operator.cc 19.6 KB
Newer Older
1
/* Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserved.
Q
Qiao Longfei 已提交
2 3 4 5 6 7 8 9 10 11 12 13

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. */
D
dzhwinter 已提交
14
#include <gflags/gflags.h>
D
dzhwinter 已提交
15
#include <glog/logging.h>
Q
Qiao Longfei 已提交
16

17
#include <algorithm>
D
dzhwinter 已提交
18

Y
Yi Wang 已提交
19 20 21 22 23
#include "paddle/fluid/framework/data_transform.h"
#include "paddle/fluid/framework/executor.h"
#include "paddle/fluid/framework/operator.h"
#include "paddle/fluid/framework/shape_inference.h"
#include "paddle/fluid/framework/var_type.h"
Q
Qiao Longfei 已提交
24

D
dzhwinter 已提交
25
DECLARE_bool(benchmark);
D
dzhwinter 已提交
26

Q
Qiao Longfei 已提交
27 28 29
namespace paddle {
namespace framework {

30 31 32 33 34 35
std::vector<std::tuple<platform::Place, LibraryType>> kKernelPriority = {
    std::make_tuple(platform::CUDAPlace(0), LibraryType::kCUDNN),
    std::make_tuple(platform::CUDAPlace(0), LibraryType::kPlain),
    std::make_tuple(platform::CPUPlace(), LibraryType::kMKLDNN),
    std::make_tuple(platform::CPUPlace(), LibraryType::kPlain),
};
D
dzhwinter 已提交
36

37 38
static DDim GetDims(const Scope& scope, const std::string& name) {
  Variable* var = scope.FindVar(name);
Q
qiaolongfei 已提交
39 40
  if (var == nullptr) {
    return DDim({-1});
Q
Qiao Longfei 已提交
41 42 43
  }

  if (var->IsType<LoDTensor>()) {
44 45 46 47 48 49 50 51
    return var->Get<LoDTensor>().dims();
  } else if (var->IsType<SelectedRows>()) {
    return var->Get<SelectedRows>().GetCompleteDims();
  } else {
    return DDim({-1});
  }
}

Q
Qiao Longfei 已提交
52 53 54 55 56 57 58 59 60 61 62 63 64 65 66
static LoD GetLoD(const Scope& scope, const std::string& name) {
  Variable* var = scope.FindVar(name);
  auto default_lod = LoD({{}});

  if (var == nullptr) {
    return default_lod;
  }

  if (var->IsType<LoDTensor>()) {
    return var->Get<LoDTensor>().lod();
  } else {
    return default_lod;
  }
}

67 68 69 70 71 72 73 74 75 76 77 78
void OperatorBase::Run(const Scope& scope, const platform::Place& place) {
  if (platform::is_gpu_place(place)) {
#ifndef PADDLE_WITH_CUDA
    PADDLE_THROW("Cannot run operator on place %s", place);
#else
    auto dev_id = boost::get<platform::CUDAPlace>(place).device;
    platform::SetDeviceId(dev_id);
#endif
  }
  RunImpl(scope, place);
}

79
std::string OperatorBase::Input(const std::string& name) const {
Y
Yu Yang 已提交
80
  auto& ins = Inputs(name);
Y
Yu Yang 已提交
81
  PADDLE_ENFORCE_LE(ins.size(), 1UL,
82 83
                    "Operator %s's input %s should contain only one variable.",
                    type_, name);
Y
Yu Yang 已提交
84
  return ins.empty() ? kEmptyVarName : ins[0];
Y
Yan Chunwei 已提交
85 86
}

Y
Yu Yang 已提交
87 88
const std::vector<std::string>& OperatorBase::Inputs(
    const std::string& name) const {
Y
Yu Yang 已提交
89
  auto it = inputs_.find(name);
90 91
  PADDLE_ENFORCE(it != inputs_.end(), "Operator %s does not have the input %s.",
                 type_, name);
Y
Yu Yang 已提交
92
  return it->second;
Y
Yan Chunwei 已提交
93 94
}

