backward_test.cc 14.2 KB
Newer Older
Y
Yu Yang 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14
/* Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserve.

   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. */

Y
Yu Yang 已提交
15
#include "paddle/framework/backward.h"
D
dongzhihong 已提交
16

Y
Yu Yang 已提交
17
#include <gtest/gtest.h>
Y
Yu Yang 已提交
18
#include "paddle/framework/net.h"
Y
Yu Yang 已提交
19
#include "paddle/framework/op_registry.h"
Y
Yu Yang 已提交
20

Y
Yu Yang 已提交
21 22 23 24 25 26 27 28 29 30
namespace paddle {
namespace framework {

class EmptyOp : public OperatorBase {
 public:
  void InferShape(const std::shared_ptr<Scope> &scope) const override {}
  void Run(const std::shared_ptr<Scope> &scope,
           const platform::DeviceContext &dev_ctx) const override {}
};

Y
Yu Yang 已提交
31
class RowWiseAddOpMaker : public OpProtoAndCheckerMaker {
Y
Yu Yang 已提交
32
 public:
Y
Yu Yang 已提交
33
  RowWiseAddOpMaker(OpProto *proto, OpAttrChecker *op_checker)
Y
Yu Yang 已提交
34 35 36 37 38 39 40 41
      : OpProtoAndCheckerMaker(proto, op_checker) {
    AddInput("X", "Input X of Add").IgnoreGradient();
    AddInput("b", "Bias of Add").IgnoreGradient();
    AddOutput("Out", "Out of Add").IgnoreGradient();
    AddComment("Add Op");
  }
};

Y
Yu Yang 已提交
42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62
class MulOpMaker : public OpProtoAndCheckerMaker {
 public:
  MulOpMaker(OpProto *proto, OpAttrChecker *op_checker)
      : OpProtoAndCheckerMaker(proto, op_checker) {
    AddInput("A", "A");
    AddInput("B", "B");
    AddOutput("Out", "Out");
    AddComment("Mul");
  }
};

class SigmoidOpMaker : public OpProtoAndCheckerMaker {
 public:
  SigmoidOpMaker(OpProto *proto, OpAttrChecker *op_checker)
      : OpProtoAndCheckerMaker(proto, op_checker) {
    AddInput("X", "X");
    AddOutput("Y", "Y");
    AddComment("Sigmoid");
  }
};

D
dongzhihong 已提交
63 64 65 66 67 68 69 70 71 72
class NoGradOpMaker : public OpProtoAndCheckerMaker {
 public:
  NoGradOpMaker(OpProto *proto, OpAttrChecker *op_checker)
      : OpProtoAndCheckerMaker(proto, op_checker) {
    AddInput("X", "X input");
    AddOutput("Y", "Y output");
    AddComment("NoGradOp, same input output. no Grad");
  }
};

Y
Yu Yang 已提交
73 74 75 76
class FcOp : public NetOp {
 public:
  void Init() override {
    AddOp(OpRegistry::CreateOp("mul", {Input("X"), Input("W")},
Y
Yu Yang 已提交
77
                               {Output("mul_result")}, {}));
Y
Yu Yang 已提交
78
    auto b_name = Input("b");
Y
Yu Yang 已提交
79
    std::string before_act = "mul_result";
Y
Yu Yang 已提交
80
    if (b_name != EMPTY_VAR_NAME()) {
Y
Yu Yang 已提交
81 82 83 84 85 86 87 88
      AddOp(OpRegistry::CreateOp("rowwise_add", {Output("mul_result"), b_name},
                                 {Output("add_result")}, {}));
      before_act = "add_result";
    } else {
      auto out_varname = Output("add_result");
      if (out_varname != EMPTY_VAR_NAME()) {
        this->Rename(out_varname, EMPTY_VAR_NAME());
      }
Y
Yu Yang 已提交
89
    }
Y
Yu Yang 已提交
90 91 92

