layer.h 13.1 KB
Newer Older
1
// Copyright (c) 2019 PaddlePaddle Authors. All Rights Reserved.
X
xiexionghang 已提交
2 3 4 5 6 7 8 9 10 11 12 13 14 15
//
// 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.

#pragma once
16 17
#include <algorithm>
#include <atomic>
X
xiexionghang 已提交
18
#include <cstdint>
19
#include <list>
X
xiexionghang 已提交
20 21 22 23 24 25 26
#include <map>     // NOLINT
#include <memory>  // NOLINT
#include <mutex>   // NOLINT
#include <set>
#include <string>         // NOLINT
#include <unordered_map>  // NOLINT
#include <utility>
27
#include <vector>
X
xiexionghang 已提交
28 29 30
#include "paddle/fluid/framework/op_desc.h"
#include "paddle/fluid/framework/operator.h"
#include "paddle/fluid/framework/var_type_inference.h"
31
#include "paddle/fluid/framework/variable.h"
X
xiexionghang 已提交
32
#include "paddle/fluid/imperative/flags.h"
33 34 35
#include "paddle/fluid/imperative/type_defs.h"
#include "paddle/fluid/platform/enforce.h"
#include "paddle/fluid/platform/macros.h"
X
xiexionghang 已提交
36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55

namespace paddle {
namespace imperative {

class OpBase;

class ThreadSafeNameSet {
 public:
  void Insert(const std::string& name);

  void Remove(const std::string& name);

  std::vector<std::string> Names() const;

 private:
  std::multiset<std::string> set_;
  mutable std::mutex mtx_;
};

class VarBase {
56 57
  DISABLE_COPY_AND_ASSIGN(VarBase);

X
xiexionghang 已提交
58 59
 public:
  static std::vector<std::string> AliveVarNames();
60
  explicit VarBase(bool has_grad, const std::string& name)
X
xiexionghang 已提交
61
      : name_(name),
62
        grad_var_(has_grad ? new VarBase(false, GradVarName()) : nullptr) {
X
xiexionghang 已提交
63
    if (IsDebugEnabled()) {
64
      VLOG(10) << "Construct VarBase: " << name;
X
xiexionghang 已提交
65 66 67 68
      name_set_.Insert(name_);
    }
  }

69 70 71 72
  explicit VarBase(const std::string& name) : VarBase(true, name) {}

  ~VarBase() {
    VLOG(10) << "Destruct VarBase: " << name_;
X
xiexionghang 已提交
73 74 75 76 77
    if (IsDebugEnabled()) {
      name_set_.Remove(name_);
    }
  }

78
  const framework::Variable& Var() const { return var_; }
X
xiexionghang 已提交
79

80
  framework::Variable* MutableVar() { return &var_; }
X
xiexionghang 已提交
81

82
  bool HasGradVar() const { return grad_var_ != nullptr; }
X
xiexionghang 已提交
83

84
  const std::shared_ptr<VarBase>& GradVarBase() const { return grad_var_; }
X
xiexionghang 已提交
85

86 87 88
  const framework::Variable& GradVar() const {
    PADDLE_ENFORCE_NOT_NULL(grad_var_, "Gradient of %s does not exist", name_);
    return grad_var_->var_;
X
xiexionghang 已提交
89 90
  }

91 92 93
  framework::Variable* MutableGradVar() {
    PADDLE_ENFORCE_NOT_NULL(grad_var_, "Gradient of %s does not exist", name_);
    return &(grad_var_->var_);
X
xiexionghang 已提交
94 95
  }

96 97 98 99
  // This is used for python api
  void SetOverridedStopGradient(bool stop_gradient) {
    if (stop_gradient) {
      overrided_stop_gradient_ = 1;
X
xiexionghang 已提交
100
    } else {
101
      overrided_stop_gradient_ = 0;
X
xiexionghang 已提交
102
    }
103 104
    if (grad_var_) {
      grad_var_->SetOverridedStopGradient(stop_gradient);
X
xiexionghang 已提交
105 106
    }
  }
107 108 109 110 111 112
  // This is used for python api
  bool OverridedStopGradient() const {
    if (overrided_stop_gradient_ == 0) {
      return false;
    } else {
      return true;
X
xiexionghang 已提交
113 114 115
    }
  }