95
std::string OperatorBase::Output(const std::string& name) const {
Y
Yu Yang 已提交
96
  auto& outs = Outputs(name);
Y
Yu Yang 已提交
97
  PADDLE_ENFORCE_LE(outs.size(), 1UL,
98 99
                    "Operator %s's output %s should contain only one variable.",
                    type_, name);
Y
Yu Yang 已提交
100
  return outs.empty() ? kEmptyVarName : outs[0];
Y
Yan Chunwei 已提交
101 102
}

Y
Yu Yang 已提交
103 104
const std::vector<std::string>& OperatorBase::Outputs(
    const std::string& name) const {
Y
Yu Yang 已提交
105
  auto it = outputs_.find(name);
106 107
  PADDLE_ENFORCE(it != outputs_.end(),
                 "Operator %s does not have an output called %s.", type_, name);
Y
Yu Yang 已提交
108
  return it->second;
Y
Yan Chunwei 已提交
109 110
}

111
std::string OperatorBase::DebugStringEx(const Scope* scope) const {
Q
Qiao Longfei 已提交
112
  std::stringstream ss;
Y
Yu Yang 已提交
113
  ss << "Op(" << type_ << "), inputs:{";
Y
Yu Yang 已提交
114 115
  for (auto it = inputs_.begin(); it != inputs_.end();) {
    auto& input = *it;
Y
Yu Yang 已提交
116 117 118
    ss << input.first << "[";
    for (size_t i = 0; i < input.second.size(); ++i) {
      ss << input.second[i];
119
      if (scope) {
Q
Qiao Longfei 已提交
120 121
        ss << "[" << GetDims(*scope, input.second[i]) << "]";
        ss << "(" << GetLoD(*scope, input.second[i]) << ")";
122
      }
Y
Yu Yang 已提交
123 124 125
      if (i != input.second.size() - 1) {
        ss << ", ";
      }
126
    }
Y
Yu Yang 已提交
127
    ss << "]";
Y
Yu Yang 已提交
128 129
    ++it;
    if (it != inputs_.end()) {
130 131
      ss << ", ";
    }
Q
Qiao Longfei 已提交
132
  }
Y
Yu Yang 已提交
133
  ss << "}, outputs:{";
Y
Yu Yang 已提交
134 135
  for (auto it = outputs_.begin(); it != outputs_.end();) {
    auto& output = *it;
Y
Yu Yang 已提交
136 137 138
    ss << output.first << "[";
    for (size_t i = 0; i < output.second.size(); ++i) {
      ss << output.second[i];
139
      if (scope) {
Q
Qiao Longfei 已提交
140 141
        ss << "[" << GetDims(*scope, output.second[i]) << "]";
        ss << "(" << GetLoD(*scope, output.second[i]) << ")";
142
      }
Y
Yu Yang 已提交
143 144 145
      if (i != output.second.size() - 1) {
        ss << ", ";
      }
146
    }
Y
Yu Yang 已提交
147
    ss << "]";
Y
Yu Yang 已提交
148 149
    ++it;
    if (it != outputs_.end()) {
150 151
      ss << ", ";
    }
Q
Qiao Longfei 已提交
152
  }
Y
Yu Yang 已提交
153
  ss << "}.";
Q
Qiao Longfei 已提交
154 155 156
  return ss.str();
}

D
dongzhihong 已提交
157 158
void OperatorBase::Rename(const std::string& old_name,
                          const std::string& new_name) {
Y
Yu Yang 已提交
159 160 161 162 163 164 165
  for (auto& input : inputs_) {
    std::replace(input.second.begin(), input.second.end(), old_name, new_name);
  }
  for (auto& output : outputs_) {
    std::replace(output.second.begin(), output.second.end(), old_name,
                 new_name);
  }
D
dongzhihong 已提交
166 167
}

