merge develop

Dockerfile

@@ -75,8 +75,9 @@ RUN curl -s -q https://glide.sh/get | sh
 #    and its size is only one-third of the official one.
 # 2. Manually add ~IPluginFactory() in IPluginFactory class of NvInfer.h, otherwise, it couldn't work in paddle.
 #    See https://github.com/PaddlePaddle/Paddle/issues/10129 for details.
-RUN wget -qO- http://paddlepaddledeps.cdn.bcebos.com/TensorRT-4.0.0.3.Ubuntu-16.04.4.x86_64-gnu.cuda-8.0.cudnn7.0.tar.gz | \
-    tar -xz -C /usr/local && \
+
+RUN wget -q https://paddlepaddledeps.cdn.bcebos.com/TensorRT-4.0.1.6-ubuntu14.04.x86_64-gnu.cuda.8.0.cudnn7.0.tar.gz --no-check-certificate && \
+    tar -zxf TensorRT-4.0.1.6-ubuntu14.04.x86_64-gnu.cuda.8.0.cudnn7.0.tar.gz -C /usr/local && \
     cp -rf /usr/local/TensorRT/include /usr && \
     cp -rf /usr/local/TensorRT/lib /usr
 

paddle/fluid/API.spec

@@ -238,7 +238,7 @@ paddle.fluid.layers.load (ArgSpec(args=['out', 'file_path', 'load_as_fp16'], var
 paddle.fluid.layers.create_tensor (ArgSpec(args=['dtype', 'name', 'persistable'], varargs=None, keywords=None, defaults=(None, False)), ('document', 'c0c3d0194f83fff8ea99ce0820657dae'))
 paddle.fluid.layers.create_parameter (ArgSpec(args=['shape', 'dtype', 'name', 'attr', 'is_bias', 'default_initializer'], varargs=None, keywords=None, defaults=(None, None, False, None)), ('document', 'd62b866c899bc1fedb5385f95b88e1f8'))
 paddle.fluid.layers.create_global_var (ArgSpec(args=['shape', 'value', 'dtype', 'persistable', 'force_cpu', 'name'], varargs=None, keywords=None, defaults=(False, False, None)), ('document', 'ab914fac893607e29ac6e52bbdbea1a4'))
-paddle.fluid.layers.cast (ArgSpec(args=['x', 'dtype'], varargs=None, keywords=None, defaults=None), ('document', '60cb8f843d625abf33f8bf12455b8f99'))
+paddle.fluid.layers.cast (ArgSpec(args=['x', 'dtype'], varargs=None, keywords=None, defaults=None), ('document', '992eb42590fc1c380841a6db72ce78b3'))
 paddle.fluid.layers.tensor_array_to_tensor (ArgSpec(args=['input', 'axis', 'name'], varargs=None, keywords=None, defaults=(1, None)), ('document', 'b12717d3d4567e6119589f7f655b0cbb'))
 paddle.fluid.layers.concat (ArgSpec(args=['input', 'axis', 'name'], varargs=None, keywords=None, defaults=(0, None)), ('document', 'b19b79be4f05e85d1d6cec642c9fb535'))
 paddle.fluid.layers.sums (ArgSpec(args=['input', 'out'], varargs=None, keywords=None, defaults=(None,)), ('document', '42912092418620b4be07f36af31e7816'))
@@ -262,7 +262,7 @@ paddle.fluid.layers.Switch.default (ArgSpec(args=['self'], varargs=None, keyword
 paddle.fluid.layers.increment (ArgSpec(args=['x', 'value', 'in_place'], varargs=None, keywords=None, defaults=(1.0, True)), ('document', '73bb96ec4783ec1a11e760e8851b0e77'))
 paddle.fluid.layers.array_write (ArgSpec(args=['x', 'i', 'array'], varargs=None, keywords=None, defaults=(None,)), ('document', '40b6d15f4c86b2b09df340d7778ad713'))
 paddle.fluid.layers.create_array (ArgSpec(args=['dtype'], varargs=None, keywords=None, defaults=None), ('document', '2d4f20087080ba5105b55205ad5c5b6a'))
-paddle.fluid.layers.less_than (ArgSpec(args=['x', 'y', 'force_cpu', 'cond'], varargs=None, keywords='ignored', defaults=(None, None)), ('document', '067bbc799c66289ca8b8924c26b6673f'))
+paddle.fluid.layers.less_than (ArgSpec(args=['x', 'y', 'force_cpu', 'cond'], varargs=None, keywords=None, defaults=(None, None)), ('document', '067bbc799c66289ca8b8924c26b6673f'))
 paddle.fluid.layers.equal (ArgSpec(args=['x', 'y', 'cond'], varargs=None, keywords=None, defaults=(None,)), ('document', '80c29b1dc64718f0116de90d1ac88a77'))
 paddle.fluid.layers.array_read (ArgSpec(args=['array', 'i'], varargs=None, keywords=None, defaults=None), ('document', '0275133f1dde2aed528b4d3230edf823'))
 paddle.fluid.layers.array_length (ArgSpec(args=['array'], varargs=None, keywords=None, defaults=None), ('document', 'ffb8b9578ec66db565b223d313aa82a2'))
@@ -287,7 +287,7 @@ paddle.fluid.layers.StaticRNN.step_output (ArgSpec(args=['self', 'o'], varargs=N
 paddle.fluid.layers.StaticRNN.update_memory (ArgSpec(args=['self', 'mem', 'var'], varargs=None, keywords=None, defaults=None), ('document', '6adf97f83acf6453d4a6a4b1070f3754'))
 paddle.fluid.layers.reorder_lod_tensor_by_rank (ArgSpec(args=['x', 'rank_table'], varargs=None, keywords=None, defaults=None), ('document', '3545f529ef04e8f6ecb76b47fa3df01a'))
