executor.cc 14.6 KB
Newer Older
1
/* Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserved.
Q
qijun 已提交
2 3 4 5 6 7 8 9 10 11 12 13 14

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
Yi Wang 已提交
15
#include "paddle/fluid/framework/executor.h"
Y
Yang Yang 已提交
16

17
#include "paddle/fluid/framework/channel.h"
Y
Yi Wang 已提交
18 19 20 21 22 23
#include "paddle/fluid/framework/feed_fetch_method.h"
#include "paddle/fluid/framework/lod_rank_table.h"
#include "paddle/fluid/framework/lod_tensor_array.h"
#include "paddle/fluid/framework/op_registry.h"
#include "paddle/fluid/framework/reader.h"
#include "paddle/fluid/platform/place.h"
X
Xin Pan 已提交
24
#include "paddle/fluid/platform/profiler.h"
Y
Yang Yu 已提交
25

D
dzhwinter 已提交
26
DECLARE_bool(benchmark);
27
DEFINE_bool(use_mkldnn, false, "Use MKLDNN to run");
Q
qijun 已提交
28 29 30

namespace paddle {
namespace framework {
X
Xin Pan 已提交
31 32 33 34 35
namespace {
// block id starts from 0. This id is used to represent the codeblock
// wrapping the first block 0.
int kProgramId = -1;
}  // namespace
Q
qijun 已提交
36

Q
Qiao Longfei 已提交
37 38 39
ExecutorPrepareContext::ExecutorPrepareContext(
    const framework::ProgramDesc& prog, size_t block_id)
    : prog_(prog), block_id_(block_id) {}
Y
Yu Yang 已提交
40

Q
Qiao Longfei 已提交
41 42 43
ExecutorPrepareContext::~ExecutorPrepareContext() {
  VLOG(5) << "destroy ExecutorPrepareContext";
}
Y
Yu Yang 已提交
44

D
dzhwinter 已提交
45
Executor::Executor(const platform::Place& place) : place_(place) {}
Q
qijun 已提交
46

Y
Stash  
Yu Yang 已提交
47
void InitializeVariable(Variable* var, proto::VarType::Type var_type) {
48
  if (var_type == proto::VarType::LOD_TENSOR) {
Q
QI JUN 已提交
49
    var->GetMutable<LoDTensor>();
50
  } else if (var_type == proto::VarType::SELECTED_ROWS) {
Q
QI JUN 已提交
51
    var->GetMutable<SelectedRows>();
52
  } else if (var_type == proto::VarType::FEED_MINIBATCH) {
Q
QI JUN 已提交
53
    var->GetMutable<FeedFetchList>();
54
  } else if (var_type == proto::VarType::FETCH_LIST) {
Q
QI JUN 已提交
55
    var->GetMutable<FeedFetchList>();
56
  } else if (var_type == proto::VarType::STEP_SCOPES) {
Y
Yu Yang 已提交
57
    var->GetMutable<std::vector<framework::Scope>>();
58
  } else if (var_type == proto::VarType::LOD_RANK_TABLE) {
Y
Yu Yang 已提交
59
    var->GetMutable<LoDRankTable>();
60
  } else if (var_type == proto::VarType::LOD_TENSOR_ARRAY) {
Y
Yu Yang 已提交
61
    var->GetMutable<LoDTensorArray>();
62
  } else if (var_type == proto::VarType::PLACE_LIST) {
Y
Yang Yu 已提交
63
    var->GetMutable<platform::PlaceList>();
64
  } else if (var_type == proto::VarType::READER) {
F
fengjiayi 已提交
65
    var->GetMutable<ReaderHolder>();
66 67
  } else if (var_type == proto::VarType::CHANNEL) {
    var->GetMutable<ChannelHolder>();
T
typhoonzero 已提交
68 69
  } else if (var_type == proto::VarType::RAW) {
    // GetMutable will be called in operator
Q
QI JUN 已提交
70 71
  } else {
    PADDLE_THROW(
Y
Yu Yang 已提交
72
        "Variable type %d is not in "
F
fengjiayi 已提交
73
        "[LOD_TENSOR, SELECTED_ROWS, FEED_MINIBATCH, FETCH_LIST, "
T
typhoonzero 已提交
74
        "LOD_RANK_TABLE, PLACE_LIST, READER, CHANNEL, RAW]",
Y
Yu Yang 已提交
75
        var_type);
Q
QI JUN 已提交
76 77 78
  }
}