Y
Yu Yang 已提交
168
OperatorBase::OperatorBase(const std::string& type,
Y
Yu Yang 已提交
169 170
                           const VariableNameMap& inputs,
                           const VariableNameMap& outputs,
Y
Yu Yang 已提交
171 172
                           const AttributeMap& attrs)
    : type_(type), inputs_(inputs), outputs_(outputs), attrs_(attrs) {
173 174
  GenerateTemporaryNames();
  CheckAllInputOutputSet();
Y
Yu Yang 已提交
175
}
176

Q
qijun 已提交
177 178
std::vector<std::string> OperatorBase::InputVars() const {
  std::vector<std::string> ret_val;
Y
Yu Yang 已提交
179
  for (auto& o : inputs_) {
Q
qijun 已提交
180 181 182 183 184 185
    ret_val.reserve(ret_val.size() + o.second.size());
    ret_val.insert(ret_val.end(), o.second.begin(), o.second.end());
  }
  return ret_val;
}

Y
Yu Yang 已提交
186 187 188 189 190 191 192 193 194 195
std::vector<std::string> OperatorBase::OutputVars(bool has_intermediate) const {
  std::vector<std::string> ret_val;
  if (has_intermediate) {
    // push all outputs into ret_val
    for (auto& o : outputs_) {
      ret_val.reserve(ret_val.size() + o.second.size());
      ret_val.insert(ret_val.end(), o.second.begin(), o.second.end());
    }
    return ret_val;
  }
Y
Yu Yang 已提交
196
  auto& info = OpInfoMap::Instance().Get(Type());
Y
Yu Yang 已提交
197 198

  // get all OpProto::Var for outputs
Y
Yu Yang 已提交
199
  for (auto& o : info.Proto().outputs()) {
Y
Yu Yang 已提交
200 201 202 203 204 205 206 207 208
    // ignore all intermediate output
    if (o.intermediate()) continue;
    auto out = outputs_.find(o.name());
    if (out != outputs_.end()) {
      ret_val.reserve(ret_val.size() + out->second.size());
      ret_val.insert(ret_val.end(), out->second.begin(), out->second.end());
    }
  }
  return ret_val;
D
dongzhihong 已提交
209 210
}

211 212 213
void OperatorBase::CheckAllInputOutputSet() const {
  auto& info_map = OpInfoMap::Instance();
  auto* op_info = info_map.GetNullable(Type());
Y
Yu Yang 已提交
214
  if (op_info == nullptr || op_info->proto_ == nullptr) return;
215 216 217

  for (auto& in : op_info->Proto().inputs()) {
    PADDLE_ENFORCE(inputs_.find(in.name()) != inputs_.end(),
Y
Yu Yang 已提交
218
                   "Type %s's input %s is not set", Type(), in.name());
219 220 221 222
  }

  for (auto& out : op_info->Proto().outputs()) {
    PADDLE_ENFORCE(outputs_.find(out.name()) != outputs_.end(),
Y
Yu Yang 已提交
223
                   "Type %s's output %s is not set", Type(), out.name());
224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239
  }
}

void OperatorBase::GenerateTemporaryNames() {
  static std::atomic<size_t> gUniqId(0UL);
  for (auto& output : outputs_) {
    for (auto& output_name : output.second) {
      if (output_name == kTempVarName) {
        output_name += type_;
        output_name += "@";
        output_name += std::to_string(gUniqId.fetch_add(1));
      }
    }
  }
}

240 241 242 243
static bool VarIsTensor(const Variable* var) {
  return var->IsType<LoDTensor>() || var->IsType<SelectedRows>();
}