 paddle.fluid.layers.Print (ArgSpec(args=['input', 'first_n', 'message', 'summarize', 'print_tensor_name', 'print_tensor_type', 'print_tensor_shape', 'print_tensor_lod', 'print_phase'], varargs=None, keywords=None, defaults=(-1, None, -1, True, True, True, True, 'both')), ('document', '5fef91b0e21c93610785f2b1f7161732'))
-paddle.fluid.layers.is_empty (ArgSpec(args=['x', 'cond'], varargs=None, keywords='ignored', defaults=(None,)), ('document', 'bbe578dbb49ad13e15b014e98c22b519'))
+paddle.fluid.layers.is_empty (ArgSpec(args=['x', 'cond'], varargs=None, keywords=None, defaults=(None,)), ('document', 'bbe578dbb49ad13e15b014e98c22b519'))
 paddle.fluid.layers.sigmoid (ArgSpec(args=['x', 'name'], varargs=None, keywords=None, defaults=(None,)), ('document', '29a25ba78de79152076cacfc5443137d'))
 paddle.fluid.layers.logsigmoid (ArgSpec(args=['x', 'name'], varargs=None, keywords=None, defaults=(None,)), ('document', '81ccb7acafd06c7728e11581f5d342e3'))
 paddle.fluid.layers.exp (ArgSpec(args=['x', 'name'], varargs=None, keywords=None, defaults=(None,)), ('document', 'e6b3e769413d96aab4176f96db25984b'))
@@ -329,6 +329,8 @@ paddle.fluid.layers.polygon_box_transform (ArgSpec(args=['input', 'name'], varar
 paddle.fluid.layers.yolov3_loss (ArgSpec(args=['x', 'gtbox', 'gtlabel', 'anchors', 'anchor_mask', 'class_num', 'ignore_thresh', 'downsample_ratio', 'name'], varargs=None, keywords=None, defaults=(None,)), ('document', '991e934c3e09abf0edec7c9c978b4691'))
 paddle.fluid.layers.box_clip (ArgSpec(args=['input', 'im_info', 'name'], varargs=None, keywords=None, defaults=(None,)), ('document', '397e9e02b451d99c56e20f268fa03f2e'))
 paddle.fluid.layers.multiclass_nms (ArgSpec(args=['bboxes', 'scores', 'score_threshold', 'nms_top_k', 'keep_top_k', 'nms_threshold', 'normalized', 'nms_eta', 'background_label', 'name'], varargs=None, keywords=None, defaults=(0.3, True, 1.0, 0, None)), ('document', 'ca7d1107b6c5d2d6d8221039a220fde0'))
+paddle.fluid.layers.distribute_fpn_proposals (ArgSpec(args=['fpn_rois', 'min_level', 'max_level', 'refer_level', 'refer_scale', 'name'], varargs=None, keywords=None, defaults=(None,)), ('document', '7bb011ec26bace2bc23235aa4a17647d'))
+paddle.fluid.layers.box_decoder_and_assign (ArgSpec(args=['prior_box', 'prior_box_var', 'target_box', 'box_score', 'box_clip', 'name'], varargs=None, keywords=None, defaults=(None,)), ('document', '005a5ae47d6c8fff721931d69d072b9f'))
 paddle.fluid.layers.accuracy (ArgSpec(args=['input', 'label', 'k', 'correct', 'total'], varargs=None, keywords=None, defaults=(1, None, None)), ('document', '9808534c12c5e739a10f73ebb0b4eafd'))
 paddle.fluid.layers.auc (ArgSpec(args=['input', 'label', 'curve', 'num_thresholds', 'topk', 'slide_steps'], varargs=None, keywords=None, defaults=('ROC', 4095, 1, 1)), ('document', 'e0e95334fce92d16c2d9db6e7caffc47'))
 paddle.fluid.layers.exponential_decay (ArgSpec(args=['learning_rate', 'decay_steps', 'decay_rate', 'staircase'], varargs=None, keywords=None, defaults=(False,)), ('document', '98a5050bee8522fcea81aa795adaba51'))

paddle/fluid/framework/details/memory_optimize_helper.cc

@@ -20,6 +20,9 @@
 #include <numeric>
 #include <sstream>
 #include <string>
+#include <unordered_map>
+#include <unordered_set>
+#include "paddle/fluid/framework/operator.h"
 #include "paddle/fluid/framework/var_desc.h"
 #include "paddle/fluid/platform/cpu_info.h"
 
@@ -302,7 +305,10 @@ std::string OrderedSet::ToString() const {
 
 bool NodeCanReused(ir::Node* node) {
   // valid the node is a var node
-  if (node == nullptr || !node->IsVar() || node->IsCtrlVar()) return false;
+  // vars can be @EMPTY@, @LR_DECAY_REUSE_ID@. For example, while_grad
+  if (node == nullptr || !node->IsVar() || node->IsCtrlVar() ||
+      node->Name() == kEmptyVarName)
+    return false;
 
   bool flag = true;
   // op output force generated in cpu, can not be reused.
@@ -348,10 +354,6 @@ bool NodeCanReused(const VarDesc& node) {
   if (shape.empty() || size < MinChunkSize()) {
     return false;
   }
-  // vars can be @EMPTY@, @LR_DECAY_REUSE_ID@. For example, while_grad
-  std::string name = node.Name();
-  if (!name.empty() && name[0] == '@' && name[name.size() - 1] == '@')
-    return false;
   return true;
 }
 

paddle/fluid/framework/details/parallel_ssa_graph_executor.cc

@@ -13,6 +13,8 @@
 // limitations under the License.
 