116 117
  // This is used inside C++
  int InnerOverridedStopGradient() const { return overrided_stop_gradient_; }
X
xiexionghang 已提交
118

119
  bool GradGenerated() const { return grad_generated_; }
X
xiexionghang 已提交
120

121 122 123 124 125 126 127 128 129 130 131 132
  void SetGradGenerated(bool generated) { grad_generated_ = generated; }
  // This is used inside C++
  void InnerSetOverridedStopGradient(bool stop_gradient) {
    if (overrided_stop_gradient_ == -1) {
      overrided_stop_gradient_ = static_cast<int>(stop_gradient);
      if (grad_var_) {
        grad_var_->InnerSetOverridedStopGradient(stop_gradient);
      }
    } else {
      VLOG(6) << "Ignore Stop gradient conversion for Var: " << Name()
              << "Set value is: " << overrided_stop_gradient_;
    }
X
xiexionghang 已提交
133 134
  }

135
  void SetPersistable(bool persistable) { persistable_ = persistable; }
X
xiexionghang 已提交
136

137
  bool Persistable() const { return persistable_; }
X
xiexionghang 已提交
138

139
  void AddGradOps(const std::weak_ptr<OpBase>& op);
X
xiexionghang 已提交
140

141 142 143 144 145 146
  std::vector<OpBase*> GradOps() {
    std::vector<OpBase*> rlt;
    // TODO(jiabin): use better data structure to remove nullptr when we find it
    for (const auto& wk_ptr : grad_ops_) {
      OpBase* tmp_op = wk_ptr.lock().get();
      if (tmp_op) rlt.emplace_back(tmp_op);
X
xiexionghang 已提交
147
    }
148
    return rlt;
X
xiexionghang 已提交
149
  }
150
  void ClearGradOps() { grad_ops_.clear(); }
X
xiexionghang 已提交
151

152 153 154 155 156 157
  const std::string& Name() const { return name_; }

  void SetName(const std::string& name) {
    name_ = name;
    if (grad_var_) {
      grad_var_->SetName(GradVarName());
X
xiexionghang 已提交
158 159 160
    }
  }

161
  std::string GradVarName() { return framework::GradVarName(name_); }
X
xiexionghang 已提交
162

163
  void SetType(framework::proto::VarType::Type type) { type_ = type; }
X
xiexionghang 已提交
164

165 166 167 168 169 170 171 172
  framework::proto::VarType::Type Type() const { return type_; }

  void SetDataType(framework::proto::VarType::Type data_type) {
    data_type_ = data_type;
    if (grad_var_) {
      grad_var_->SetDataType(data_type_);
    }
  }
X
xiexionghang 已提交
173

174
  framework::proto::VarType::Type DataType() const { return data_type_; }
X
xiexionghang 已提交
175

176
  void ClearGradient();
X
xiexionghang 已提交
177

178 179
  std::shared_ptr<VarBase> NewVarBase(const platform::Place& dst_place,
                                      const bool blocking) const;
X
xiexionghang 已提交
180

181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198
 private:
  framework::Variable var_;
  std::string name_;
  std::shared_ptr<VarBase> grad_var_;
  mutable size_t copied_counter_ = 0;

  // grad_op indicates which grad_op will this var be used as input
  std::vector<std::weak_ptr<OpBase>> grad_ops_;
  // add this property for users may set stop_gradient themselves and this
  // should override the
  // frameworks setting (-1) unset, (1) true, (0) false
  int overrided_stop_gradient_{-1};
  bool grad_generated_{false};
  bool persistable_{false};

  framework::proto::VarType::Type type_{framework::proto::VarType::LOD_TENSOR};
  framework::proto::VarType::Type data_type_{framework::proto::VarType::FP32};
  static ThreadSafeNameSet name_set_;
X
xiexionghang 已提交
199 200 201 202 203 204 205 206
};

class Layer {
 public:
  virtual ~Layer() {}

  virtual std::vector<std::shared_ptr<VarBase>> Forward(
      const std::vector<std::shared_ptr<VarBase>>& inputs) {
207
    return {};
X
xiexionghang 已提交
208 209 210 211
  }
};

