op_registry.h 11.5 KB
Newer Older
1
/* Copyright (c) 2017 PaddlePaddle Authors. All Rights Reserved.
2

L
Luo Tao 已提交
3 4 5
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
6

L
Luo Tao 已提交
7
    http://www.apache.org/licenses/LICENSE-2.0
8

L
Luo Tao 已提交
9 10 11 12 13
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. */
14 15 16

#pragma once

17
#include <map>
18
#include <memory>
19 20
#include <string>
#include <tuple>
S
sneaxiy 已提交
21
#include <type_traits>
M
minqiyang 已提交
22 23
#include <unordered_map>
#include <unordered_set>
24
#include <vector>
25

Y
Yi Wang 已提交
26
#include "paddle/fluid/framework/grad_op_desc_maker.h"
D
dzhwinter 已提交
27
#include "paddle/fluid/framework/inplace_op_inference.h"
S
sneaxiy 已提交
28
#include "paddle/fluid/framework/no_need_buffer_vars_inference.h"
Y
Yi Wang 已提交
29 30 31 32
#include "paddle/fluid/framework/op_info.h"
#include "paddle/fluid/framework/op_proto_maker.h"
#include "paddle/fluid/framework/operator.h"
#include "paddle/fluid/framework/var_type_inference.h"
H
hong 已提交
33 34
#include "paddle/fluid/imperative/dygraph_grad_maker.h"
#include "paddle/fluid/imperative/type_defs.h"
35 36 37 38 39 40 41 42

namespace paddle {
namespace framework {
namespace details {

enum OpInfoFillType {
  kOperator = 0,
  kOpProtoAndCheckerMaker = 1,
Y
Yu Yang 已提交
43
  kGradOpDescMaker = 2,
44
  kVarTypeInference = 3,
D
dzhwinter 已提交
45
  kShapeInference = 4,
S
sneaxiy 已提交
46 47
  kInplaceOpInference = 5,
  kNoNeedBufferVarsInference = 6,
H
hong 已提交
48
  kGradOpBaseMaker = 7,
S
sneaxiy 已提交
49
  kUnknown = -1
50 51
};

S
sneaxiy 已提交
52 53 54 55 56 57 58 59 60 61 62
namespace internal {
template <typename T, OpInfoFillType kType>
struct TypePair {
  using Type = T;
  static constexpr OpInfoFillType kFillType = kType;
};

using OpRegistryClasses = std::tuple<                                // NOLINT
    TypePair<OperatorBase, kOperator>,                               // NOLINT
    TypePair<OpProtoAndCheckerMaker, kOpProtoAndCheckerMaker>,       // NOLINT
    TypePair<GradOpDescMakerBase, kGradOpDescMaker>,                 // NOLINT
H
hong 已提交
63
    TypePair<imperative::GradOpBaseMakerBase, kGradOpBaseMaker>,     // NOLINT
S
sneaxiy 已提交
64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86
    TypePair<VarTypeInference, kVarTypeInference>,                   // NOLINT
    TypePair<InferShapeBase, kShapeInference>,                       // NOLINT
    TypePair<InplaceOpInference, kInplaceOpInference>,               // NOLINT
    TypePair<NoNeedBufferVarsInference, kNoNeedBufferVarsInference>  // NOLINT
    >;

static constexpr int kOpRegistryClassNumber =
    std::tuple_size<OpRegistryClasses>::value;

template <typename T, int kPos, bool kIsBounded /* = true*/>
struct IsMatchedBaseTypeImpl {
  using PairType = typename std::tuple_element<kPos, OpRegistryClasses>::type;
  static constexpr bool kValue =
      std::is_base_of<typename PairType::Type, T>::value;
};

template <typename T, int kPos>
struct IsMatchedBaseTypeImpl<T, kPos, false> {
  static constexpr bool kValue = false;
};

template <typename T, int kPos>
static inline constexpr bool IsMatchedBaseType() {
87 88 89 90
  return IsMatchedBaseTypeImpl<T,
                               kPos,
                               (kPos >= 0 &&
                                kPos < kOpRegistryClassNumber)>::kValue;
S
sneaxiy 已提交
91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107
}

template <typename T, int kStart, int kEnd, bool kIsEnd, bool kIsMatched>
struct OpInfoFillTypeGetterImpl {};