 #include "paddle/fluid/framework/details/parallel_ssa_graph_executor.h"
+#include <memory>
+#include <utility>
 #include "paddle/fluid/framework/ir/graph_helper.h"
 
 namespace paddle {
@@ -29,6 +31,11 @@ ParallelSSAGraphExecutor::SeparateMultiDevicesGraph(ir::Graph *graph) {
     auto &g = graphs.back();
     g->Set(kGraphVars, new GraphVars(1UL));
     g->Set(kGraphDepVars, new GraphDepVars);
+    auto &stale_ops =
+        graph->Get<const std::vector<OpDesc *>>(details::kStaleProgramOpDescs);
+    g->Erase(details::kStaleProgramOpDescs);
+    g->Set<const std::vector<OpDesc *>>(details::kStaleProgramOpDescs,
+                                        new std::vector<OpDesc *>(stale_ops));
   }
   auto op_handles = ir::FilterByNodeWrapper<OpHandleBase>(*graph);
 

paddle/fluid/framework/inlined_stack.h

@@ -14,7 +14,6 @@
 
 #pragma once
 
-#include <array>
 #include <deque>
 #include "paddle/fluid/platform/enforce.h"
 
@@ -22,7 +21,7 @@ namespace paddle {
 namespace framework {
 
 template <typename T, size_t N>
-class SmallStack {
+class InlinedStack {
   static_assert(N > 0, "N must be larger than 0");
 
  public:
@@ -66,8 +65,8 @@ class SmallStack {
 
  private:
   T head_[N];
+  size_t size_{0};
   std::deque<T> tail_;
-  size_t size_;
 };
 
 }  // namespace framework

paddle/fluid/framework/ir/fuse_pass_base.h

@@ -14,6 +14,7 @@
 
 #pragma once
 
+#include <string>
 #include "paddle/fluid/framework/ir/graph.h"
 #include "paddle/fluid/framework/ir/pass.h"
 #include "paddle/fluid/framework/scope.h"
@@ -24,6 +25,10 @@ namespace ir {
 
 static const char kParamScopeAttr[] = "__param_scope__";
 static const char kFuseStatisAttr[] = "__fuse_statis__";
+// When we use trt or other third_party lib, the parameters are managed by
+// the lib, but not the fluid. So we need to record them to avoid duplicate
+// allocation.
+static const char kRepetitiveParamAttr[] = "__repetitive_param__";
 
 enum FuseOptions {
   DO_NOT_FUSE,  // fusing will not be done

paddle/fluid/framework/ir/graph_helper.cc

@@ -130,15 +130,21 @@ std::map<ir::Node *, std::unordered_set<ir::Node *>> BuildOperationAdjList(
     if (adj_list.find(n) == adj_list.end()) {
       adj_list[n] = std::unordered_set<ir::Node *>();
     }
+    std::vector<ir::Node *> nodes;
     for (auto &var : n->inputs) {
       for (auto &adj_n : var->inputs) {
         PADDLE_ENFORCE(adj_n->NodeType() == ir::Node::Type::kOperation);
         VLOG(4) << "adj " << adj_n->Name() << reinterpret_cast<void *>(adj_n)
                 << " -> " << n->Name() << reinterpret_cast<void *>(n)
                 << "  via " << var->Name() << reinterpret_cast<void *>(var);
-        adj_list[n].insert(adj_n);
+        nodes.push_back(adj_n);
       }
     }
+    std::sort(nodes.begin(), nodes.end(), [](ir::Node *node1, ir::Node *node2) {
+      return node1->id() > node2->id();
+    });
+    adj_list[n].insert(std::make_move_iterator(nodes.begin()),
+                       std::make_move_iterator(nodes.end()));
   }
   return adj_list;
 }

paddle/fluid/framework/operator.cc

@@ -467,12 +467,6 @@ const Variable* ExecutionContext::InputVar(const std::string& name) const {
   return it->second.empty() ? nullptr : it->second[0];
 }
 
-const Variable* ExecutionContext::LegacyInputVar(
-    const std::string& name) const {
-  auto ipt = op_.Input(name);
-  return ipt == kEmptyVarName ? nullptr : scope_.FindVar(ipt);
-}
-
 Variable* ExecutionContext::OutputVar(const std::string& name) const {
   auto it = ctx_.outputs.find(name);
   if (it == ctx_.outputs.end()) return nullptr;
@@ -483,22 +477,11 @@ Variable* ExecutionContext::OutputVar(const std::string& name) const {
   return it->second.empty() ? nullptr : it->second[0];
 }
 