244
static const Tensor* GetTensorFromVar(Variable* var) {
Q
QI JUN 已提交
245
  if (var->IsType<LoDTensor>()) {
246
    return var->GetMutable<LoDTensor>();
Q
QI JUN 已提交
247
  } else if (var->IsType<SelectedRows>()) {
248
    return var->GetMutable<SelectedRows>()->mutable_value();
Q
QI JUN 已提交
249
  } else {
Y
Yang Yang 已提交
250 251
    PADDLE_THROW("Variable type_id %s, expect LoDTensor/SelectedRows.",
                 var->Type().name());
Q
QI JUN 已提交
252 253 254 255 256
  }
}

static Tensor* GetMutableTensorFromVar(Variable* var) {
  if (var->IsType<LoDTensor>()) {
257
    return var->GetMutable<LoDTensor>();
Q
QI JUN 已提交
258
  } else if (var->IsType<SelectedRows>()) {
259
    return var->GetMutable<SelectedRows>()->mutable_value();
Q
QI JUN 已提交
260
  } else {
Y
Yang Yang 已提交
261 262
    PADDLE_THROW("Variable type_id %s, expect LoDTensor/SelectedRows.",
                 var->Type().name());
Q
QI JUN 已提交
263 264 265
  }
}

266
template <>
267
const Tensor* ExecutionContext::Input<Tensor>(const std::string& name) const {
268
  auto* var = InputVar(name);
269 270
  return var == nullptr ? nullptr
                        : GetTensorFromVar(const_cast<Variable*>(var));
271 272 273
}

template <>
274
const std::vector<const Tensor*> ExecutionContext::MultiInput<Tensor>(
275 276 277 278
    const std::string& name) const {
  auto names = op().Inputs(name);
  std::vector<const Tensor*> res;
  res.reserve(names.size());
279 280 281 282 283
  std::transform(names.begin(), names.end(), std::back_inserter(res),
                 [&](const std::string& sub_name) {
                   auto var = scope_.FindVar(sub_name);
                   return var == nullptr ? nullptr : GetTensorFromVar(var);
                 });
284 285 286 287
  return res;
}

template <>
288
Tensor* ExecutionContext::Output<Tensor>(const std::string& name) const {
289
  auto var = OutputVar(name);
Q
QI JUN 已提交
290
  return var == nullptr ? nullptr : GetMutableTensorFromVar(var);
291 292 293
}

template <>
294
std::vector<Tensor*> ExecutionContext::MultiOutput<Tensor>(
295 296 297 298
    const std::string& name) const {
  auto names = op().Outputs(name);
  std::vector<Tensor*> res;
  res.reserve(names.size());
299 300
  std::transform(names.begin(), names.end(), std::back_inserter(res),
                 [&](const std::string& sub_name) {
301 302
                   auto var = scope_.FindVar(sub_name);
                   return var == nullptr ? nullptr
Q
QI JUN 已提交
303
                                         : GetMutableTensorFromVar(var);
304
                 });
305 306 307
  return res;
}

Y
Yu Yang 已提交
308 309 310 311 312
bool OpSupportGPU(const std::string& op_type) {
  auto& all_kernels = OperatorWithKernel::AllOpKernels();
  auto it = all_kernels.find(op_type);
  if (it == all_kernels.end()) {
    // All control operator must support GPU
313

Y
Yu Yang 已提交
314 315 316 317 318 319 320 321 322 323
    return true;
  }
  for (auto& kern_pair : it->second) {
    if (platform::is_gpu_place(kern_pair.first.place_)) {
      return true;
    }
  }
  return false;
}

324 325 326 327 328 329 330 331 332 333 334
class RuntimeInferShapeContext : public InferShapeContext {
 public:
  RuntimeInferShapeContext(const OperatorBase& op, const Scope& scope)
      : op_(op), scope_(scope) {}

  bool HasInput(const std::string& name) const override {
    auto& ins = Inputs(name);
    size_t length = ins.size();
    if (length == 0) {
      return false;
    }
F
fengjiayi 已提交
335 336
    PADDLE_ENFORCE_EQ(length, 1UL,
                      "Input %s should not have more than one inputs", name);
337 338 339 340 341 342 343 344 345 346 347
    auto ipt = ins[0];
    auto* var = ipt == kEmptyVarName ? nullptr : scope_.FindVar(ipt);
    return var != nullptr;
  }