// This case should not happen
template <typename T, int kStart, int kEnd>
struct OpInfoFillTypeGetterImpl<T, kStart, kEnd, true, true> {};

template <typename T, int kStart, int kEnd>
struct OpInfoFillTypeGetterImpl<T, kStart, kEnd, true, false> {
  static constexpr OpInfoFillType kType = kUnknown;
};

template <typename T, int kStart, int kEnd>
struct OpInfoFillTypeGetterImpl<T, kStart, kEnd, false, false> {
  static constexpr OpInfoFillType kType =
108 109 110 111
      OpInfoFillTypeGetterImpl<T,
                               kStart + 1,
                               kEnd,
                               kStart + 1 == kEnd,
S
sneaxiy 已提交
112 113 114 115 116 117 118 119 120 121 122
                               IsMatchedBaseType<T, kStart + 1>()>::kType;
};

template <typename T, int kStart, int kEnd>
struct OpInfoFillTypeGetterImpl<T, kStart, kEnd, false, true> {
  using PairType = typename std::tuple_element<kStart, OpRegistryClasses>::type;
  static constexpr OpInfoFillType kType = PairType::kFillType;
};

template <typename T>
using OpInfoFillTypeGetter =
123 124 125
    OpInfoFillTypeGetterImpl<T,
                             0,
                             kOpRegistryClassNumber,
S
sneaxiy 已提交
126 127 128 129 130
                             kOpRegistryClassNumber == 0,
                             IsMatchedBaseType<T, 0>()>;

}  // namespace internal

131 132 133
template <typename T>
struct OpInfoFillTypeID {
  static constexpr OpInfoFillType ID() {
S
sneaxiy 已提交
134
    return internal::OpInfoFillTypeGetter<T>::kType;
135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166
  }
};

template <typename T, OpInfoFillType = OpInfoFillTypeID<T>::ID()>
struct OpInfoFiller;

template <size_t I, bool at_end, typename... ARGS>
class OperatorRegistrarRecursive;

template <size_t I, typename... ARGS>
class OperatorRegistrarRecursive<I, false, ARGS...> {
 public:
  using T = typename std::tuple_element<I, std::tuple<ARGS...>>::type;
  OperatorRegistrarRecursive(const char* op_type, OpInfo* info) {
    OpInfoFiller<T> fill;
    fill(op_type, info);
    constexpr auto size = sizeof...(ARGS);
    OperatorRegistrarRecursive<I + 1, I + 1 == size, ARGS...> reg(op_type,
                                                                  info);
    (void)(reg);
  }
};

template <size_t I, typename... ARGS>
class OperatorRegistrarRecursive<I, true, ARGS...> {
 public:
  OperatorRegistrarRecursive(const char* op_type, OpInfo* info) {}
};

template <typename T>
struct OpInfoFiller<T, kOperator> {
  void operator()(const char* op_type, OpInfo* info) const {
167 168
    PADDLE_ENFORCE_EQ(info->creator_,
                      nullptr,
Z
Zeng Jinle 已提交
169 170
                      platform::errors::AlreadyExists(
                          "OpCreator of %s has been registered", op_type));
171 172
    info->creator_ = [](const std::string& type,
                        const VariableNameMap& inputs,
173 174 175 176
                        const VariableNameMap& outputs,
                        const AttributeMap& attrs) {
      return new T(type, inputs, outputs, attrs);
    };
Z
Zeng Jinle 已提交
177 178 179

    if (std::is_base_of<OperatorWithKernel, T>::value) {
      PADDLE_ENFORCE_EQ(
180 181
          info->infer_shape_,
          nullptr,
Z
Zeng Jinle 已提交
182 183 184
          platform::errors::AlreadyExists(
              "Duplicate InferShapeFN of %s has been registered", op_type));