-Variable* ExecutionContext::LegacyOutputVar(const std::string& name) const {
-  auto opt = op_.Output(name);
-  return opt == kEmptyVarName ? nullptr : scope_.FindVar(opt);
-}
-
 template <>
 const Tensor* ExecutionContext::Input<Tensor>(const std::string& name) const {
   return Input<LoDTensor>(name);
 }
 
-template <>
-const Tensor* ExecutionContext::LegacyInput<Tensor>(
-    const std::string& name) const {
-  return LegacyInput<LoDTensor>(name);
-}
-
 template <>
 const std::vector<const Tensor*> ExecutionContext::MultiInput<Tensor>(
     const std::string& name) const {
@@ -521,35 +504,11 @@ const std::vector<const Tensor*> ExecutionContext::MultiInput<Tensor>(
   return res;
 }
 
-template <>
-const std::vector<const Tensor*> ExecutionContext::LegacyMultiInput<Tensor>(
-    const std::string& name) const {
-  auto names = op().Inputs(name);
-  std::vector<const Tensor*> res;
-  res.reserve(names.size());
-  std::transform(names.begin(), names.end(), std::back_inserter(res),
-                 [&](const std::string& sub_name) -> const Tensor* {
-                   auto var = scope_.FindVar(sub_name);
-                   if (var == nullptr) return nullptr;
-                   PADDLE_ENFORCE(
-                       var->IsType<LoDTensor>(),
-                       "%s should be LoDTensor, but the received type is %s",
-                       sub_name, ToTypeName(var->Type()));
-                   return &(var->Get<LoDTensor>());
-                 });
-  return res;
-}
-
 template <>
 Tensor* ExecutionContext::Output<Tensor>(const std::string& name) const {
   return Output<LoDTensor>(name);
 }
 
-template <>
-Tensor* ExecutionContext::LegacyOutput<Tensor>(const std::string& name) const {
-  return LegacyOutput<LoDTensor>(name);
-}
-
 template <>
 std::vector<Tensor*> ExecutionContext::MultiOutput<Tensor>(
     const std::string& name) const {

paddle/fluid/framework/operator.h

@@ -16,9 +16,11 @@ limitations under the License. */
 
 #include <algorithm>
 #include <atomic>
+#include <memory>
 #include <string>
 #include <tuple>
 #include <unordered_map>
+#include <utility>
 #include <vector>
 
 #include "glog/logging.h"  // For VLOG
@@ -253,31 +255,6 @@ class ExecutionContext {
     return it->second;
   }
 
-  const std::vector<Variable*> LegacyMultiInputVar(
-      const std::string& name) const {
-    auto names = op_.Inputs(name);
-    std::vector<Variable*> res;
-    res.reserve(names.size());
-    std::transform(names.begin(), names.end(), std::back_inserter(res),
-                   [this](const std::string& name) {
-                     return name == kEmptyVarName ? nullptr
-                                                  : scope_.FindVar(name);
-                   });
-    return res;
-  }
-
-  std::vector<Variable*> LegacyMultiOutputVar(const std::string& name) const {
-    auto names = op_.Outputs(name);
-    std::vector<Variable*> res;
-    res.reserve(names.size());
-    std::transform(names.begin(), names.end(), std::back_inserter(res),
-                   [this](const std::string& name) {
-                     return name == kEmptyVarName ? nullptr
-                                                  : scope_.FindVar(name);
-                   });
-    return res;
-  }
-
   template <typename T>
   const T* Input(const std::string& name) const {
     auto* var = InputVar(name);
@@ -290,22 +267,6 @@ class ExecutionContext {
     return var == nullptr ? nullptr : var->GetMutable<T>();
   }
 
-  template <typename T>
-  const T* LegacyInput(const std::string& name) const {
-    auto* var = LegacyInputVar(name);
-    return var == nullptr ? nullptr : &var->Get<T>();
-  }
-
-  template <typename T>
-  T* LegacyOutput(const std::string& name) const {
-    auto var = LegacyOutputVar(name);
-    return var == nullptr ? nullptr : var->GetMutable<T>();
-  }
-
-  const Variable* LegacyInputVar(const std::string& name) const;
-
-  Variable* LegacyOutputVar(const std::string& name) const;
-
   template <typename T>
   const std::vector<const T*> MultiInput(const std::string& name) const {
     auto it = ctx_.inputs.find(name);
@@ -338,32 +299,6 @@ class ExecutionContext {
     return res;
   }
 
-  template <typename T>
-  const std::vector<const T*> LegacyMultiInput(const std::string& name) const {
-    auto names = op_.Inputs(name);
-    std::vector<const T*> res;
-    res.reserve(names.size());
-    std::transform(names.begin(), names.end(), std::back_inserter(res),
-                   [&](const std::string& sub_name) -> const T* {
-                     auto var = scope_.FindVar(sub_name);
-                     return var == nullptr ? nullptr : &var->Get<T>();
-                   });
-    return res;
-  }
-
-  template <typename T>
-  std::vector<T*> LegacyMultiOutput(const std::string& name) const {
-    auto names = op_.Outputs(name);
-    std::vector<T*> res;
-    res.reserve(names.size());
-    std::transform(names.begin(), names.end(), std::back_inserter(res),
-                   [&](const std::string& sub_name) -> T* {
-                     auto var = scope_.FindVar(sub_name);
-                     return var == nullptr ? nullptr : var->GetMutable<T>();
-                   });
-    return res;
-  }
-
   platform::Place GetPlace() const { return device_context_.GetPlace(); }
 
   template <typename DeviceContextType>
@@ -433,24 +368,13 @@ class ExecutionContext {
 template <>
 const Tensor* ExecutionContext::Input<Tensor>(const std::string& name) const;
 