    AddOp(OpRegistry::CreateOp("sigmoid", {Output(before_act)}, {Output("Out")},
                               {}));
Y
Yu Yang 已提交
93 94 95 96 97 98 99 100 101 102 103
    CompleteAddOp(false);
  }
};

class FcOpMaker : public OpProtoAndCheckerMaker {
 public:
  FcOpMaker(OpProto *proto, OpAttrChecker *op_checker)
      : OpProtoAndCheckerMaker(proto, op_checker) {
    AddInput("X", "x");
    AddInput("W", "w");
    AddInput("b", "b");
Y
Yu Yang 已提交
104 105
    AddOutput("mul_result", "").SetTemporary();
    AddOutput("add_result", "").SetTemporary();
Y
Yu Yang 已提交
106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130
    AddOutput("Out", "");
    AddComment("");
  }
};

class ManyOutputOpMaker : public OpProtoAndCheckerMaker {
 public:
  ManyOutputOpMaker(OpProto *proto, OpAttrChecker *op_checker)
      : OpProtoAndCheckerMaker(proto, op_checker) {
    AddInput("x", "x");
    AddOutput("y", "y");
    AddOutput("z", "z");
    AddComment("");
  }
};

class FillZeroOpMaker : public OpProtoAndCheckerMaker {
 public:
  FillZeroOpMaker(OpProto *proto, OpAttrChecker *op_checker)
      : OpProtoAndCheckerMaker(proto, op_checker) {
    AddInput("x", "x");
    AddOutput("out", "out");
    AddComment("");
  }
};
Y
Yu Yang 已提交
131 132 133 134 135 136 137 138 139 140

class AddOpMaker : public OpProtoAndCheckerMaker {
 public:
  AddOpMaker(OpProto *proto, OpAttrChecker *op_checker)
      : OpProtoAndCheckerMaker(proto, op_checker) {
    AddInput("X", "x").SetMultiple();
    AddOutput("Y", "y");
    AddComment("");
  }
};
Y
Yu Yang 已提交
141 142 143 144
}  // namespace framework
}  // namespace paddle

namespace f = paddle::framework;
Y
Yu Yang 已提交
145 146 147 148 149 150 151
using EnforceNotMet = paddle::platform::EnforceNotMet;
REGISTER_OP(rowwise_add, f::EmptyOp, f::RowWiseAddOpMaker);
REGISTER_GRADIENT_OP(rowwise_add, rowwise_add_grad, f::EmptyOp);
REGISTER_OP(mul, f::EmptyOp, f::MulOpMaker);
REGISTER_GRADIENT_OP(mul, mul_grad, f::EmptyOp);
REGISTER_OP(sigmoid, f::EmptyOp, f::SigmoidOpMaker);
REGISTER_GRADIENT_OP(sigmoid, sigmoid_grad, f::EmptyOp);
D
dongzhihong 已提交
152
REGISTER_OP(nograd, f::EmptyOp, f::NoGradOpMaker);
Y
Yu Yang 已提交
153
REGISTER_OP(fill_zeros_like, f::EmptyOp, f::FillZeroOpMaker);
Y
Yu Yang 已提交
154 155
REGISTER_OP(add, f::EmptyOp, f::AddOpMaker);
REGISTER_GRADIENT_OP(add, add_grad, f::EmptyOp);
D
dongzhihong 已提交
156 157 158
REGISTER_OP(fc, f::FcOp, f::FcOpMaker);
REGISTER_OP(many_output_op, f::EmptyOp, f::ManyOutputOpMaker);
REGISTER_GRADIENT_OP(many_output_op, many_output_op_grad, f::EmptyOp);
Y
Yu Yang 已提交
159

Y
Yu Yang 已提交
160
TEST(Backward, simple_op_grad) {
Y
Yu Yang 已提交
161 162
  auto fwd = f::OpRegistry::CreateOp("rowwise_add", {"X", "b"}, {"Out"}, {});
  ASSERT_NE(fwd, nullptr);
Y
Yu Yang 已提交
163
  auto gop = f::OpRegistry::CreateGradOp(*fwd);
164
  ASSERT_EQ(1UL, gop->inputs_.size());
Y
Yu Yang 已提交
165 166 167 168 169
  ASSERT_EQ("Out" + f::OperatorBase::GRAD_VAR_SUFFIX(), gop->inputs_[0]);
  ASSERT_EQ("rowwise_add_grad", gop->type_);
  ASSERT_EQ("X" + f::OperatorBase::GRAD_VAR_SUFFIX(), gop->outputs_[0]);
  ASSERT_EQ("b" + f::OperatorBase::GRAD_VAR_SUFFIX(), gop->outputs_[1]);