L
Liu Yiqun 已提交
79 80 81
void Executor::CreateVariables(const ProgramDesc& pdesc, Scope* scope,
                               int block_id) {
  auto& global_block = pdesc.Block(block_id);
82 83 84 85 86 87 88 89 90 91 92 93 94 95

  const Scope* ancestor_scope = scope;
  while (ancestor_scope->parent()) {
    ancestor_scope = ancestor_scope->parent();
  }

  if (ancestor_scope != scope) {
    for (auto& var : global_block.AllVars()) {
      if (var->Name() == framework::kEmptyVarName) {
        continue;
      }

      if (var->Persistable()) {
        auto* ptr = const_cast<Scope*>(ancestor_scope)->Var(var->Name());
96
        InitializeVariable(ptr, var->GetType());
97 98 99 100
        VLOG(3) << "Create Variable " << var->Name()
                << " global, which pointer is " << ptr;
      } else {
        auto* ptr = scope->Var(var->Name());
101
        InitializeVariable(ptr, var->GetType());
102 103 104 105 106 107 108
        VLOG(3) << "Create Variable " << var->Name()
                << " locally, which pointer is " << ptr;
      }
    }
  } else {
    for (auto& var : global_block.AllVars()) {
      auto* ptr = scope->Var(var->Name());
109
      InitializeVariable(ptr, var->GetType());
110 111 112 113 114 115
      VLOG(3) << "Create variable " << var->Name() << ", which pointer is "
              << ptr;
    }
  }
}

Y
Yu Yang 已提交
116
void Executor::Run(const ProgramDesc& pdesc, Scope* scope, int block_id,
T
typhoonzero 已提交
117
                   bool create_local_scope, bool create_vars) {
X
Xin Pan 已提交
118
  platform::RecordBlock b(block_id);
119
  if (FLAGS_use_mkldnn) EnableMKLDNN(pdesc);
Q
Qiao Longfei 已提交
120 121
  auto ctx = Prepare(pdesc, block_id);
  RunPreparedContext(ctx.get(), scope, create_local_scope, create_vars);
Q
qijun 已提交
122 123
}

124 125 126 127 128 129 130
// Check whether the block already has feed operators and feed_holder.
// Return false if the block does not have any feed operators.
// If some feed operators have been prepended to the block, check that
// the info contained in these feed operators matches the feed_targets
// and feed_holder_name. Raise exception when any mismatch is found.
// Return true if the block has feed operators and holder of matching info.
static bool has_feed_operators(
131
    const BlockDesc& block,
L
Liu Yiqun 已提交
132
    const std::map<std::string, const LoDTensor*>& feed_targets,
133 134
    const std::string& feed_holder_name) {
  size_t feed_count = 0;
135
  for (auto* op : block.AllOps()) {
136 137
    if (op->Type() == kFeedOpType) {
      feed_count++;
L
Liu Yiqun 已提交
138
      // The input variable's name of feed_op should be feed_holder_name.
139 140 141 142 143 144 145 146 147 148 149 150 151 152 153
      PADDLE_ENFORCE_EQ(op->Input("X")[0], feed_holder_name,
                        "Input to feed op should be '%s'", feed_holder_name);
      std::string feed_target_name = op->Output("Out")[0];
      PADDLE_ENFORCE(
          feed_targets.find(feed_target_name) != feed_targets.end(),
          "Feed operator output name '%s' cannot be found in 'feed_targets'",
          feed_target_name);
    }
  }

  if (feed_count > 0) {
    PADDLE_ENFORCE_EQ(
        feed_count, feed_targets.size(),
        "The number of feed operators should match 'feed_targets'");