-template <>
-const Tensor* ExecutionContext::LegacyInput<Tensor>(
-    const std::string& name) const;
-
 template <>
 const std::vector<const Tensor*> ExecutionContext::MultiInput<Tensor>(
     const std::string& name) const;
 
-template <>
-const std::vector<const Tensor*> ExecutionContext::LegacyMultiInput<Tensor>(
-    const std::string& name) const;
-
 template <>
 Tensor* ExecutionContext::Output<Tensor>(const std::string& name) const;
 
-template <>
-Tensor* ExecutionContext::LegacyOutput<Tensor>(const std::string& name) const;
-
 template <>
 std::vector<Tensor*> ExecutionContext::MultiOutput<Tensor>(
     const std::string& name) const;

paddle/fluid/inference/analysis/argument.h

@@ -23,8 +23,12 @@
 
 #pragma once
 
+#include <memory>
 #include <string>
+#include <unordered_map>
+#include <unordered_set>
 #include <vector>
+
 #include "paddle/fluid/framework/ir/graph.h"
 #include "paddle/fluid/framework/program_desc.h"
 #include "paddle/fluid/framework/scope.h"
@@ -133,6 +137,8 @@ struct Argument {
   DECL_ARGUMENT_FIELD(tensorrt_min_subgraph_size, TensorRtMinSubgraphSize, int);
   DECL_ARGUMENT_FIELD(tensorrt_precision_mode, TensorRtPrecisionMode,
                       AnalysisConfig::Precision);
+  DECL_ARGUMENT_FIELD(tensorrt_use_static_engine, TensorRtUseStaticEngine,
+                      bool);
 
   // Memory optimized related.
   DECL_ARGUMENT_FIELD(enable_memory_optim, EnableMemoryOptim, bool);

paddle/fluid/inference/analysis/helper.h

@@ -17,10 +17,12 @@ limitations under the License. */
 #include <sys/stat.h>
 #include <cstdio>
 #include <fstream>
+#include <memory>
 #include <set>
 #include <string>
 #include <typeindex>
 #include <unordered_map>
+#include <utility>
 #include <vector>
 
 #include "paddle/fluid/framework/framework.pb.h"
@@ -217,6 +219,35 @@ static std::string GetTrtCalibTableData(const std::string &model_opt_cache_dir,
   return "";
 }
 
+static std::string GetTrtEngineSerializedPath(const std::string &model_root,
+                                              const std::string &engine_key) {
+  return model_root + "/trt_serialized_" + engine_key;
+}
+
+static std::string GetTrtEngineSerializedData(
+    const std::string &model_opt_cache_dir, const std::string &engine_key) {
+  std::string trt_serialized_path =
+      GetTrtEngineSerializedPath(model_opt_cache_dir, engine_key);
+  if (FileExists(trt_serialized_path)) {
+    VLOG(3) << "Trt serialized file: " << trt_serialized_path
+            << "is found here";
+    std::ifstream infile(trt_serialized_path, std::ios::in);
+    std::stringstream buffer;
+    buffer << infile.rdbuf();
+    std::string trt_engine_serialized_data(buffer.str());
+    return trt_engine_serialized_data;
+  }
+  return "";
+}
+
+static void SaveTrtEngineSerializedDataToFile(
+    const std::string &trt_serialized_path,
+    const std::string &engine_serialized_data) {
+  std::ofstream outfile(trt_serialized_path);
+  outfile << engine_serialized_data;
+  outfile.close();
+}
+
 }  // namespace analysis
 }  // namespace inference
 }  // namespace paddle

paddle/fluid/inference/analysis/ir_pass_manager.cc

@@ -81,6 +81,9 @@ void IRPassManager::CreatePasses(Argument *argument,
       pass->Set(
           "model_opt_cache_dir",
           new std::string(GetOrCreateModelOptCacheDir(model_opt_cache_dir)));
+      pass->Set("gpu_device_id", new int(argument->gpu_device_id()));
+      pass->Set("use_static_engine",
+                new bool(argument->tensorrt_use_static_engine()));
     }
 
     pre_pass = pass_name;

paddle/fluid/inference/analysis/ir_pass_manager.h

@@ -22,7 +22,10 @@
 
 #pragma once
 
+#include <memory>
 #include <string>
+#include <unordered_set>
+#include <utility>
 #include <vector>
 #include "paddle/fluid/framework/ir/graph.h"
 #include "paddle/fluid/framework/ir/pass.h"

paddle/fluid/inference/analysis/ir_passes/tensorrt_subgraph_pass.h

@@ -13,7 +13,12 @@
 // limitations under the License.
 