Y
Stash  
Yu Yang 已提交
170 171
  ASSERT_EQ("X" + f::OperatorBase::GRAD_VAR_SUFFIX(),
            gop->Output("X" + f::OperatorBase::GRAD_VAR_SUFFIX()));
Y
Yu Yang 已提交
172 173
}

D
dongzhihong 已提交
174
TEST(Backward, simple_op_not_need_grad) {
D
dongzhihong 已提交
175
  auto fwd = f::OpRegistry::CreateOp("rowwise_add", {"X", "b"}, {"Out"}, {});
D
dongzhihong 已提交
176
  ASSERT_NE(fwd, nullptr);
D
dongzhihong 已提交
177
  auto gop = f::Backward(*fwd, {"X"});
D
dongzhihong 已提交
178 179
  ASSERT_EQ(std::find(gop->outputs_.begin(), gop->outputs_.end(),
                      "X" + f::OperatorBase::GRAD_VAR_SUFFIX()),
D
dongzhihong 已提交
180
            gop->outputs_.end());
D
dongzhihong 已提交
181

D
dongzhihong 已提交
182 183
  auto no_input_gop = f::Backward(*fwd, {"X", "b"});
  ASSERT_NE(no_input_gop, nullptr);
Y
Yu Yang 已提交
184 185
  ASSERT_TRUE(no_input_gop->IsNetOp());
  ASSERT_EQ(0UL, std::static_pointer_cast<f::NetOp>(no_input_gop)->ops_.size());
D
dongzhihong 已提交
186 187
}

Y
Yu Yang 已提交
188
TEST(Backward, net_fc_backward_normal) {
Y
Yu Yang 已提交
189
  std::shared_ptr<f::OperatorBase> fwd = f::OpRegistry::CreateOp(
Y
Yu Yang 已提交
190
      "fc", {"X", "w", "b"}, {"mul_result", "add_result", "out"}, {});
Y
Stash  
Yu Yang 已提交
191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209
  ASSERT_NE(fwd, nullptr);
  std::shared_ptr<f::OperatorBase> gop = f::Backward(*fwd, {});
  ASSERT_TRUE(gop->IsNetOp());
  auto net = static_cast<f::NetOp *>(gop.get());

  ASSERT_NO_THROW(net->DebugString());

  ASSERT_EQ(3UL, net->ops_.size());

  f::OperatorBase &d_sigmoid = *net->ops_[0];
  ASSERT_EQ("sigmoid_grad", d_sigmoid.type_);

  f::OperatorBase &d_add = *net->ops_[1];
  ASSERT_EQ("rowwise_add_grad", d_add.type_);

  f::OperatorBase &d_mul = *net->ops_[2];
  ASSERT_EQ("mul_grad", d_mul.type_);
}

Y
Yu Yang 已提交
210
TEST(Backward, net_fc_backward_not_have_b) {
Y
Stash  
Yu Yang 已提交
211
  std::shared_ptr<f::OperatorBase> fwd = f::OpRegistry::CreateOp(
Y
Yu Yang 已提交
212
      "fc", {"X", "w", f::OperatorBase::EMPTY_VAR_NAME()},
Y
Yu Yang 已提交
213
      {"mul_result", "add_result", "tmp"}, {});
Y
Stash  
Yu Yang 已提交
214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229
  ASSERT_NE(fwd, nullptr);
  std::shared_ptr<f::OperatorBase> gop = f::Backward(*fwd, {});
  ASSERT_TRUE(gop->IsNetOp());
  auto net = static_cast<f::NetOp *>(gop.get());

  ASSERT_NO_THROW(net->DebugString());

