paddle_analysis_config.h 9.2 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18
// Copyright (c) 2018 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

#include <cassert>
#include <memory>
#include <string>
19
#include <unordered_set>
20 21
#include <vector>

22 23
/*! \file */

24 25 26 27 28 29 30 31 32 33
// Here we include some header files with relative paths, for that in deploy,
// the abstract path of this header file will be changed.
#include "paddle_api.h"           // NOLINT
#include "paddle_pass_builder.h"  // NOLINT

namespace paddle {

class AnalysisPredictor;

// NOTE WIP, not stable yet.
34 35
struct AnalysisConfig {
  AnalysisConfig() = default;
36
  explicit AnalysisConfig(const AnalysisConfig& other);
37 38 39
  explicit AnalysisConfig(const std::string& model_dir);
  explicit AnalysisConfig(const std::string& prog_file,
                          const std::string& params_file);
N
nhzlx 已提交
40 41 42 43
  enum class Precision {
    kFloat32 = 0,
    kInt8,
  };
44

45 46
  /** Set model with a directory.
   */
47
  void SetModel(const std::string& model_dir) { model_dir_ = model_dir; }
48 49
  /** Set model with two specific pathes for program and parameters.
   */
50 51
  void SetModel(const std::string& prog_file_path,
                const std::string& params_file_path);
52 53
  /** Set program file path.
   */
54
  void SetProgFile(const std::string& x) { prog_file_ = x; }
55 56
  /** Set parameter composed file path.
   */
57
  void SetParamsFile(const std::string& x) { params_file_ = x; }
58 59
  /** Get the model directory path.
   */
60
  const std::string& model_dir() const { return model_dir_; }
61 62
  /** Get the program file path.
   */
63
  const std::string& prog_file() const { return prog_file_; }
64 65
  /** Get the composed parameters file.
   */
66 67 68
  const std::string& params_file() const { return params_file_; }

  // GPU related.
69 70 71 72 73 74

  /**
   * \brief Turn on GPU.
   * @param memory_pool_init_size_mb initial size of the GPU memory pool in MB.
   * @param device_id the GPU card to use (default is 0).
   */
75
  void EnableUseGpu(uint64_t memory_pool_init_size_mb, int device_id = 0);
76 77
  /** Turn off the GPU.
   */
78
  void DisableGpu();
79 80
  /** A bool state telling whether the GPU is turned on.
   */
81
  bool use_gpu() const { return use_gpu_; }
82 83
  /** Get the GPU device id.
   */
84
  int gpu_device_id() const { return device_id_; }
85 86
  /** Get the initial size in MB of the GPU memory pool.
   */
87
  int memory_pool_init_size_mb() const { return memory_pool_init_size_mb_; }
88 89
  /** Get the proportion of the initial memory pool size compared to the device.
   */
90
  float fraction_of_gpu_memory_for_pool() const;
91

92 93 94 95
  /** \brief Control whether to perform IR graph optimization.
   *
   * If turned off, the AnalysisConfig will act just like a NativeConfig.
   */
96
  void SwitchIrOptim(int x = true) { enable_ir_optim_ = x; }
97 98
  /** A boolean state tell whether the ir graph optimization is actived.
   */
99
  bool ir_optim() const { return enable_ir_optim_; }
100

101 102 103 104
  /** \brief INTERNAL Determine whether to use the feed and fetch operators.
   * Just for internal development, not stable yet.
   * When ZeroCopyTensor is used, this should turned off.
   */
105
  void SwitchUseFeedFetchOps(int x = true) { use_feed_fetch_ops_ = x; }
106 107
  /** A boolean state telling whether to use the feed and fetch operators.
   */
108
  bool use_feed_fetch_ops_enabled() const { return use_feed_fetch_ops_; }
109

110 111 112 113 114 115
  /** \brief Control whether to specify the inputs' names.
   *
   * The PaddleTensor type has a `name` member, assign it with the corresponding
   * variable name. This is used only when the input PaddleTensors passed to the
   * `PaddlePredictor.Run(...)` cannot follow the order in the training phase.
   */
116
  void SwitchSpecifyInputNames(bool x = true) { specify_input_name_ = x; }
117 118 119 120

  /** A boolean state tell whether the input PaddleTensor names specified should
   * be used to reorder the inputs in `PaddlePredictor.Run(...)`.
   */
121
  bool specify_input_name() const { return specify_input_name_; }
122

123 124 125 126 127 128 129 130 131 132 133 134 135
  /**
   * \brief Turn on the TensorRT engine.
   *
   * The TensorRT engine will accelerate some subgraphes in the original Fluid
   * computation graph. In some models such as TensorRT50, GoogleNet and so on,
   * it gains significant performance acceleration.
   *
   * @param workspace_size the memory size(in byte) used for TensorRT workspace.
   * @param max_batch_size the maximum batch size of this prediction task,
   * better set as small as possible, or performance loss.
   * @param min_subgrpah_size the minimum TensorRT subgraph size needed, if a
   * subgraph is less than this, it will not transfer to TensorRT engine.
   */
136
  void EnableTensorRtEngine(int workspace_size = 1 << 20,
N
nhzlx 已提交
137
                            int max_batch_size = 1, int min_subgraph_size = 3,
N
nhzlx 已提交
138 139
                            Precision precision = Precision::kFloat32,
                            bool use_static = true);
140 141
  /** A boolean state telling whether the TensorRT engine is used.
   */
142 143
  bool tensorrt_engine_enabled() const { return use_tensorrt_; }