185 186
      OperatorWithKernel* op = dynamic_cast<OperatorWithKernel*>(info->creator_(
          std::string{}, VariableNameMap{}, VariableNameMap{}, AttributeMap{}));
187 188 189
      PADDLE_ENFORCE_NOT_NULL(
          op,
          platform::errors::InvalidArgument("%s should have kernels", op_type));
Z
Zeng Jinle 已提交
190 191 192 193
      info->infer_shape_ = [op](InferShapeContext* ctx) {
        op->InferShape(ctx);
      };
    }
194 195 196 197 198 199
  }
};

template <typename T>
struct OpInfoFiller<T, kOpProtoAndCheckerMaker> {
  void operator()(const char* op_type, OpInfo* info) const {
200 201
    PADDLE_ENFORCE_EQ(info->proto_,
                      nullptr,
Z
Zeng Jinle 已提交
202
                      platform::errors::AlreadyExists(
203
                          "OpProto of %s has been registered.", op_type));
204 205
    PADDLE_ENFORCE_EQ(info->checker_,
                      nullptr,
Z
Zeng Jinle 已提交
206
                      platform::errors::AlreadyExists(
207
                          "OpAttrChecker of %s has been registered.", op_type));
208
    info->proto_ = new proto::OpProto;
209
    info->checker_ = new OpAttrChecker();
210
    info->proto_->set_type(op_type);
Y
Yu Yang 已提交
211
    T maker;
Y
yuyang18 已提交
212
    maker(info->proto_, info->checker_);
213
    PADDLE_ENFORCE_EQ(
214 215
        info->proto_->IsInitialized(),
        true,
216 217
        platform::errors::PreconditionNotMet(
            "Fail to initialize %s's OpProto, because %s is not initialized.",
218 219
            op_type,
            info->proto_->InitializationErrorString()));
220 221 222 223 224 225
  }
};

template <typename T>
struct OpInfoFiller<T, kGradOpDescMaker> {
  void operator()(const char* op_type, OpInfo* info) const {
Z
Zeng Jinle 已提交
226
    PADDLE_ENFORCE_EQ(
227 228
        info->grad_op_maker_,
        nullptr,
Z
Zeng Jinle 已提交
229 230 231
        platform::errors::AlreadyExists(
            "GradOpDescMaker of %s has been registered", op_type));

232 233 234 235 236 237 238 239
    info->grad_op_maker_ =
        [](const OpDesc& fwd_op,
           const std::unordered_set<std::string>& no_grad_set,
           std::unordered_map<std::string, std::string>* grad_to_var,
           const std::vector<BlockDesc*>& grad_block) {
          T maker(fwd_op, no_grad_set, grad_to_var, grad_block);
          return maker();
        };
S
sneaxiy 已提交
240 241

    info->use_default_grad_op_desc_maker_ =
H
hong 已提交
242
        std::is_base_of<DefaultGradOpMaker<OpDesc, true>, T>::value ||
243 244 245 246 247 248 249 250 251
        std::is_base_of<DefaultGradOpMaker<OpDesc, false>, T>::value ||
        std::is_base_of<DefaultGradOpMaker<imperative::OpBase, true>,
                        T>::value ||
        std::is_base_of<DefaultGradOpMaker<imperative::OpBase, false>,
                        T>::value;

    info->use_empty_grad_op_desc_maker_ =
        std::is_base_of<EmptyGradOpMaker<OpDesc>, T>::value ||
        std::is_base_of<EmptyGradOpMaker<imperative::OpBase>, T>::value;
H
hong 已提交
252 253 254 255 256 257
  }
};

template <typename T>
struct OpInfoFiller<T, kGradOpBaseMaker> {
  void operator()(const char* op_type, OpInfo* info) const {
Z
Zeng Jinle 已提交
258
    PADDLE_ENFORCE_EQ(
259 260
        info->dygraph_grad_op_maker_,
        nullptr,
Z
Zeng Jinle 已提交
261 262 263
        platform::errors::AlreadyExists(
            "GradOpBaseMaker of %s has been registered", op_type));