 #pragma once
-#include <paddle/fluid/framework/ir/fuse_pass_base.h>
+#include <memory>
+#include <string>
+#include <unordered_map>
+#include <unordered_set>
+#include <vector>
+#include "paddle/fluid/framework/ir/fuse_pass_base.h"
 #include "paddle/fluid/framework/ir/pass.h"
 
 namespace paddle {
@@ -26,8 +31,9 @@ class TensorRtSubgraphPass : public framework::ir::FusePassBase {
       std::unique_ptr<framework::ir::Graph> graph) const override;
 
  private:
-  void CreateTensorRTOp(framework::ir::Node *x,
-                        framework::ir::Graph *graph) const;
+  void CreateTensorRTOp(framework::ir::Node *x, framework::ir::Graph *graph,
+                        const std::vector<std::string> &graph_params,
+                        std::vector<std::string> *repetitive_params) const;
   void CleanIntermediateOutputs(framework::ir::Node *node);
 };
 

paddle/fluid/inference/analysis/passes/ir_params_sync_among_devices_pass.cc

@@ -31,6 +31,13 @@ void IrParamsSyncAmongDevicesPass::RunImpl(Argument *argument) {
   // The parameters are on the cpu, therefore, synchronization is not necessary.
   if (!argument->use_gpu()) return;
 
+  auto &graph = argument->main_graph();
+  std::vector<std::string> repetitive_params;
+
+  if (graph.Has(framework::ir::kRepetitiveParamAttr))
+    repetitive_params = graph.Get<std::vector<std::string>>(
+        framework::ir::kRepetitiveParamAttr);
+
   LOG(INFO) << "Sync params from CPU to GPU";
 
   PADDLE_ENFORCE(argument->gpu_device_id_valid());
@@ -43,6 +50,10 @@ void IrParamsSyncAmongDevicesPass::RunImpl(Argument *argument) {
   // Because there exists the case that new parameter variables are not added to
   // the program in the analysis pass.
   for (auto &var_name : all_vars) {
+    if (std::count(repetitive_params.begin(), repetitive_params.end(),
+                   var_name)) {
+      continue;
+    }
     auto *var = scope->FindLocalVar(var_name);
     PADDLE_ENFORCE(var != nullptr);
     if (var->IsType<framework::LoDTensor>() ||

paddle/fluid/inference/analysis/passes/ir_params_sync_among_devices_pass.h

@@ -17,6 +17,7 @@
 #include <string>
 #include <vector>
 
+#include "paddle/fluid/framework/ir/fuse_pass_base.h"
 #include "paddle/fluid/framework/scope.h"
 #include "paddle/fluid/inference/analysis/analysis_pass.h"
 #include "paddle/fluid/platform/place.h"

paddle/fluid/inference/api/analysis_config.cc

@@ -103,6 +103,7 @@ AnalysisConfig::AnalysisConfig(const AnalysisConfig &other) {
   CP_MEMBER(tensorrt_max_batchsize_);
   CP_MEMBER(tensorrt_min_subgraph_size_);
   CP_MEMBER(tensorrt_precision_mode_);
+  CP_MEMBER(trt_use_static_engine_);
   // MKLDNN related.
   CP_MEMBER(use_mkldnn_);
   CP_MEMBER(mkldnn_enabled_op_types_);
@@ -144,7 +145,7 @@ void AnalysisConfig::EnableMKLDNN() {
 
 void AnalysisConfig::EnableTensorRtEngine(
     int workspace_size, int max_batch_size, int min_subgraph_size,
-    AnalysisConfig::Precision precision_mode) {
+    AnalysisConfig::Precision precision_mode, bool use_static) {
 #ifdef PADDLE_WITH_CUDA
   if (!use_gpu()) {
     LOG(ERROR) << "To use TensorRT engine, please call EnableGpu() first";
@@ -156,6 +157,7 @@ void AnalysisConfig::EnableTensorRtEngine(
   tensorrt_max_batchsize_ = max_batch_size;
   tensorrt_min_subgraph_size_ = min_subgraph_size;
   tensorrt_precision_mode_ = precision_mode;
+  trt_use_static_engine_ = use_static;
 
   Update();
 #else

paddle/fluid/inference/api/analysis_predictor.cc

@@ -183,6 +183,9 @@ void AnalysisPredictor::SetMkldnnThreadID(int tid) {
 bool AnalysisPredictor::Run(const std::vector<PaddleTensor> &inputs,
                             std::vector<PaddleTensor> *output_data,
                             int batch_size) {
+  if (UNLIKELY(config_.cpu_math_library_num_threads() > 1)) {
+    paddle::platform::SetNumThreads(config_.cpu_math_library_num_threads());
+  }
   VLOG(3) << "Predictor::predict";
   inference::Timer timer;
   timer.tic();
@@ -362,6 +365,7 @@ void AnalysisPredictor::OptimizeInferenceProgram() {
     argument_.SetTensorRtMaxBatchSize(config_.tensorrt_max_batchsize_);
     argument_.SetTensorRtMinSubgraphSize(config_.tensorrt_min_subgraph_size_);
     argument_.SetTensorRtPrecisionMode(config_.tensorrt_precision_mode_);
+    argument_.SetTensorRtUseStaticEngine(config_.trt_use_static_engine_);
   }
 
   if (config_.use_mkldnn_) {
@@ -435,12 +439,14 @@ void AnalysisPredictor::PrepareFeedFetch() {
       }
       feeds_[idx] = op;
       feed_names_[op->Output("Out")[0]] = idx;
+      idx2feeds_[idx] = op->Output("Out")[0];
     } else if (op->Type() == "fetch") {
       int idx = boost::get<int>(op->GetAttr("col"));
       if (fetches_.size() <= static_cast<size_t>(idx)) {
         fetches_.resize(idx + 1);
       }
       fetches_[idx] = op;
+      idx2fetches_[idx] = op->Input("X")[0];
     }
   }
 }
@@ -453,6 +459,22 @@ void AnalysisPredictor::CreateFeedFetchVar(framework::Scope *scope) {
   var->GetMutable<framework::FeedFetchList>();
 }
 
