-6))
+            return optimizer
+
+        not_fuse_op_first_loss, not_fuse_op_last_loss = self.check_network_convergence(
+            model,
+            feed_dict={"image": img,
+                       "label": label},
+            use_cuda=use_cuda,
+            fuse_all_reduce_ops=False,
+            memory_opt=False,
+            optimizer=_optimizer)
+        fuse_op_first_loss, fuse_op_last_loss = self.check_network_convergence(
+            model,
+            feed_dict={"image": img,
+                       "label": label},
+            use_cuda=use_cuda,
+            fuse_all_reduce_ops=True,
+            memory_opt=False,
+            optimizer=_optimizer)
+
+        for loss in zip(not_fuse_op_first_loss, fuse_op_first_loss):
+            self.assertAlmostEquals(loss[0], loss[1], delta=1e-6)
+        for loss in zip(not_fuse_op_last_loss, fuse_op_last_loss):
+            self.assertAlmostEquals(loss[0], loss[1], delta=1e-6)
+
+    def test_simple_fc_with_fuse_op(self):
+        self._compare_fuse_all_reduce_ops(simple_fc_net, True)
+        self._compare_fuse_all_reduce_ops(simple_fc_net, False)
+
+    def test_batchnorm_fc_with_fuse_op(self):
+        self._compare_fuse_all_reduce_ops(fc_with_batchnorm, True)
+        self._compare_fuse_all_reduce_ops(fc_with_batchnorm, False)
+
+
+if __name__ == '__main__':
+    unittest.main()

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

@@ -87,5 +87,31 @@ class TestFP16(TestSliceOp):
                 place, ['Input'], 'Out', max_relative_error=0.006)
 
 
+@unittest.skipIf(not core.is_compiled_with_cuda(),
+                 "core is not compiled with CUDA")
+class TestFP16_2(TestSliceOp):
+    def config(self):
+        self.dtype = "float16"
+        self.input = np.random.random([3, 4, 5]).astype(self.dtype)
+        self.starts = [0]
+        self.ends = [1]
+        self.axes = [1]
+        self.out = self.input[:, 0:1, :]
+
+    def test_check_output(self):
+        place = core.CUDAPlace(0)
+        if core.is_float16_supported(place):
+            self.check_output_with_place(place, atol=1e-5)
+
+    def test_check_grad_normal(self):
+        place = core.CUDAPlace(0)
+        if core.is_float16_supported(place):
+            self.check_grad_with_place(
+                place, ['Input'],
+                'Out',
+                max_relative_error=0.006,
+                numeric_grad_delta=0.5)
+
+
 if __name__ == '__main__':
     unittest.main()

python/paddle/reader/__init__.py

@@ -38,9 +38,8 @@ items. It can be any function with no parameter that creates a iterable
 Element produced from the iterable should be a **single** entry of data,
 **not** a mini batch. That entry of data could be a single item, or a tuple of
 items.
-Item should be of `supported type <http://www.paddlepaddle.org/doc/ui/data_provider
-/pydataprovider2.html?highlight=dense_vector#input-types>`_ (e.g., numpy 1d
-array of float32, int, list of int)
+Item should be of supported type (e.g., numpy array or list/tuple of float 
+or int).
 
 An example implementation for single item data reader creator:
 
@@ -62,8 +61,6 @@ An example implementation for multiple item data reader creator:
                 yield numpy.random.uniform(-1, 1, size=width*height), label
     return reader
 
-
-TODO(yuyang18): Should we add whole design doc here?
 """
 
 import paddle.reader.decorator