154
    if (!feed_holder_name.empty()) {
L
Liu Yiqun 已提交
155
      // When feed operator are present, so should be feed_holder.
156 157 158 159 160 161 162
      auto var = block.FindVar(feed_holder_name);
      PADDLE_ENFORCE_NOT_NULL(var, "Block should already have a '%s' variable",
                              feed_holder_name);
      PADDLE_ENFORCE_EQ(var->GetType(), proto::VarType::FEED_MINIBATCH,
                        "'%s' variable should be 'FEED_MINIBATCH' type",
                        feed_holder_name);
    }
163 164 165 166 167 168 169 170 171 172 173 174
  }

  return feed_count > 0;
}

// Check whether the block already has fetch operators and fetch_holder.
// Return false if the block does not have any fetch operators.
// If some fetch operators have been appended to the block, check that
// the info contained in these fetch operators matches the fetch_targets
// and fetch_holder_name. Raise exception when any mismatch is found.
// Return true if the block has fetch operators and holder of matching info.
static bool has_fetch_operators(
L
Liu Yiqun 已提交
175 176
    const BlockDesc& block,
    const std::map<std::string, LoDTensor*>& fetch_targets,
177 178
    const std::string& fetch_holder_name) {
  size_t fetch_count = 0;
179
  for (auto* op : block.AllOps()) {
180 181
    if (op->Type() == kFetchOpType) {
      fetch_count++;
L
Liu Yiqun 已提交
182
      // The output variable's name of fetch_op should be fetch_holder_name.
183 184 185 186 187 188 189 190 191 192 193 194 195 196 197
      PADDLE_ENFORCE_EQ(op->Output("Out")[0], fetch_holder_name,
                        "Output of fetch op should be '%s'", fetch_holder_name);
      std::string fetch_target_name = op->Input("X")[0];
      PADDLE_ENFORCE(
          fetch_targets.find(fetch_target_name) != fetch_targets.end(),
          "Fetch operator input name '%s' cannot be found in 'fetch_targets'",
          fetch_target_name);
    }
  }

  if (fetch_count > 0) {
    PADDLE_ENFORCE_EQ(
        fetch_count, fetch_targets.size(),
        "The number of fetch operators should match 'fetch_targets'");

198
    if (!fetch_holder_name.empty()) {
L
Liu Yiqun 已提交
199
      // When fetch operator are present, so should be fetch_holder.
200 201 202 203 204 205 206
      auto var = block.FindVar(fetch_holder_name);
      PADDLE_ENFORCE_NOT_NULL(var, "Block should already have a '%s' variable",
                              fetch_holder_name);
      PADDLE_ENFORCE_EQ(var->GetType(), proto::VarType::FETCH_LIST,
                        "'%s' variable should be 'FETCH_LIST' type",
                        fetch_holder_name);
    }
207 208 209 210 211 212
  }

  return fetch_count > 0;
}

void Executor::Run(const ProgramDesc& program, Scope* scope,
213 214
                   std::map<std::string, const LoDTensor*>* feed_targets,
                   std::map<std::string, LoDTensor*>* fetch_targets,
W
Wu Yi 已提交
215 216
                   bool create_local_scope, bool create_vars,
                   const std::string& feed_holder_name,
217
                   const std::string& fetch_holder_name) {
X
Xin Pan 已提交
218
  platform::RecordBlock b(kProgramId);
219
  if (FLAGS_use_mkldnn) EnableMKLDNN(program);
220
  bool has_feed_ops =
221
      has_feed_operators(program.Block(0), *feed_targets, feed_holder_name);
222
  bool has_fetch_ops =
223
      has_fetch_operators(program.Block(0), *fetch_targets, fetch_holder_name);
224 225

  ProgramDesc* copy_program = const_cast<ProgramDesc*>(&program);
S
sneaxiy 已提交
226
  std::unique_ptr<ProgramDesc> unique_ptr_of_copy_program;
227
  if (!has_feed_ops || !has_fetch_ops) {
S
sneaxiy 已提交
228 229
    unique_ptr_of_copy_program.reset(new ProgramDesc(program));
    copy_program = unique_ptr_of_copy_program.get();
230
  }
231 232
  auto* global_block = copy_program->MutableBlock(0);

