operator.cc 8.2 KB
Newer Older
Q
Qiao Longfei 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
/* 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. */

#include "paddle/framework/operator.h"
16 17
#include <algorithm>
#include "paddle/framework/op_registry.h"
Q
Qiao Longfei 已提交
18 19 20 21

namespace paddle {
namespace framework {

Q
qijun 已提交
22
template <>
23
Eigen::DefaultDevice& ExecutionContext::GetEigenDevice<
Q
qijun 已提交
24
    platform::CPUPlace, Eigen::DefaultDevice>() const {
D
dongzhihong 已提交
25
  return *device_context_->get_eigen_device<Eigen::DefaultDevice>();
Q
qijun 已提交
26 27 28 29
}

#ifndef PADDLE_ONLY_CPU
template <>
30
Eigen::GpuDevice&
31
ExecutionContext::GetEigenDevice<platform::GPUPlace, Eigen::GpuDevice>() const {
D
dongzhihong 已提交
32
  return *device_context_->get_eigen_device<Eigen::GpuDevice>();
Q
qijun 已提交
33 34 35
}
#endif

36
std::string OperatorBase::Input(const std::string& name) const {
Y
Yu Yang 已提交
37
  auto& ins = Inputs(name);
Y
Yu Yang 已提交
38 39 40 41
  PADDLE_ENFORCE_LE(ins.size(), 1UL,
                    "Op %s input %s should contain only one variable", type_,
                    name);
  return ins.empty() ? kEmptyVarName : ins[0];
Y
Yan Chunwei 已提交
42 43
}

Y
Yu Yang 已提交
44 45
const std::vector<std::string>& OperatorBase::Inputs(
    const std::string& name) const {
Y
Yu Yang 已提交
46 47 48 49
  auto it = inputs_.find(name);
  PADDLE_ENFORCE(it != inputs_.end(), "Op %s do not have input %s", type_,
                 name);
  return it->second;
Y
Yan Chunwei 已提交
50 51
}

52
std::string OperatorBase::Output(const std::string& name) const {
Y
Yu Yang 已提交
53
  auto& outs = Outputs(name);
Y
Yu Yang 已提交
54 55 56 57
  PADDLE_ENFORCE_LE(outs.size(), 1UL,
                    "Op %s output %s should contain only one variable", type_,
                    name);
  return outs.empty() ? kEmptyVarName : outs[0];
Y
Yan Chunwei 已提交
58 59
}

Y
Yu Yang 已提交
60 61
const std::vector<std::string>& OperatorBase::Outputs(
    const std::string& name) const {
Y
Yu Yang 已提交
62 63 64 65
  auto it = outputs_.find(name);
  PADDLE_ENFORCE(it != outputs_.end(), "Op %s does not have output %s", type_,
                 name);
  return it->second;
Y
Yan Chunwei 已提交
66 67
}

Q
Qiao Longfei 已提交
68 69
std::string OperatorBase::DebugString() const {
  std::stringstream ss;
Y
Yu Yang 已提交
70
  ss << "Op(" << type_ << "), inputs:{";
Y
Yu Yang 已提交
71 72
  for (auto it = inputs_.begin(); it != inputs_.end();) {
    auto& input = *it;
Y
Yu Yang 已提交
73 74 75 76 77 78
    ss << input.first << "[";
    for (size_t i = 0; i < input.second.size(); ++i) {
      ss << input.second[i];
      if (i != input.second.size() - 1) {
        ss << ", ";
      }
79
    }
Y
Yu Yang 已提交
80
    ss << "]";
Y
Yu Yang 已提交
81 82
    ++it;
    if (it != inputs_.end()) {
83 84
      ss << ", ";
    }
Q
Qiao Longfei 已提交
85
  }
Y
Yu Yang 已提交
86
  ss << "}, outputs:{";
Y
Yu Yang 已提交
87 88
  for (auto it = outputs_.begin(); it != outputs_.end();) {
    auto& output = *it;
Y
Yu Yang 已提交
89 90 91 92 93 94
    ss << output.first << "[";
    for (size_t i = 0; i < output.second.size(); ++i) {
      ss << output.second[i];
      if (i != output.second.size() - 1) {
        ss << ", ";
      }
95
    }
Y
Yu Yang 已提交
96
    ss << "]";
Y
Yu Yang 已提交
97 98
    ++it;
    if (it != outputs_.end()) {
99 100
      ss << ", ";
    }
Q
Qiao Longfei 已提交
101
  }
Y
Yu Yang 已提交
102
  ss << "}.";
Q
Qiao Longfei 已提交
103 104 105
  return ss.str();
}

D
dongzhihong 已提交
106 107
void OperatorBase::Rename(const std::string& old_name,
                          const std::string& new_name) {
Y
Yu Yang 已提交
108 109 110 111 112 113 114
  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
115 116
}