// infer var type context for imperative mode
212
class RuntimeInferVarTypeContext : public framework::InferVarTypeContext {
X
xiexionghang 已提交
213
 public:
214 215 216
  RuntimeInferVarTypeContext(const NameVarBaseMap& inputs,
                             const NameVarBaseMap* outputs,
                             const framework::AttributeMap& attrs_map)
X
xiexionghang 已提交
217 218 219 220 221 222 223
      : InferVarTypeContext(nullptr, nullptr),
        inputs_(inputs),
        outputs_(outputs),
        attrs_(attrs_map),
        input_names_(),
        output_names_(),
        var_set_() {
224 225 226
    input_names_.reserve(inputs_.size());
    for (auto& it : inputs_) {
      for (auto& var : it.second) {
X
xiexionghang 已提交
227
        input_names_[it.first].emplace_back(var->Name());
228
        var_set_[var->Name()] = var.get();
X
xiexionghang 已提交
229 230 231 232 233
      }
    }

    output_names_.reserve(outputs_->size());
    for (auto& it : *outputs_) {
234
      for (auto& var : it.second) {
X
xiexionghang 已提交
235
        output_names_[it.first].emplace_back(var->Name());
236
        var_set_[var->Name()] = var.get();
X
xiexionghang 已提交
237 238 239 240 241 242 243
      }
    }
  }

  virtual ~RuntimeInferVarTypeContext() {}

  framework::Attribute GetAttr(const std::string& name) const override {
244 245 246 247
    auto iter = attrs_.find(name);
    PADDLE_ENFORCE_EQ(iter != attrs_.end(), true, "Cannot find attribute %s",
                      name);
    return iter->second;
X
xiexionghang 已提交
248 249 250 251 252 253 254
  }

  bool HasVar(const std::string& name) const override {
    return var_set_.count(name) > 0;
  }

  bool HasInput(const std::string& name) const override {
255
    return inputs_.count(name) > 0;
X
xiexionghang 已提交
256 257 258 259 260 261 262 263 264
  }

  bool HasOutput(const std::string& name) const override {
    PADDLE_ENFORCE_NOT_NULL(outputs_);
    return outputs_->count(name) > 0;
  }

  const std::vector<std::string>& Input(
      const std::string& name) const override {
265 266 267 268
    auto iter = input_names_.find(name);
    PADDLE_ENFORCE_EQ(iter != input_names_.end(), true, "Cannot find input %s",
                      name);
    return iter->second;
X
xiexionghang 已提交
269 270 271 272
  }

  const std::vector<std::string>& Output(
      const std::string& name) const override {
273 274 275 276
    auto iter = output_names_.find(name);
    PADDLE_ENFORCE_EQ(iter != output_names_.end(), true,
                      "Cannot find output %s", name);
    return iter->second;
X
xiexionghang 已提交
277 278 279 280
  }

  framework::proto::VarType::Type GetType(
      const std::string& name) const override {
281 282 283 284
    auto iter = var_set_.find(name);
    PADDLE_ENFORCE_EQ(iter != var_set_.end(), true,
                      "Cannot find var %s in GetType", name);
    return iter->second->Type();
X
xiexionghang 已提交
285 286 287 288 289 290 291 292 293 294 295 296 297
  }

  void SetType(const std::string& name,
               framework::proto::VarType::Type type) override {
    if (name == "kLookupTablePath") {
      VLOG(2) << "SUPER UGLY FIX, remove this when move imperative mode in C++";
    } else {
      var_set_[name]->SetType(type);
    }
  }

  framework::proto::VarType::Type GetDataType(
      const std::string& name) const override {
298 299 300 301
    auto iter = var_set_.find(name);
    PADDLE_ENFORCE_EQ(iter != var_set_.end(), true,
                      "Cannot find var %s in GetDataType", name);
    return iter->second->DataType();
X
xiexionghang 已提交
302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337
  }

  void SetDataType(const std::string& name,
                   framework::proto::VarType::Type type) override {
    var_set_[name]->SetDataType(type);
  }

  std::vector<framework::proto::VarType::Type> GetDataTypes(
      const std::string& name) const override {
    PADDLE_THROW("GetDataTypes is not supported in runtime InferVarType");
  }

  void SetDataTypes(const std::string& name,
                    const std::vector<framework::proto::VarType::Type>&
                        multiple_data_type) override {
    PADDLE_THROW("SetDataTypes is not supported in runtime InferVarType");
  }