+std::vector<std::string> AnalysisPredictor::GetInputNames() {
+  std::vector<std::string> input_names;
+  for (auto &item : idx2feeds_) {
+    input_names.push_back(item.second);
+  }
+  return input_names;
+}
+
+std::vector<std::string> AnalysisPredictor::GetOutputNames() {
+  std::vector<std::string> output_names;
+  for (auto &item : idx2fetches_) {
+    output_names.push_back(item.second);
+  }
+  return output_names;
+}
+
 std::unique_ptr<ZeroCopyTensor> AnalysisPredictor::GetInputTensor(
     const std::string &name) {
   PADDLE_ENFORCE(executor_->scope()->FindVar(name), "no name called %s", name);
@@ -460,6 +482,13 @@ std::unique_ptr<ZeroCopyTensor> AnalysisPredictor::GetInputTensor(
       new ZeroCopyTensor(static_cast<void *>(executor_->scope())));
   res->input_or_output_ = true;
   res->SetName(name);
+  if (platform::is_cpu_place(place_)) {
+    res->SetPlace(PaddlePlace::kCPU);
+  } else {
+    auto gpu_place = boost::get<platform::CUDAPlace>(place_);
+    res->SetPlace(PaddlePlace::kGPU, gpu_place.GetDeviceId());
+  }
+
   return res;
 }
 
@@ -470,6 +499,12 @@ std::unique_ptr<ZeroCopyTensor> AnalysisPredictor::GetOutputTensor(
       new ZeroCopyTensor(static_cast<void *>(executor_->scope())));
   res->input_or_output_ = false;
   res->SetName(name);
+  if (platform::is_cpu_place(place_)) {
+    res->SetPlace(PaddlePlace::kCPU);
+  } else {
+    auto gpu_place = boost::get<platform::CUDAPlace>(place_);
+    res->SetPlace(PaddlePlace::kGPU, gpu_place.GetDeviceId());
+  }
   return res;
 }
 

paddle/fluid/inference/api/analysis_predictor.h

@@ -15,12 +15,14 @@
 #pragma once
 #include <algorithm>
 #include <map>
+#include <memory>
 #include <string>
 #include <vector>
 #include "paddle/fluid/framework/naive_executor.h"
 #include "paddle/fluid/inference/analysis/analyzer.h"
 #include "paddle/fluid/inference/api/api_impl.h"
 #include "paddle/fluid/inference/api/details/reset_tensor_array.h"
+#include "paddle/fluid/inference/api/helper.h"
 #include "paddle/fluid/inference/api/paddle_inference_api.h"
 #include "paddle/fluid/string/printf.h"
 #ifdef PADDLE_WITH_TESTING
@@ -53,6 +55,9 @@ class AnalysisPredictor : public PaddlePredictor {
            std::vector<PaddleTensor> *output_data,
            int batch_size = -1) override;
 
+  std::vector<std::string> GetInputNames();
+  std::vector<std::string> GetOutputNames();
+
   std::unique_ptr<ZeroCopyTensor> GetInputTensor(
       const std::string &name) override;
   std::unique_ptr<ZeroCopyTensor> GetOutputTensor(
@@ -131,7 +136,11 @@ class AnalysisPredictor : public PaddlePredictor {
   std::shared_ptr<framework::ProgramDesc> inference_program_;
   std::vector<framework::OpDesc *> feeds_;
   std::map<std::string, size_t> feed_names_;
+  // Sorted according to the idx.
+  std::map<size_t, std::string> idx2feeds_;
   std::vector<framework::OpDesc *> fetches_;
+  std::map<size_t, std::string> idx2fetches_;
+
   // Memory buffer for feed inputs. The temporary LoDTensor will cause serious
   // concurrency problems, wrong results and memory leak, so cache them.
   std::vector<framework::LoDTensor> feed_tensors_;

paddle/fluid/inference/api/api_impl.cc

@@ -131,6 +131,9 @@ NativePaddlePredictor::~NativePaddlePredictor() {
 bool NativePaddlePredictor::Run(const std::vector<PaddleTensor> &inputs,
                                 std::vector<PaddleTensor> *output_data,
                                 int batch_size) {
+  if (UNLIKELY(config_.cpu_math_library_num_threads() > 1)) {
+    paddle::platform::SetNumThreads(config_.cpu_math_library_num_threads());
+  }
   VLOG(3) << "Predictor::predict";
   Timer timer;
   timer.tic();

paddle/fluid/inference/api/details/zero_copy_tensor.cc

@@ -15,6 +15,7 @@
 #include "paddle/fluid/framework/lod_tensor.h"
 #include "paddle/fluid/framework/scope.h"
 #include "paddle/fluid/inference/api/paddle_inference_api.h"
+#include "paddle/fluid/memory/memcpy.h"
 #include "paddle/fluid/platform/enforce.h"
 
 namespace paddle {
@@ -73,6 +74,61 @@ T *ZeroCopyTensor::data(PaddlePlace *place, int *size) const {
   return res;
 }
 