  bool HasOutput(const std::string& name) const override {
    auto& outs = Outputs(name);
    size_t length = outs.size();
    if (length == 0) {
      return false;
    }
F
fengjiayi 已提交
348 349
    PADDLE_ENFORCE_EQ(length, 1UL,
                      "Output %s should not have more than one inputs", name);
350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392
    auto ipt = outs[0];
    auto* var = ipt == kEmptyVarName ? nullptr : scope_.FindVar(ipt);
    return var != nullptr;
  }

  bool HasInputs(const std::string& name) const override {
    auto inputs = op_.Inputs(name);
    if (inputs.empty()) {
      return false;
    }
    for (auto& input : inputs) {
      if (scope_.FindVar(input) == nullptr) {
        return false;
      }
    }
    return true;
  }

  bool HasOutputs(const std::string& name) const override {
    auto outputs = op_.Outputs(name);
    if (outputs.empty()) {
      return false;
    }
    for (auto& output : outputs) {
      if (scope_.FindVar(output) == nullptr) {
        return false;
      }
    }
    return true;
  }

  AttrReader Attrs() const override { return AttrReader(op_.Attrs()); }

  const std::vector<std::string>& Inputs(
      const std::string& name) const override {
    return op_.Inputs(name);
  }

  const std::vector<std::string>& Outputs(
      const std::string& name) const override {
    return op_.Outputs(name);
  }

Q
Qiao Longfei 已提交
393 394 395 396 397 398 399 400 401 402 403 404
  void ShareLoD(const std::string& in, const std::string& out, size_t i = 0,
                size_t j = 0) const override {
    PADDLE_ENFORCE_LT(i, Inputs(in).size());
    PADDLE_ENFORCE_LT(j, Outputs(out).size());
    Variable* in_var = scope_.FindVar(Inputs(in)[i]);
    Variable* out_var = scope_.FindVar(Outputs(out)[j]);
    if (!in_var->IsType<LoDTensor>()) return;
    PADDLE_ENFORCE(out_var->IsType<LoDTensor>(),
                   "The %d-th output of Output(%s) must be LoDTensor.", j, out);
    auto in_tensor = in_var->Get<LoDTensor>();
    auto* out_tensor = out_var->GetMutable<LoDTensor>();
    out_tensor->set_lod(in_tensor.lod());
D
dzhwinter 已提交
405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423

    // TODO(dzhwinter) : reuse ShareLoD in most operators.
    // Need to call ShareLayout explicitly in sequence related ops.
    // Shall we have a better method to shared info between in/out Tensor?
    out_tensor->set_layout(in_tensor.layout());
  }

  void ShareLayout(const std::string& in, const std::string& out, size_t i = 0,
                   size_t j = 0) const {
    PADDLE_ENFORCE_LT(i, Inputs(in).size());
    PADDLE_ENFORCE_LT(j, Outputs(out).size());
    Variable* in_var = scope_.FindVar(Inputs(in)[i]);
    Variable* out_var = scope_.FindVar(Outputs(out)[j]);
    if (!in_var->IsType<LoDTensor>()) return;
    PADDLE_ENFORCE(out_var->IsType<LoDTensor>(),
                   "The %d-th output of Output(%s) must be LoDTensor.", j, out);
    auto in_tensor = in_var->Get<LoDTensor>();
    auto* out_tensor = out_var->GetMutable<LoDTensor>();
    out_tensor->set_layout(in_tensor.layout());
Q
Qiao Longfei 已提交
424 425
  }

426 427 428
  bool IsRuntime() const override { return true; }

 protected:
429 430 431 432 433 434 435
  DDim GetDim(const std::string& name) const override {
    Variable* var = scope_.FindVar(name);
    if (var->IsType<LoDTensor>()) {
      return var->Get<LoDTensor>().dims();
    } else if (var->IsType<SelectedRows>()) {
      return var->Get<SelectedRows>().GetCompleteDims();
    } else {
F
fengjiayi 已提交
436 437 438 439 440 441 442
      PADDLE_THROW(
          "Only LoDTensor/SelectedRows support 'GetDim', but Variable %s's "
          "type_id is %s.",
          name, var->Type().name());
    }
  }