  std::vector<int64_t> GetShape(const std::string& name) const override {
    PADDLE_THROW("Do not handle Shape in runtime InferVarType");
  }

  void SetShape(const std::string& name,
                const std::vector<int64_t>& dims) override {
    PADDLE_THROW("Do not handle Shape in runtime InferVarType");
  }

  int32_t GetLoDLevel(const std::string& name) const override {
    PADDLE_THROW("Do not handle LoDLevel in runtime InferVarType");
  }

  void SetLoDLevel(const std::string& name, int32_t lod_level) override {
    PADDLE_THROW("Do not handle LoDLevel in runtime InferVarType");
  }

 private:
338 339 340
  const NameVarBaseMap& inputs_;
  const NameVarBaseMap* outputs_;
  const framework::AttributeMap& attrs_;
X
xiexionghang 已提交
341 342
  std::unordered_map<std::string, std::vector<std::string>> input_names_;
  std::unordered_map<std::string, std::vector<std::string>> output_names_;
343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428
  std::unordered_map<std::string, VarBase*> var_set_;
};

// TODO(zjl): to support py_func layer
class OpBase : public std::enable_shared_from_this<OpBase> {
  DISABLE_COPY_AND_ASSIGN(OpBase);

 public:
  ~OpBase() { VLOG(3) << "Destruct Op: " << Type() << std::endl; }

  // Developer should not rely on this method to create OpBase.
  // OpBase should be created in Tracer and managed by Tracer totally.
  template <typename... Args>
  static std::shared_ptr<OpBase> Create(Args&&... args) {
    return std::shared_ptr<OpBase>(new OpBase(std::forward<Args>(args)...));
  }

  size_t id() const { return id_; }

  const std::string& Type() const { return op_->Type(); }

  void Run(const NameVarBaseMap& ins, const NameVarBaseMap& outs);

  const framework::VariableNameMap& InputNameMap() const {
    return op_->Inputs();
  }

  const framework::VariableNameMap& OutputNameMap() const {
    return op_->Outputs();
  }

  const framework::AttributeMap& Attrs() const { return op_->Attrs(); }
  const framework::OpInfo& Info() const { return op_->Info(); }

  void ClearBackwardTrace();

  const std::vector<OpBase*>& GradPendingOps() const {
    return grad_pending_ops_;
  }

  void InsertGradPendingOps(OpBase* op) { grad_pending_ops_.emplace_back(op); }

  void SortGradPendingOps() {
    std::sort(grad_pending_ops_.begin(), grad_pending_ops_.end(),
              [](OpBase* op1, OpBase* op2) { return op1->id() > op2->id(); });
  }
  NameVarBaseMap* GetMutableOutsMap() { return &outs_; }
  NameVarBaseMap* GetMutableInsMap() { return &ins_; }
  const NameVarBaseMap& GetInsMap() { return ins_; }
  const NameVarBaseMap& GetOutsMap() { return outs_; }
  const platform::Place& place() const { return place_; }

  // TODO(jiabin) prepare for backward hook
  void RegisterBackwardHooks(const std::function<void()>& func) {
    backward_hooks_.emplace_back(func);
  }

  void InvokeBackwardHooks() {
    for (const auto& func : backward_hooks_) {
      func();
      VLOG(5) << "Invoke Backward Hook for: " << Type() << std::endl;
    }
  }

 private:
  OpBase(size_t id, const std::string& type, const NameVarBaseMap& ins,
         const NameVarBaseMap& outs, framework::AttributeMap attrs,
         const platform::Place& place);

  OpBase(size_t id, const framework::OpDesc& op_desc,
         const platform::Place& place);

  size_t id_;

  std::unique_ptr<framework::OperatorBase> op_;

  std::vector<std::function<void()>> backward_hooks_;
  platform::Place place_;

  // Not need to be std::weak_ptr, because op is binded to a certain Tracer,
  // and would not be used by a Tracer that does not create itself.
  std::vector<OpBase*> grad_pending_ops_;

  // This part is only used for backward
  NameVarBaseMap ins_;
  NameVarBaseMap outs_;
X
xiexionghang 已提交
429 430 431 432
};

}  // namespace imperative
}  // namespace paddle