parallel_executor.h 2.3 KB
Newer Older
Y
Yang Yang 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
/* Copyright (c) 2016 PaddlePaddle Authors. All Rights Reserved.

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

17
#include <string>
Y
Yang Yang 已提交
18
#include <unordered_set>
19
#include <vector>
Y
Yang Yang 已提交
20 21 22 23 24 25 26 27 28 29
#include "paddle/fluid/framework/executor.h"
#include "paddle/fluid/framework/op_info.h"
#include "paddle/fluid/framework/program_desc.h"
#include "paddle/fluid/framework/scope.h"
#include "paddle/fluid/framework/tensor.h"
#include "paddle/fluid/platform/device_context.h"

namespace paddle {
namespace framework {

30
class ParallelExecutorPrivate;
Y
Use mtx  
Yu Yang 已提交
31

Y
Yang Yang 已提交
32
class ParallelExecutor {
Y
Yu Yang 已提交
33 34
  DISABLE_COPY_AND_ASSIGN(ParallelExecutor);

35
 public:
36
  explicit ParallelExecutor(size_t num_threads, bool use_event,
Y
Yu Yang 已提交
37
                            const std::vector<platform::Place>& places,
38
                            const std::unordered_set<std::string>& params,
39
                            const std::unordered_set<std::string>& bcast_vars,
40
                            const ProgramDesc& main_program,
41
                            const std::string& loss_var_name, Scope* scope,
42
                            const std::vector<Scope*>& local_scopes,
43
                            bool allow_op_delay, bool customize_scale_loss);
44

45 46
  ~ParallelExecutor();

47 48
  std::vector<Scope*>& GetLocalScopes();

49 50 51 52 53 54 55 56 57 58
  /**
   * Feed tensors to local scopes. The size of tensors should be equal to the
   * size of local scopes.
   */
  void FeedTensorsIntoLocalScopes(
      const std::vector<std::unordered_map<std::string, LoDTensor>>& tensors);

  void FeedAndSplitTensorIntoLocalScopes(
      const std::unordered_map<std::string, LoDTensor>& tensors);

Y
Yu Yang 已提交
59
  void Run(const std::vector<std::string>& fetch_tensors,
60
           const std::string& fetched_var_name);
Y
Yang Yang 已提交
61

T
typhoonzero 已提交
62 63
  void BCastParamsToGPUs(const std::unordered_set<std::string>& vars) const;

Y
Yang Yang 已提交
64
 private:
65
  ParallelExecutorPrivate* member_;
Y
Yang Yang 已提交
66 67 68 69
};

}  // namespace framework
}  // namespace paddle
反馈
建议
客服 返回
顶部