Y
Yan Chunwei 已提交
144 145 146 147
  /** \brief Control whether to debug IR graph analysis phase.
   *
   * This will generate DOT files for visualizing the computation graph after
   * each analysis pass applied.
148
   */
Y
Yan Chunwei 已提交
149
  void SwitchIrDebug(int x = true);
150

151 152
  /** Turn on MKLDNN.
   */
L
luotao1 已提交
153
  void EnableMKLDNN();
154 155
  /** A boolean state telling whether to use the MKLDNN.
   */
156 157
  bool mkldnn_enabled() const { return use_mkldnn_; }

158 159
  /** Set and get the number of cpu math library threads.
   */
160
  void SetCpuMathLibraryNumThreads(int cpu_math_library_num_threads);
161 162
  /** An int state telling how many threads are used in the CPU math library.
   */
163 164 165 166
  int cpu_math_library_num_threads() const {
    return cpu_math_library_num_threads_;
  }

167 168
  /** Transform the AnalysisConfig to NativeConfig.
   */
Y
Yan Chunwei 已提交
169
  NativeConfig ToNativeConfig() const;
170 171 172
  /** Specify the operator type list to use MKLDNN acceleration.
   * @param op_list the operator type list.
   */
173 174 175
  void SetMKLDNNOp(std::unordered_set<std::string> op_list) {
    mkldnn_enabled_op_types_ = op_list;
  }
176

177 178 179 180 181 182
  /** Specify the memory buffer of program and parameter
   * @param prog_buffer the memory buffer of program.
   * @param prog_buffer_size the size of the data.
   * @param params_buffer the memory buffer of the composed parameters file.
   * @param params_buffer_size the size of the commposed parameters data.
   */
T
Tao Luo 已提交
183
  void SetModelBuffer(const char* prog_buffer, size_t prog_buffer_size,
184 185 186
                      const char* params_buffer, size_t params_buffer_size);
  /** A boolean state telling whether the model is set from the CPU memory.
   */
T
Tao Luo 已提交
187
  bool model_from_memory() const { return model_from_memory_; }
T
Tao Luo 已提交
188

Y
Yan Chunwei 已提交
189 190 191
  /** Turn on memory optimize
   * NOTE still in development, will release latter.
   */
Y
Yan Chunwei 已提交
192 193
  void EnableMemoryOptim(bool static_optim = false,
                         bool force_update_static_cache = false);
Y
Yan Chunwei 已提交
194 195 196
  /** Tell whether the memory optimization is activated. */
  bool enable_memory_optim() const;

197 198
  friend class ::paddle::AnalysisPredictor;

199 200 201
  /** NOTE just for developer, not an official API, easily to be broken.
   * Get a pass builder for customize the passes in IR analysis phase.
   */
202 203 204 205 206 207 208 209
  PassStrategy* pass_builder() const;

 protected:
  // Update the config.
  void Update();

  std::string SerializeInfoCache();

210
 protected:
211 212 213 214 215
  // Model pathes.
  std::string model_dir_;
  std::string prog_file_;
  std::string params_file_;

S
Sylwester Fraczek 已提交
216
  // GPU related.
217 218 219 220
  bool use_gpu_{false};
  int device_id_{0};
  uint64_t memory_pool_init_size_mb_{100};  // initial size is 100MB.

S
Sylwester Fraczek 已提交
221
  // TensorRT related.
222
  bool use_tensorrt_{false};
223 224
  // For workspace_size, refer it from here:
  // https://docs.nvidia.com/deeplearning/sdk/tensorrt-developer-guide/index.html#troubleshooting
225
  int tensorrt_workspace_size_;
226 227 228 229
  // While TensorRT allows an engine optimized for a given max batch size
  // to run at any smaller size, the performance for those smaller
  // sizes may not be as well-optimized. Therefore, Max batch is best
  // equivalent to the runtime batch size.
230
  int tensorrt_max_batchsize_;
231 232 233 234 235
  //  We transform the Ops that can be converted into TRT layer in the model,
  //  and aggregate these Ops into subgraphs for TRT execution.
  //  We set this variable to control the minimum number of nodes in the
  //  subgraph, 3 as default value.
  int tensorrt_min_subgraph_size_{3};
N
nhzlx 已提交
236
  Precision tensorrt_precision_mode_;
N
nhzlx 已提交
237
  bool trt_use_static_engine_;
238

Y
Yan Chunwei 已提交
239 240
  // memory reuse related.
  bool enable_memory_optim_{false};
Y
Yan Chunwei 已提交
241 242
  bool static_memory_optim_{false};
  bool static_memory_optim_force_update_{false};
Y
Yan Chunwei 已提交
243

244 245 246
  bool use_mkldnn_{false};
  std::unordered_set<std::string> mkldnn_enabled_op_types_;

T
Tao Luo 已提交
247
  bool model_from_memory_{false};
248

249 250 251 252 253 254 255 256 257 258 259 260
  bool enable_ir_optim_{true};
  bool use_feed_fetch_ops_{true};
  bool ir_debug_{false};

  bool specify_input_name_{false};

  int cpu_math_library_num_threads_{1};

  // A runtime cache, shouldn't be transferred to others.
  std::string serialized_info_cache_;

  mutable std::unique_ptr<PassStrategy> pass_builder_;
261 262 263
};

}  // namespace paddle