  ASSERT_EQ(2UL, net->ops_.size());

  f::OperatorBase &d_sigmoid = *net->ops_[0];
  ASSERT_EQ("sigmoid_grad", d_sigmoid.type_);

  f::OperatorBase &d_mul = *net->ops_[1];
  ASSERT_EQ("mul_grad", d_mul.type_);
}

Y
Yu Yang 已提交
230
TEST(Backward, net_input_of_network_not_need_grad) {
Y
Stash  
Yu Yang 已提交
231
  f::NetOp net;
Y
Yu Yang 已提交
232
  net.AddOp(f::OpRegistry::CreateOp("fc", {"X", "W1", "b1"},
Y
Yu Yang 已提交
233
                                    {"mul_tmp_0", "add_tmp_0", "hidden0"}, {}));
Y
Yu Yang 已提交
234
  net.AddOp(f::OpRegistry::CreateOp("fc", {"hidden0", "W2", "b2"},
Y
Yu Yang 已提交
235
                                    {"mul_tmp_1", "add_tmp_1", "hidden1"}, {}));
Y
Yu Yang 已提交
236
  net.CompleteAddOp();
Y
Stash  
Yu Yang 已提交
237 238 239 240 241 242 243 244 245 246 247 248
  auto bwd = Backward(net, {"X"});  // X@GRAD is not need.
  ASSERT_TRUE(bwd->IsNetOp());
  auto bwd_net = static_cast<f::NetOp *>(bwd.get());

  std::unordered_set<std::string> all_output = std::unordered_set<std::string>(
      bwd_net->outputs_.begin(), bwd_net->outputs_.end());
  all_output.erase(f::OperatorBase::EMPTY_VAR_NAME());

  for (auto &out : {"W1", "b1", "hidden0", "W2", "b2"}) {
    ASSERT_NE(all_output.find(out + f::OperatorBase::GRAD_VAR_SUFFIX()),
              all_output.end());
  }
Y
Yu Yang 已提交
249 250 251 252 253

  // Not Generated X
  ASSERT_EQ(all_output.find("X" + f::OperatorBase::GRAD_VAR_SUFFIX()),
            all_output.end());

D
dongzhihong 已提交
254
  ASSERT_EQ(2UL, bwd_net->ops_.size());
Y
Yu Yang 已提交
255 256
  ASSERT_TRUE(bwd_net->ops_[1]->IsNetOp());
  auto first_fc_grad = static_cast<f::NetOp *>(bwd_net->ops_[1].get());
D
dongzhihong 已提交
257
  ASSERT_EQ(3UL, first_fc_grad->ops_.size());
Y
Yu Yang 已提交
258 259 260
  ASSERT_EQ(
      f::OperatorBase::EMPTY_VAR_NAME(),
      first_fc_grad->ops_[2]->Output("A" + f::OperatorBase::GRAD_VAR_SUFFIX()));
Y
Yu Yang 已提交
261 262 263 264 265 266 267 268 269 270 271 272
}

TEST(Backward, net_shared_weight) {
  f::NetOp net;
  net.AddOp(f::OpRegistry::CreateOp("mul", {"X", "W"}, {"Out"}, {}));
  net.AddOp(f::OpRegistry::CreateOp("mul", {"Out", "W"}, {"FinalOut"}, {}));
  net.CompleteAddOp();

  auto bwd = f::Backward(net, {});
  ASSERT_TRUE(bwd->IsNetOp());
  auto bwd_net = static_cast<f::NetOp *>(bwd.get());
  ASSERT_EQ(3UL, bwd_net->ops_.size());
Y
Yu Yang 已提交
273
  ASSERT_EQ("add", bwd_net->ops_[2]->type_);
Y
Stash  
Yu Yang 已提交
274 275
}

Y
Yu Yang 已提交
276
TEST(Backward, op_register_grad_not_for_network) {
Y
Yu Yang 已提交
277 278 279
  auto fwd = f::OpRegistry::CreateOp(
      "fc", {"X", "W", "b"}, {"mul_out", "add_out", "out1"},
      {{"temporary_index", std::vector<int>{0, 1}}});
D
dongzhihong 已提交
280

Y
Yu Yang 已提交
281 282 283
  ASSERT_THROW(f::OpRegistry::CreateGradOp(*fwd), EnforceNotMet);
}