Y
Yu Yang 已提交
117
OperatorBase::OperatorBase(const std::string& type,
Y
Yu Yang 已提交
118 119
                           const VariableNameMap& inputs,
                           const VariableNameMap& outputs,
Y
Yu Yang 已提交
120 121
                           const AttributeMap& attrs)
    : type_(type), inputs_(inputs), outputs_(outputs), attrs_(attrs) {
122 123
  GenerateTemporaryNames();
  CheckAllInputOutputSet();
Y
Yu Yang 已提交
124
}
125

Q
qijun 已提交
126 127 128 129 130 131 132 133 134
std::vector<std::string> OperatorBase::InputVars() const {
  std::vector<std::string> 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
135 136 137 138 139 140 141 142 143 144
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
145
  auto& info = OpInfoMap::Instance().Get(Type());
Y
Yu Yang 已提交
146 147

  // get all OpProto::Var for outputs
Y
Yu Yang 已提交
148
  for (auto& o : info.Proto().outputs()) {
Y
Yu Yang 已提交
149 150 151 152 153 154 155 156 157
    // 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
158 159
}

160 161 162
void OperatorBase::CheckAllInputOutputSet() const {
  auto& info_map = OpInfoMap::Instance();
  auto* op_info = info_map.GetNullable(Type());
Y
Yu Yang 已提交
163
  if (op_info == nullptr || op_info->proto_ == nullptr) return;
164 165 166

  for (auto& in : op_info->Proto().inputs()) {
    PADDLE_ENFORCE(inputs_.find(in.name()) != inputs_.end(),
Y
Yu Yang 已提交
167
                   "Type %s's input %s is not set", Type(), in.name());
168 169 170 171
  }

  for (auto& out : op_info->Proto().outputs()) {
    PADDLE_ENFORCE(outputs_.find(out.name()) != outputs_.end(),
Y
Yu Yang 已提交
172
                   "Type %s's output %s is not set", Type(), out.name());
173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188
  }
}

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));
      }
    }
  }
}

189 190 191
template <>
const Tensor* InferShapeContext::Input<Tensor>(const std::string& name) const {
  auto* var = InputVar(name);
192
  return var == nullptr ? nullptr : GetTensorFromVar(var);
193 194 195 196 197 198 199 200
}

template <>
const std::vector<const Tensor*> InferShapeContext::MultiInput<Tensor>(
    const std::string& name) const {
  auto names = op().Inputs(name);
  std::vector<const Tensor*> res;
  res.reserve(names.size());
201 202 203 204 205
  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);
                 });
206 207 208 209 210 211 212
  return res;
}

template <>
Tensor* ExecutionContext::Output<Tensor>(const std::string& name) const {
  auto* var = OutputVar(name);
  if (var == nullptr) return nullptr;
213
  return GetTensorFromVar(var);
214 215 216 217 218 219 220 221
}

template <>
std::vector<Tensor*> ExecutionContext::MultiOutput<Tensor>(
    const std::string& name) const {
  auto names = op().Outputs(name);
  std::vector<Tensor*> res;
  res.reserve(names.size());
222 223 224 225 226
  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);
                 });
227 228 229
  return res;
}

230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267
void OpProtoAndCheckerMaker::Validate() {
  validated_ = true;
  CheckNoDuplicatedInOutAttrs();
}

OpProtoAndCheckerMaker::VariableBuilder OpProtoAndCheckerMaker::AddInput(
    const std::string& name, const std::string& comment) {
  auto* input = proto_->add_inputs();
  input->set_name(name);
  input->set_comment(comment);
  return OpProtoAndCheckerMaker::VariableBuilder{input};
}

OpProtoAndCheckerMaker::VariableBuilder OpProtoAndCheckerMaker::AddOutput(
    const std::string& name, const std::string& comment) {
  auto* output = proto_->add_outputs();
  output->set_name(name);
  output->set_comment(comment);
  return OpProtoAndCheckerMaker::VariableBuilder{output};
}

void OpProtoAndCheckerMaker::CheckNoDuplicatedInOutAttrs() {
  std::unordered_set<std::string> names;
  auto checker = [&](const std::string& name) {
    PADDLE_ENFORCE(!names.count(name), "[%s] is duplicated", name);
    names.insert(name);
  };
  for (auto& attr : proto_->attrs()) {
    checker(attr.name());
  }
  for (auto& input : proto_->inputs()) {
    checker(input.name());
  }
  for (auto& output : proto_->outputs()) {
    checker(output.name());
  }
}

Q
Qiao Longfei 已提交
268
}  // namespace framework
L
liaogang 已提交
269
}  // namespace paddle