F
fengjiayi 已提交
443
  std::vector<DDim> GetRepeatedDims(const std::string& name) const override {
F
fengjiayi 已提交
444 445 446 447 448
    Variable* var = scope_.FindVar(name);
    if (var->IsType<ReaderHolder>()) {
      return var->Get<ReaderHolder>().shapes();
    } else {
      PADDLE_THROW(
F
fengjiayi 已提交
449
          "Only ReaderHolder support 'GetRepeatedDims', but Variable %s's "
F
fengjiayi 已提交
450 451
          "type_id is %s.",
          name, var->Type().name());
452 453 454 455 456 457 458 459 460 461
    }
  }

  void SetDim(const std::string& name, const DDim& dim) override {
    Variable* var = scope_.FindVar(name);
    if (var->IsType<LoDTensor>()) {
      var->GetMutable<LoDTensor>()->Resize(dim);
    } else if (var->IsType<SelectedRows>()) {
      var->GetMutable<SelectedRows>()->set_height(dim[0]);
    } else {
Y
Yang Yang 已提交
462 463
      PADDLE_THROW("Variable %s type_id %s, expect LoDTensor/SelectedRows.",
                   name, var->Type().name());
464 465 466
    }
  }

F
fengjiayi 已提交
467 468 469 470 471 472 473 474 475 476 477 478 479
  void SetRepeatedDims(const std::string& name,
                       const std::vector<DDim>& dims) override {
    Variable* var = scope_.FindVar(name);
    if (var->IsType<ReaderHolder>()) {
      var->GetMutable<ReaderHolder>()->set_shapes(dims);
    } else {
      PADDLE_THROW(
          "Only ReaderHolder support 'SetRepeatedDims', but Variable %s's "
          "type_id is %s.",
          name, var->Type().name());
    }
  }

480
  proto::VarType::Type GetVarType(const std::string& name) const override {
481 482 483 484
    auto* var = scope_.FindVar(name);
    return ToVarType(var->Type());
  }

F
fengjiayi 已提交
485 486 487 488
  InferShapeVarPtr GetVarPtr(const std::string& name) override {
    return scope_.FindVar(name);
  }

489
 private:
490 491 492 493
  const OperatorBase& op_;
  const Scope& scope_;
};

494 495
void OperatorWithKernel::RunImpl(const Scope& scope,
                                 const platform::Place& place) const {
496 497
  RuntimeInferShapeContext infer_shape_ctx(*this, scope);
  this->InferShape(&infer_shape_ctx);
Y
Yu Yang 已提交
498 499
  platform::DeviceContextPool& pool = platform::DeviceContextPool::Instance();
  auto dev_ctx = pool.Get(place);
500 501 502 503 504

  // check if op[type] has kernel registered.
  auto& all_op_kernels = AllOpKernels();
  auto kernels_iter = all_op_kernels.find(type_);
  if (kernels_iter == all_op_kernels.end()) {
Y
Yu Yang 已提交
505 506
    PADDLE_THROW(
        "There are no kernels which are registered in the %s operator.", type_);
507 508
  }

D
dzhwinter 已提交
509
  ExecutionContext ctx(*this, scope, *dev_ctx);
510

Q
qiaolongfei 已提交
511 512
  OpKernelMap& kernels = kernels_iter->second;

513 514
  // TODO(dzhwinter) : kernel fallback mechanism will be added when all the
  // transform functions are ready.
Q
qiaolongfei 已提交
515

516 517 518 519 520
  // for (auto& candidate : kKernelPriority) {
  //   Do selection
  // }

  auto expected_kernel_key = this->GetExpectedKernelType(ctx);
Q
qiaolongfei 已提交
521 522
  VLOG(3) << "expected_kernel_key:" << expected_kernel_key;

523 524 525 526 527 528 529
  auto kernel_iter = kernels.find(expected_kernel_key);
  if (kernel_iter == kernels.end()) {
    PADDLE_THROW("op %s does not have kernel for %s", type_,
                 KernelTypeToString(expected_kernel_key));
  }