Y
Yu Yang 已提交
284
TEST(Backward, op_all_input_are_not_need) {
Y
Yu Yang 已提交
285 286 287 288 289 290 291
  auto fwd = f::OpRegistry::CreateOp("rowwise_add", {"X", "b"}, {"Out"}, {});
  auto backward = f::Backward(*fwd, {"X", "b"});
  ASSERT_TRUE(backward->IsNetOp());
  auto net = static_cast<f::NetOp *>(backward.get());
  ASSERT_TRUE(net->ops_.empty());
}

Y
Yu Yang 已提交
292
TEST(Backward, op_all_output_are_not_need) {
Y
Yu Yang 已提交
293 294 295 296 297 298 299
  auto fwd = f::OpRegistry::CreateOp("rowwise_add", {"X", "b"}, {"Out"}, {});
  auto backward = f::Backward(*fwd, {"Out"});
  ASSERT_TRUE(backward->IsNetOp());
  auto net = static_cast<f::NetOp *>(backward.get());
  ASSERT_TRUE(net->ops_.empty());
}

Y
Yu Yang 已提交
300
TEST(Backward, op_part_of_output_are_not_need) {
Y
Yu Yang 已提交
301 302 303 304
  auto fwd = f::OpRegistry::CreateOp("many_output_op", {"X"}, {"Y", "Z"}, {});
  auto backward = f::Backward(*fwd, {"Z"});
  ASSERT_TRUE(backward->IsNetOp());
  auto net = static_cast<f::NetOp *>(backward.get());
Y
Stash  
Yu Yang 已提交
305
  ASSERT_EQ(net->ops_.size(), 2UL);
Y
Yu Yang 已提交
306 307 308

  auto &fill_zero = *net->ops_[0];
  ASSERT_EQ("fill_zeros_like", fill_zero.type_);
309
  ASSERT_EQ(1UL, fill_zero.inputs_.size());
Y
Yu Yang 已提交
310
  ASSERT_EQ("Z", fill_zero.inputs_[0]);
311 312
  ASSERT_EQ(1UL, fill_zero.outputs_.size());
  ASSERT_EQ("Z" + f::OperatorBase::ZERO_VAR_SUFFIX(), fill_zero.outputs_[0]);
Y
Yu Yang 已提交
313 314 315

  auto &d_many_out = *net->ops_[1];
  ASSERT_EQ("many_output_op_grad", d_many_out.type_);
316 317 318 319 320 321 322
  ASSERT_EQ(1UL + 2UL + 2UL, d_many_out.inputs_.size());  // I/O/OG
  ASSERT_EQ("Z" + f::OperatorBase::ZERO_VAR_SUFFIX(),
            d_many_out.Input("z" + f::OperatorBase::GRAD_VAR_SUFFIX()));
  ASSERT_EQ("Y" + f::OperatorBase::GRAD_VAR_SUFFIX(),
            d_many_out.Input("y" + f::OperatorBase::GRAD_VAR_SUFFIX()));
  ASSERT_EQ("X" + f::OperatorBase::GRAD_VAR_SUFFIX(),
            d_many_out.Output("x" + f::OperatorBase::GRAD_VAR_SUFFIX()));
Y
Yu Yang 已提交
323 324
}

Y
Yu Yang 已提交
325
TEST(Backward, op_part_of_input_are_not_need) {
326 327
  auto fwd = f::OpRegistry::CreateOp("mul", {"a", "b"}, {"out"}, {});
  auto backward = f::Backward(*fwd, {"a"});
D
dongzhihong 已提交
328
  auto &grad_mul = *backward;
329 330 331 332 333 334 335 336 337
  ASSERT_EQ(grad_mul.type_, "mul_grad");
  ASSERT_EQ(grad_mul.inputs_.size(), 2UL + 1UL + 1UL);
  ASSERT_EQ(grad_mul.outputs_.size(), 2UL);
  ASSERT_EQ(grad_mul.Output("A" + f::OperatorBase::GRAD_VAR_SUFFIX()),
            f::OperatorBase::EMPTY_VAR_NAME());
  ASSERT_EQ(grad_mul.Output("B" + f::OperatorBase::GRAD_VAR_SUFFIX()),
            "b" + f::OperatorBase::GRAD_VAR_SUFFIX());
  ASSERT_EQ(grad_mul.Input("Out" + f::OperatorBase::GRAD_VAR_SUFFIX()),
            "out" + f::OperatorBase::GRAD_VAR_SUFFIX());
338 339 340 341 342
  ASSERT_EQ(grad_mul.Input("A"), "a");
  ASSERT_EQ(grad_mul.Input("B"), "b");
  ASSERT_EQ(grad_mul.Input("Out"), "out");
}