233
  if (!has_feed_ops) {
234 235
    // create feed_holder variable
    auto* feed_holder = global_block->Var(feed_holder_name);
236
    feed_holder->SetType(proto::VarType::FEED_MINIBATCH);
237 238 239
    feed_holder->SetPersistable(true);

    int i = 0;
240
    for (auto& feed_target : (*feed_targets)) {
241 242 243 244 245 246 247 248 249 250 251 252 253 254 255
      std::string var_name = feed_target.first;
      VLOG(3) << "feed target's name: " << var_name;

      // prepend feed op
      auto* op = global_block->PrependOp();
      op->SetType(kFeedOpType);
      op->SetInput("X", {feed_holder_name});
      op->SetOutput("Out", {var_name});
      op->SetAttr("col", {static_cast<int>(i)});
      op->CheckAttrs();

      i++;
    }
  }

256
  if (!has_fetch_ops) {
257 258
    // create fetch_holder variable
    auto* fetch_holder = global_block->Var(fetch_holder_name);
259
    fetch_holder->SetType(proto::VarType::FETCH_LIST);
260 261 262
    fetch_holder->SetPersistable(true);

    int i = 0;
263
    for (auto& fetch_target : (*fetch_targets)) {
264 265 266 267 268 269 270 271 272 273 274 275 276 277 278
      std::string var_name = fetch_target.first;
      VLOG(3) << "fetch target's name: " << var_name;

      // append fetch op
      auto* op = global_block->AppendOp();
      op->SetType(kFetchOpType);
      op->SetInput("X", {var_name});
      op->SetOutput("Out", {fetch_holder_name});
      op->SetAttr("col", {static_cast<int>(i)});
      op->CheckAttrs();

      i++;
    }
  }

279
  auto ctx = Prepare(*copy_program, 0);
W
Wu Yi 已提交
280 281 282
  RunPreparedContext(ctx.get(), scope, feed_targets, fetch_targets,
                     create_local_scope, create_vars, feed_holder_name,
                     fetch_holder_name);
283 284
}

Q
Qiao Longfei 已提交
285 286
std::unique_ptr<ExecutorPrepareContext> Executor::Prepare(
    const ProgramDesc& program, int block_id) {
Y
Yu Yang 已提交
287 288 289 290 291 292
  auto* ctx = new ExecutorPrepareContext(program, block_id);
  PADDLE_ENFORCE_LT(static_cast<size_t>(block_id), program.Size());
  auto& block = program.Block(block_id);
  for (auto& op_desc : block.AllOps()) {
    ctx->ops_.push_back(OpRegistry::CreateOp(*op_desc));
  }
Q
Qiao Longfei 已提交
293
  return std::unique_ptr<ExecutorPrepareContext>(ctx);
Y
Yu Yang 已提交
294 295
}

T
refine  
typhoonzero 已提交
296
std::vector<std::shared_ptr<ExecutorPrepareContext>> Executor::Prepare(
T
typhoonzero 已提交
297 298 299 300 301 302 303 304 305 306 307 308 309 310
    const ProgramDesc& program, const std::vector<int>& block_ids) {
  std::vector<std::shared_ptr<ExecutorPrepareContext>> result;
  for (auto& bid : block_ids) {
    auto* ctx = new ExecutorPrepareContext(program, bid);
    PADDLE_ENFORCE_LT(static_cast<size_t>(bid), program.Size());
    auto& block = program.Block(bid);
    for (auto& op_desc : block.AllOps()) {
      ctx->ops_.push_back(OpRegistry::CreateOp(*op_desc));
    }
    result.push_back(std::shared_ptr<ExecutorPrepareContext>(ctx));
  }
  return result;
}