264 265 266 267 268 269 270 271 272 273 274
    info->dygraph_grad_op_maker_ =
        [](const std::string& type,
           const imperative::NameVarBaseMap& var_base_map_in,
           const imperative::NameVarBaseMap& var_base_map_out,
           const framework::AttributeMap& attrs,
           const framework::AttributeMap& default_attrs,
           const std::map<std::string, std::string>& inplace_map) {
          T maker(type, var_base_map_in, var_base_map_out, attrs, inplace_map);
          maker.SetDygraphDefaultAttrsMap(default_attrs);
          return maker();
        };
275 276
  }
};
Y
Yu Yang 已提交
277 278 279 280

template <typename T>
struct OpInfoFiller<T, kVarTypeInference> {
  void operator()(const char* op_type, OpInfo* info) const {
Z
Zeng Jinle 已提交
281
    PADDLE_ENFORCE_EQ(
282 283
        info->infer_var_type_,
        nullptr,
Z
Zeng Jinle 已提交
284 285
        platform::errors::AlreadyExists(
            "VarTypeInference of %s has been registered", op_type));
M
minqiyang 已提交
286
    info->infer_var_type_ = [](InferVarTypeContext* context) {
Y
Yu Yang 已提交
287
      T inference;
M
minqiyang 已提交
288
      inference(context);
Y
Yu Yang 已提交
289 290 291 292
    };
  }
};

293 294 295
template <typename T>
struct OpInfoFiller<T, kShapeInference> {
  void operator()(const char* op_type, OpInfo* info) const {
296 297
    // Note: if fill InferShapeFN by this Filler, the infershape here
    // will overwrite the op->InferShape func registered in kOperator Filler
298 299 300 301 302 303 304
    info->infer_shape_ = [](InferShapeContext* ctx) {
      T inference;
      inference(ctx);
    };
  }
};

D
dzhwinter 已提交
305 306 307
template <typename T>
struct OpInfoFiller<T, kInplaceOpInference> {
  void operator()(const char* op_type, OpInfo* info) const {
Z
Zeng Jinle 已提交
308
    PADDLE_ENFORCE_EQ(
309 310
        info->infer_inplace_,
        nullptr,
Z
Zeng Jinle 已提交
311 312
        platform::errors::AlreadyExists(
            "InplaceOpInference of %s has been registered", op_type));
313
    info->infer_inplace_ = [](bool use_cuda) {
D
dzhwinter 已提交
314
      T infer;
315
      return infer(use_cuda);
D
dzhwinter 已提交
316 317 318 319
    };
  }
};

S
sneaxiy 已提交
320 321 322
template <typename T>
struct OpInfoFiller<T, kNoNeedBufferVarsInference> {
  void operator()(const char* op_type, OpInfo* info) const {
Z
Zeng Jinle 已提交
323
    PADDLE_ENFORCE_EQ(
324 325
        info->infer_no_need_buffer_vars_,
        nullptr,
Z
Zeng Jinle 已提交
326 327
        platform::errors::AlreadyExists(
            "NoNeedBufferVarsInference of %s has been registered", op_type));
328
    info->infer_no_need_buffer_vars_.Reset(std::make_shared<T>());
S
sneaxiy 已提交
329 330 331
  }
};

332 333 334 335 336 337
// A fake OpInfoFiller of void
template <>
struct OpInfoFiller<void, kUnknown> {
  void operator()(const char* op_type, OpInfo* info) const {}
};

338 339 340 341
}  // namespace details

}  // namespace framework
}  // namespace paddle