Merge branch 'feature/dynamic-recurrent-op' into feature/dynamic-recu…

…rrent-op-forward-test
Superjomn · Oct 9, 2017 · 4a0cc85 · 4a0cc85
2 parents 50c364e + d30ada2
commit 4a0cc85
Show file tree

Hide file tree

Showing 6 changed files with 44 additions and 12 deletions.
diff --git a/cmake/configure.cmake b/cmake/configure.cmake
@@ -24,6 +24,10 @@ if(WITH_DOUBLE)
     add_definitions(-DPADDLE_TYPE_DOUBLE)
 endif(WITH_DOUBLE)
 
+if(WITH_TESTING)
+    add_definitions(-DPADDLE_WITH_TESTING)
+endif(WITH_TESTING)
+
 if(NOT WITH_TIMER)
     add_definitions(-DPADDLE_DISABLE_TIMER)
 endif(NOT WITH_TIMER)

diff --git a/paddle/operators/dynamic_recurrent_op.cc b/paddle/operators/dynamic_recurrent_op.cc
@@ -30,11 +30,11 @@ inline void CreateVariables(Scope& scope,
 
 }  // namespace detail
 
-class DynamicRecurrentAlgorithmProtoAndCheckerMaker
+class DynamicRecurrentOpProtoAndCheckerMaker
     : public framework::OpProtoAndCheckerMaker {
  public:
-  DynamicRecurrentAlgorithmProtoAndCheckerMaker(
-      framework::OpProto* proto, framework::OpAttrChecker* op_checker)
+  DynamicRecurrentOpProtoAndCheckerMaker(framework::OpProto* proto,
+                                         framework::OpAttrChecker* op_checker)
       : OpProtoAndCheckerMaker(proto, op_checker) {
     const auto& name = DynamicRecurrentOp::kArgName;
     // inputs and outputs stored in proto
@@ -268,5 +268,5 @@ void DynamicRecurrentGradientOp::Run(
 }  // namespace paddle
 
 REGISTER_OP_WITHOUT_GRADIENT(
-    dynamic_recurrent_op, paddle::operators::DynamicRecurrentOp,
-    paddle::operators::DynamicRecurrentAlgorithmProtoAndCheckerMaker);
+    dynamic_recurrent, paddle::operators::DynamicRecurrentOp,
+    paddle::operators::DynamicRecurrentOpProtoAndCheckerMaker);
diff --git a/paddle/operators/dynamic_recurrent_op.h b/paddle/operators/dynamic_recurrent_op.h
@@ -14,7 +14,9 @@
 
 #pragma once
 
+#ifdef PADDLE_WITH_TESTING
 #include "gtest/gtest.h"
+#endif
 
 #include "paddle/framework/lod_tensor.h"
 #include "paddle/framework/operator.h"
@@ -142,6 +144,7 @@ class DynamicRecurrentOp : public framework::OperatorBase {
   mutable rnn::Argument arg_;
   mutable ArgCache cache_;
 
+#ifdef PADDLE_WITH_TESTING
   friend class DynamicRecurrentOpTestHelper;
   FRIEND_TEST(DynamicRecurrentOpTestHelper, SplitInputs);
   FRIEND_TEST(DynamicRecurrentOpTestHelper, CreateCache);
@@ -150,6 +153,7 @@ class DynamicRecurrentOp : public framework::OperatorBase {
   FRIEND_TEST(DynamicRecurrentOpTestHelper, WriteStepOutputs);
   FRIEND_TEST(DynamicRecurrentOpTestHelper, InitStates);
   FRIEND_TEST(DynamicRecurrentOpTestHelper, ConcatOutputs);
+#endif
 };
 
 class DynamicRecurrentGradientOp : public framework::OperatorBase {

diff --git a/paddle/operators/dynamic_recurrent_op_test.cc b/paddle/operators/dynamic_recurrent_op_test.cc
@@ -86,7 +86,10 @@ class DynamicRecurrentOpTestHelper : public ::testing::Test {
     CreateVar(scope, "out0", framework::make_ddim({10, 20}), place);
     auto* in0 = CreateVar(scope, "in0", framework::make_ddim({10, 8}), place);
     // 10 instanes with 4 sentences, length is 4, 3, 2, 1 respectively.
-    framework::LoD in0_lod({{0, 4, 7, 9, 10}});
+    framework::LoD in0_lod(1);
+    for (int x : std::vector<int>{0, 4, 7, 9, 10}) {
+      in0_lod[0].push_back(x);
+    }
     in0->set_lod(in0_lod);
     in0->Resize(framework::make_ddim({10, 8}));
     // set the content, each sentence content is seqid.batchid

diff --git a/paddle/pybind/pybind.cc b/paddle/pybind/pybind.cc
@@ -376,14 +376,13 @@ All parameter, weight, gradient are variables in Paddle.
                -> void { self.SetStepNet(net.Clone()); })
       .def("get_state",
            [](operators::DynamicRecurrentOp &self, const std::string &name)
-               -> TensorArray & { return self.state(name); })
+               -> const TensorArray & { return self.state(name); })
       .def("get_step_input",
            [](operators::DynamicRecurrentOp &self, const std::string &name)
-               -> TensorArray & { return self.step_input(name); })
-      .def("get_step_output", [](operators::DynamicRecurrentOp &self,
-                                 const std::string &name) -> TensorArray & {
-        return self.step_output(name);
-      });
+               -> const TensorArray & { return self.step_input(name); })
+      .def("get_step_output",
+           [](operators::DynamicRecurrentOp &self, const std::string &name)
+               -> const TensorArray & { return self.step_output(name); });
 
   // cond_op
   py::class_<operators::CondOp, OperatorBase>(m, "CondOp")

diff --git a/python/paddle/v2/framework/op.py b/python/paddle/v2/framework/op.py
@@ -219,6 +219,27 @@ def __call__(self, *args, **kwargs):
         return core.RecurrentOp.create(proto.SerializeToString())
 
 
+class __DynamicRecurrentOp__(object):
+    __proto__ = None
+    type = "dynamic_recurrent"
+
+    def __init__(self):
+        # cache recurrent_op's proto
+        if self.__proto__ is None:
+            for op_proto in get_all_op_protos():
+                if op_proto.type == self.type:
+                    self.__proto__ = op_proto
+
+    def __call__(self, *args, **kwargs):
+        if self.type not in args and "type" not in kwargs:
+            kwargs["type"] = self.type
+        # create proto
+        create_method = OpDescCreationMethod(self.__proto__)
+        proto = create_method(*args, **kwargs)
+        # create rnnop
+        return core.DynamicRecurrentOp.create(proto.SerializeToString())
+
+
 class __CondOp__(object):
     __proto__ = None
     type = "cond"
@@ -242,4 +263,5 @@ def __call__(self, *args, **kwargs):
 
 Operator = OperatorFactory()  # The default global factory
 RecurrentOp = __RecurrentOp__()
+DynamicRecurrentOp = __DynamicRecurrentOp__()
 CondOp = __CondOp__()