api_anakin_engine.cc 17.2 KB
Newer Older
1
// Copyright (c) 2019 PaddlePaddle Authors. All Rights Reserved.
Y
Yan Chunwei 已提交
2 3 4 5 6 7 8 9 10 11 12 13 14
//
// 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.

T
Tao Luo 已提交
15 16 17
#include <map>
#include <string>
#include <utility>
L
Luo Tao 已提交
18
#include <vector>
Y
Yan Chunwei 已提交
19

20 21 22
#include "paddle/fluid/inference/api/api_anakin_engine.h"
#include "paddle/fluid/inference/api/paddle_api.h"

T
Tao Luo 已提交
23 24 25 26
#include "framework/core/net/net.h"
#include "framework/operators/ops.h"
#include "saber/funcs/timer.h"

Y
Yan Chunwei 已提交
27 28
namespace paddle {

Y
Yan Chunwei 已提交
29
using paddle::contrib::AnakinConfig;
30 31 32 33
template <typename T, Precision P, OpRunType R>
extern std::mutex PaddleInferenceAnakinPredictor<T, P, R>::mutex_;
template <typename T, Precision P, OpRunType R>
extern std::once_flag PaddleInferenceAnakinPredictor<T, P, R>::init_anakin_;
Y
Yan Chunwei 已提交
34