  // do data transform
530 531 532 533 534 535 536 537 538 539
  Scope& new_scope = scope.NewScope();

  for (auto& var_name_item : this->Inputs()) {
    for (auto& var_name : var_name_item.second) {
      auto* var = scope.FindVar(var_name);
      if (var && VarIsTensor(var)) {
        auto* tensor_in = GetTensorFromVar(var);
        if (tensor_in->IsInitialized()) {
          auto kernel_type_for_var = this->GetKernelTypeForVar(
              var_name_item.first, *tensor_in, expected_kernel_key);
540
          if (TransFromNeeded(kernel_type_for_var, expected_kernel_key)) {
541 542 543 544 545 546 547 548
            auto out_var_names = OutputVars(true);
            if (std::find(out_var_names.begin(), out_var_names.end(),
                          var_name) != out_var_names.end()) {
              PADDLE_THROW(
                  "var %s is both input and output, "
                  "does not support transform",
                  var_name);
            }
549 550
            VLOG(3) << "Transform Variable " << var_name << " from "
                    << kernel_type_for_var << " to " << expected_kernel_key;
551
            auto* trans_var = new_scope.Var(var_name);
552 553 554 555
            std::shared_ptr<Tensor> out(new Tensor);
            DataTransform(expected_kernel_key, kernel_type_for_var, *tensor_in,
                          out.get());
            CopyVariableWithTensor(*var, *(out.get()), *trans_var);
556
          }
Q
QI JUN 已提交
557 558
        }
      }
559 560
    }
  }
Q
QI JUN 已提交
561

D
dzhwinter 已提交
562 563 564 565 566
  auto* new_dev_ctx = pool.Get(expected_kernel_key.place_);
  kernel_iter->second->Compute(
      ExecutionContext(*this, new_scope, *new_dev_ctx));

  /*For profiling/benchmark only*/
D
dzhwinter 已提交
567
  if (FLAGS_benchmark) {
D
dzhwinter 已提交
568 569
    new_dev_ctx->Wait();
  }
Q
Qiao Longfei 已提交
570 571
}

572
proto::DataType OperatorWithKernel::IndicateDataType(
Y
Yu Yang 已提交
573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597
    const ExecutionContext& ctx) const {
  auto& scope = ctx.scope();
  int data_type = -1;
  for (auto& input : this->inputs_) {
    for (auto& ipt_name : input.second) {
      auto* var = scope.FindVar(ipt_name);
      if (var != nullptr) {
        const Tensor* t = nullptr;
        if (var->IsType<Tensor>()) {
          t = &var->Get<Tensor>();
        } else if (var->IsType<LoDTensor>()) {
          t = &var->Get<LoDTensor>();
        } else if (var->IsType<SelectedRows>()) {
          t = &(var->Get<SelectedRows>().value());
        }
        if (t != nullptr) {
          int tmp = static_cast<int>(ToDataType(t->type()));
          PADDLE_ENFORCE(tmp == data_type || data_type == -1,
                         "DataType of Paddle Op %s must be the same.", Type());
          data_type = tmp;
        }
      }
    }
  }
  PADDLE_ENFORCE(data_type != -1, "DataType should be indicated by input");
598
  return static_cast<proto::DataType>(data_type);
Y
Yu Yang 已提交
599
}
600

601 602 603 604 605 606 607 608 609 610 611
OpKernelType OperatorWithKernel::GetExpectedKernelType(
    const ExecutionContext& ctx) const {
  return OpKernelType(IndicateDataType(ctx), ctx.GetPlace());
}

OpKernelType OperatorWithKernel::GetKernelTypeForVar(
    const std::string& var_name, const Tensor& tensor,
    const OpKernelType& expected_kernel_type) const {
  return OpKernelType(expected_kernel_type.data_type_, tensor.place());
}

Q
Qiao Longfei 已提交
612
}  // namespace framework
L
liaogang 已提交
613
}  // namespace paddle