Y
Yu Yang 已提交
311 312 313 314 315 316
void Executor::RunPreparedContext(ExecutorPrepareContext* ctx, Scope* scope,
                                  bool create_local_scope, bool create_vars) {
  Scope* local_scope = scope;
  if (create_vars) {
    if (create_local_scope) {
      local_scope = &scope->NewScope();
317 318
    }
    CreateVariables(ctx->prog_, local_scope, ctx->block_id_);
L
Liu Yiqun 已提交
319
  }
Y
Yu Yang 已提交
320 321 322

  for (auto& op : ctx->ops_) {
    VLOG(3) << place_ << " " << op->DebugStringEx(local_scope);
323
    op->Run(*local_scope, place_);
Y
Yang Yang 已提交
324

Y
Yu Yang 已提交
325 326 327 328 329
    if (FLAGS_benchmark) {
      VLOG(2) << "Memory used after operator " + op->Type() + " running: "
              << memory::memory_usage(place_);
    }
  }
330
  platform::DeviceContextPool::Instance().Get(place_)->Wait();
Y
Yu Yang 已提交
331 332
  if (create_vars && create_local_scope) {
    scope->DeleteScope(local_scope);
333 334 335
  } else {
    // Delete the local scopes created in operators.
    scope->DropKids();
Y
Yu Yang 已提交
336 337 338 339 340 341 342 343 344
  }
  if (FLAGS_benchmark) {
    VLOG(2) << "-------------------------------------------------------";
    VLOG(2) << "Memory used after deleting local scope: "
            << memory::memory_usage(place_);
    VLOG(2) << "-------------------------------------------------------";
  }
}

345 346
void Executor::RunPreparedContext(
    ExecutorPrepareContext* ctx, Scope* scope,
347
    std::map<std::string, const LoDTensor*>* feed_targets,
W
Wu Yi 已提交
348 349 350
    std::map<std::string, LoDTensor*>* fetch_targets, bool create_local_scope,
    bool create_vars, const std::string& feed_holder_name,
    const std::string& fetch_holder_name) {
351 352
  auto& global_block = ctx->prog_.Block(ctx->block_id_);

353
  PADDLE_ENFORCE(
354
      has_feed_operators(global_block, *feed_targets, feed_holder_name),
355 356
      "Program in ExecutorPrepareContext should has feed_ops.");
  PADDLE_ENFORCE(
357
      has_fetch_operators(global_block, *fetch_targets, fetch_holder_name),
358 359
      "Program in the prepared context should has fetch_ops.");

360 361 362 363 364
  // map the data of feed_targets to feed_holder
  for (auto* op : global_block.AllOps()) {
    if (op->Type() == kFeedOpType) {
      std::string feed_target_name = op->Output("Out")[0];
      int idx = boost::get<int>(op->GetAttr("col"));
365 366
      SetFeedVariable(scope, *(*feed_targets)[feed_target_name],
                      feed_holder_name, idx);
367 368 369
    }
  }

W
Wu Yi 已提交
370
  RunPreparedContext(ctx, scope, create_local_scope, create_vars);
371 372 373 374 375 376

  // obtain the data of fetch_targets from fetch_holder
  for (auto* op : global_block.AllOps()) {
    if (op->Type() == kFetchOpType) {
      std::string fetch_target_name = op->Input("X")[0];
      int idx = boost::get<int>(op->GetAttr("col"));
377
      *(*fetch_targets)[fetch_target_name] =
378 379 380 381 382
          GetFetchVariable(*scope, fetch_holder_name, idx);
    }
  }
}

383 384 385 386 387 388 389 390 391 392 393 394 395 396
void Executor::EnableMKLDNN(const ProgramDesc& program) {
#ifdef PADDLE_WITH_MKLDNN
  VLOG(3) << "use_mkldnn=True";
  for (size_t bid = 0; bid < program.Size(); ++bid) {
    auto* block = const_cast<ProgramDesc&>(program).MutableBlock(bid);
    for (auto* op : block->AllOps()) {
      if (op->HasAttr("use_mkldnn")) {
        op->SetAttr("use_mkldnn", true);
      }
    }
  }
#endif
}

Q
qijun 已提交
397 398
}  // namespace framework
}  // namespace paddle