+template <typename T>
+void ZeroCopyTensor::copy_from_cpu(const T *data) {
+  EAGER_GET_TENSOR;
+  PADDLE_ENFORCE_GE(
+      tensor->numel(), 0,
+      "You should call ZeroCopyTensor::Reshape(const std::vector<int> &shape)"
+      "function before copy data from cpu.");
+  size_t ele_size = tensor->numel() * sizeof(T);
+
+  if (place_ == PaddlePlace::kCPU) {
+    auto *t_data = tensor->mutable_data<T>(platform::CPUPlace());
+    std::memcpy(static_cast<void *>(t_data), data, ele_size);
+  } else {
+#ifdef PADDLE_WITH_CUDA
+    platform::DeviceContextPool &pool = platform::DeviceContextPool::Instance();
+    platform::CUDAPlace gpu_place(device_);
+    auto *t_data = tensor->mutable_data<T>(gpu_place);
+    auto *dev_ctx =
+        static_cast<const platform::CUDADeviceContext *>(pool.Get(gpu_place));
+
+    memory::Copy(gpu_place, static_cast<void *>(t_data), platform::CPUPlace(),
+                 data, ele_size, dev_ctx->stream());
+#else
+    PADDLE_THROW("Not compile with CUDA, should not reach here.");
+#endif
+  }
+}
+
+template <typename T>
+void ZeroCopyTensor::copy_to_cpu(T *data) {
+  EAGER_GET_TENSOR;
+  auto ele_num = tensor->numel();
+  auto *t_data = tensor->data<T>();
+  auto t_place = tensor->place();
+
+  if (platform::is_cpu_place(t_place)) {
+    std::memcpy(static_cast<void *>(data), t_data, ele_num * sizeof(T));
+  } else {
+#ifdef PADDLE_WITH_CUDA
+    platform::DeviceContextPool &pool = platform::DeviceContextPool::Instance();
+    auto gpu_place = boost::get<platform::CUDAPlace>(t_place);
+    auto *dev_ctx =
+        static_cast<const platform::CUDADeviceContext *>(pool.Get(gpu_place));
+    memory::Copy(platform::CPUPlace(), static_cast<void *>(data), gpu_place,
+                 t_data, ele_num * sizeof(T), dev_ctx->stream());
+#else
+    PADDLE_THROW("Not compile with CUDA, should not reach here.");
+#endif
+  }
+}
+template void ZeroCopyTensor::copy_from_cpu<float>(const float *data);
+template void ZeroCopyTensor::copy_from_cpu<int64_t>(const int64_t *data);
+template void ZeroCopyTensor::copy_to_cpu<float>(float *data);
+template void ZeroCopyTensor::copy_to_cpu<int64_t>(int64_t *data);
+
 template float *ZeroCopyTensor::data<float>(PaddlePlace *place,
                                             int *size) const;
 template int64_t *ZeroCopyTensor::data<int64_t>(PaddlePlace *place,
@@ -92,10 +148,10 @@ void *ZeroCopyTensor::FindTensor() const {
   return tensor;
 }
 
-std::vector<int64_t> ZeroCopyTensor::shape() const {
+std::vector<int> ZeroCopyTensor::shape() const {
   EAGER_GET_TENSOR;
   PADDLE_ENFORCE(tensor_, "not found tensor called %s in the scope", name_);
-  return framework::vectorize(tensor->dims());
+  return framework::vectorize2int(tensor->dims());
 }
 
 void ZeroCopyTensor::SetLoD(const std::vector<std::vector<size_t>> &x) {

paddle/fluid/inference/api/details/zero_copy_tensor_dummy.cc

@@ -37,7 +37,7 @@ template int64_t *ZeroCopyTensor::mutable_data(PaddlePlace place);
 
 void *ZeroCopyTensor::FindTensor() const { return nullptr; }
 
-std::vector<int64_t> ZeroCopyTensor::shape() const { return {}; }
+std::vector<int> ZeroCopyTensor::shape() const { return {}; }
 
 void ZeroCopyTensor::SetLoD(const std::vector<std::vector<size_t>> &x) {}
 

paddle/fluid/inference/api/helper.h

@@ -50,6 +50,11 @@ class Timer {
   }
 };
 
+static int GetUniqueId() {
+  static int id = 0;
+  return id++;
+}
+
 static void split(const std::string &str, char sep,
                   std::vector<std::string> *pieces) {
   pieces->clear();

paddle/fluid/inference/api/paddle_analysis_config.h

@@ -135,7 +135,8 @@ struct AnalysisConfig {
    */
   void EnableTensorRtEngine(int workspace_size = 1 << 20,
                             int max_batch_size = 1, int min_subgraph_size = 3,
-                            Precision precision = Precision::kFloat32);
+                            Precision precision = Precision::kFloat32,
+                            bool use_static = true);
   /** A boolean state telling whether the TensorRT engine is used.
    */
   bool tensorrt_engine_enabled() const { return use_tensorrt_; }
@@ -233,6 +234,7 @@ struct AnalysisConfig {
   //  subgraph, 3 as default value.
   int tensorrt_min_subgraph_size_{3};
   Precision tensorrt_precision_mode_;
+  bool trt_use_static_engine_;
 
   // memory reuse related.
   bool enable_memory_optim_{false};