F
fengjiayi 已提交
343
TEST(Backward, linear_net_intermediate_variable_has_no_grad) {
344
  f::NetOp net;
Y
Yu Yang 已提交
345 346 347 348 349 350
  net.AddOp(f::OpRegistry::CreateOp("fc", {"x1", "w1", "b1"},
                                    {"mul_out1", "add_out1", "out1"}, {}));
  net.AddOp(f::OpRegistry::CreateOp("fc", {"out1", "w2", "b2"},
                                    {"mul_out2", "tmp_out2", "out2"}, {}));
  net.AddOp(f::OpRegistry::CreateOp("fc", {"out2", "w3", "b3"},
                                    {"mul_out3", "tmp_out3", "out3"}, {}));
351
  net.CompleteAddOp();
352
  auto backward = f::Backward(net, {"mul_out2", "tmp_out2", "out2"});
353 354
  ASSERT_TRUE(backward->IsNetOp());
  auto bwd_net = static_cast<f::NetOp *>(backward.get());
355
  ASSERT_EQ(bwd_net->ops_.size(), 3UL);
356
  auto &grad_fc = *bwd_net->ops_[0];
357 358 359 360 361 362 363 364 365
  EXPECT_EQ(grad_fc.inputs_.size(),
            3UL       /* external input number */
                + 1UL /* external output number*/
                + 1UL /* number of gradient of external output*/
                - 1UL /*ignoreGradient varable number*/
                + 2U /* internal variable number*/);
  EXPECT_EQ(grad_fc.outputs_.size(), 2UL       /* input number of mul*/
                                         + 2UL /* input number of rowwise_add */
                                         + 1UL /* input number of sigmod */);
F
fengjiayi 已提交
366 367 368 369 370
  EXPECT_EQ(bwd_net->ops_[1]->inputs_.size(), 0UL);
  EXPECT_EQ(bwd_net->ops_[1]->outputs_.size(), 0UL);
  EXPECT_EQ(bwd_net->ops_[2]->inputs_.size(), 0UL);
  EXPECT_EQ(bwd_net->ops_[2]->outputs_.size(), 0UL);

371 372 373
  /*
    EXPECT_EQ(grad_fc.Output("X" + f::OperatorBase::GRAD_VAR_SUFFIX()),
              f::OperatorBase::EMPTY_VAR_NAME());
374
  EXPECT_EQ(grad_fc.Output("W" + f::OperatorBase::GRAD_VAR_SUFFIX()),
375
    "w3" + f::OperatorBase::GRAD_VAR_SUFFIX());
376
  EXPECT_EQ(grad_fc.Output("b" + f::OperatorBase::GRAD_VAR_SUFFIX()),
377 378 379
    "b3" + f::OperatorBase::GRAD_VAR_SUFFIX());
  EXPECT_EQ(grad_fc.Output("mul_result" + f::OperatorBase::GRAD_VAR_SUFFIX()),
  "mul_out3" + f::OperatorBase::GRAD_VAR_SUFFIX());
380

381 382
  EXPECT_EQ(grad_fc.Input("Out" + f::OperatorBase::GRAD_VAR_SUFFIX()),
  "out3" + f::OperatorBase::GRAD_VAR_SUFFIX());
383 384 385 386 387
  EXPECT_EQ(grad_fc.Input("X"), "out2");
  EXPECT_EQ(grad_fc.Input("W"), "w3");
  EXPECT_EQ(grad_fc.Input("mul_result"), "mul_out3");
  EXPECT_EQ(grad_fc.Input("add_result"), "tmp_out3");
  EXPECT_EQ(grad_fc.Input("Out"), "out3");
388
  */
D
dongzhihong 已提交
389
}