35 36 37 38 39
template <typename T, Precision P, OpRunType R>
void PaddleInferenceAnakinPredictor<T, P, R>::InitEnv() {
  std::call_once(this->init_anakin_, [this]() {
    anakin::Env<T>::env_init(this->config_.max_stream);
  });
石晓伟 已提交
40
  anakin::TargetWrapper<T>::set_device(this->config_.device_id);
Y
Yan Chunwei 已提交
41
}
42 43 44
template <typename T, Precision P, OpRunType R>
void PaddleInferenceAnakinPredictor<T, P, R>::InitNet() {
  std::unique_lock<std::mutex> lock(this->mutex_);
45
  delete this->executor_p_;
46
  this->executor_p_ = new anakin::Net<T, P, R>(*this->graph_p_, true);
T
Tao Luo 已提交
47
}
48 49 50 51 52 53 54 55 56 57
template <typename T, Precision P, OpRunType R>
void PaddleInferenceAnakinPredictor<T, P, R>::SetContext() {
  this->ctx_p_ = std::make_shared<anakin::Context<T>>(
      this->config_.device_id, this->config_.data_stream_id,
      this->config_.compute_stream_id);
}
template <typename T, Precision P, OpRunType R>
void PaddleInferenceAnakinPredictor<T, P, R>::InitGraph() {
  this->graph_p_ =
      std::make_shared<anakin::graph::Graph<T, anakin::Precision::FP32>>();
石晓伟 已提交
58 59 60 61 62 63 64
  if (!this->config_.model_file.empty()) {
    this->graph_p_->load(this->config_.model_file);
  } else if (this->config_.model_buf_p) {
    this->graph_p_->load(this->config_.model_buf_p,
                         this->config_.model_buf_len);
  } else {
    LOG(FATAL) << "Model load error.";
C
cuichaowen 已提交
65
  }
66 67 68
  this->input_names_ = this->graph_p_->get_ins();
  this->output_names_ = this->graph_p_->get_outs();
  for (auto &input_str : this->input_names_) {
69 70
    if (this->config_.init_inputs_shape.find(input_str) ==
        this->config_.init_inputs_shape.end()) {
石晓伟 已提交
71
      LOG(FATAL) << input_str << " should be set in init_inputs_shape.";
72 73 74 75
    }
    std::vector<int> shape =
        this->config_.init_inputs_shape.find(input_str)->second;
    this->graph_p_->Reshape(input_str, shape);
C
cuichaowen 已提交
76
  }
77 78 79 80 81
}
template <typename T, Precision P, OpRunType R>
void PaddleInferenceAnakinPredictor<T, P, R>::OptimizeGraph() {
  if (!this->graph_p_->Optimize()) {
    LOG(FATAL) << "Graph optimization error.";
C
cuichaowen 已提交
82
  }
83 84 85 86 87 88 89 90 91 92
}
template <typename T, Precision P, OpRunType R>
void PaddleInferenceAnakinPredictor<T, P, R>::InitPredictor() {
  this->InitEnv();
  this->SetContext();
  this->InitGraph();
  this->OptimizeGraph();
  this->InitNet();
}
template <typename T, Precision P, OpRunType R>
93
void PaddleInferenceAnakinPredictor<T, P, R>::Predict(int batch_size) {
94 95 96 97 98 99 100 101 102
  anakin::TargetWrapper<T>::device_sync();
  this->executor_p_->prediction();
  anakin::TargetWrapper<T>::device_sync();
}
template <typename T, Precision P, OpRunType R>
bool PaddleInferenceAnakinPredictor<T, P, R>::Run(
    const std::vector<PaddleTensor> &inputs,
    std::vector<PaddleTensor> *output_data, int batch_size) {
  if (this->config_.re_allocable) {
103
    return this->RunImpl(inputs, output_data, batch_size);
104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191
  } else {
    // Run inputs data that exceeds batch size in batches.
    // 1. Reassign the batch size.
    if (batch_size == -1) {
      if (!inputs[0].lod.empty()) {
        batch_size = inputs[0].lod[0].size() - 1;
      } else {
        batch_size = inputs[0].shape[0];
      }
    }
    // 2. If the data don't need to be batched, run it directly.
    if (batch_size <= this->config_.init_batch_size) {
      return this->RunImpl(inputs, output_data);
    }
    // 3. Check the batch size and define temporary variables.
    std::vector<PaddleTensor> cur_inputs;
    std::vector<PaddleTensor> outputs_master;
    std::vector<std::vector<paddle::PaddleTensor>> outputs_vec;
    for (const auto &input : inputs) {
      if (!input.lod.empty()) {
        if (input.lod.size() != 1) {
          return false;
        }
        if (input.lod[0].size() - 1 != batch_size) {
          return false;
        }
      } else {
        LOG(INFO) << "Non-lod mode to be implemented.";
        return false;
      }
      PaddleTensor tensor;
      tensor.name = input.name;
      tensor.dtype = PaddleDType::FLOAT32;
      cur_inputs.push_back(tensor);
    }
    for (auto output : *output_data) {
      PaddleTensor tensor;
      tensor.name = output.name;
      outputs_master.push_back(tensor);
    }
    // 4. Batch execution.
    for (size_t start_batch = 0; start_batch < batch_size;) {
      auto end_batch = start_batch + this->config_.init_batch_size;
      if (end_batch > batch_size) {
        end_batch = batch_size;
      }
      auto cur_outputs = outputs_master;
      for (size_t i = 0; i < inputs.size(); i++) {
        auto start = inputs[i].lod[0][start_batch];
        auto end = inputs[i].lod[0][end_batch];
        std::vector<size_t> offsets;
        for (size_t j = start_batch; j <= end_batch; j++) {
          offsets.push_back(inputs[i].lod[0][j] -
                            inputs[i].lod[0][start_batch]);
        }
        auto mem_start = static_cast<float *>(inputs[i].data.data()) + start;
        cur_inputs[i].data =
            PaddleBuf(mem_start, (end - start) * sizeof(float));
        cur_inputs[i].lod = std::vector<std::vector<size_t>>({offsets});
        cur_inputs[i].shape =
            std::vector<int>({static_cast<int>(end - start), 1, 1, 1});
      }
      if (!this->RunImpl(cur_inputs, &cur_outputs)) {
        return false;
      }
      outputs_vec.push_back(cur_outputs);
      start_batch = end_batch;
    }
    // 5. Copy the results to contiguous memory.
    // Assume that each batch has the same final outputs size.
    auto count = [](const std::vector<int> &v) {
      int cnt = 1;
      for_each(v.begin(), v.end(), [&cnt](int n) { cnt *= n; });
      return cnt;
    };
    for (size_t i = 0; i < output_data->size(); i++) {
      std::vector<int> shape = outputs_vec[i][0].shape;
      shape[0] = batch_size;
      int total_cnt = count(shape);
      (*output_data)[i].shape = shape;
      (*output_data)[i].data.Resize(total_cnt * sizeof(float));
      float *addr = static_cast<float *>((*output_data)[i].data.data());
      for (const auto &single_out : outputs_vec) {
        int cnt = count(single_out[i].shape);
        memcpy(addr, single_out[i].data.data(), cnt * sizeof(float));
        addr += cnt;
      }
    }
C
cuichaowen 已提交
192
  }
Y
Yan Chunwei 已提交
193 194
  return true;
}
195 196
template <typename T, Precision P, OpRunType R>
bool PaddleInferenceAnakinPredictor<T, P, R>::RunImpl(
Y
Yan Chunwei 已提交
197
    const std::vector<PaddleTensor> &inputs,
198
    std::vector<PaddleTensor> *output_data, int batch_size) {
石晓伟 已提交
199
  anakin::TargetWrapper<T>::set_device(this->config_.device_id);
Y
Yan Chunwei 已提交
200 201
  for (const auto &input : inputs) {
    if (input.dtype != PaddleDType::FLOAT32) {
202 203
      LOG(FATAL) << "Only support float type inputs. " << input.name
                 << "'s type is not float";
Y
Yan Chunwei 已提交
204
    }
205
    auto d_tensor_p = this->executor_p_->get_in(input.name);
206
    auto net_shape = d_tensor_p->valid_shape();
C
cuichaowen 已提交
207
    if (net_shape.size() != input.shape.size()) {
208 209
      LOG(FATAL) << " input  " << input.name
                 << "'s shape size should be equal to that of net";
C
cuichaowen 已提交
210
    }
211
#ifndef ANAKIN_MLU_PLACE
C
cuichaowen 已提交
212 213 214
    int sum = 1;
    for_each(input.shape.begin(), input.shape.end(), [&](int n) { sum *= n; });
    if (sum > net_shape.count()) {
215 216 217 218 219 220 221 222 223
      if (this->config_.re_allocable) {
        this->graph_p_->Reshape(input.name, input.shape);
        this->InitNet();
        d_tensor_p = this->executor_p_->get_in(input.name);
      } else {
        LOG(FATAL)
            << "Run failed because Anakin was expected not to reallocate "
               "memory.";
      }
C
cuichaowen 已提交
224
    }
225
#endif
226
    std::vector<int> tmp_shape;
C
cuichaowen 已提交
227 228 229
    for (auto s : input.shape) {
      tmp_shape.push_back(s);
    }
230 231 232 233
    auto *data = static_cast<float *>(input.data.data());
    anakin::saber::Tensor<typename anakin::DefaultHostType<T>::Host_type>
        h_tensor(data, typename anakin::DefaultHostType<T>::Host_type(), 0,
                 tmp_shape);
234
#ifndef ANAKIN_MLU_PLACE
235
    d_tensor_p->reshape(tmp_shape);
236
#endif
T
Tao Luo 已提交
237 238
    if (input.lod.size() > 0) {
      if (input.lod.size() > 1) {
239 240
        LOG(FATAL) << " input lod first dim should <=1, but you set "
                   << input.lod.size();
T
Tao Luo 已提交
241
      }
242 243 244 245 246 247
      std::vector<int> lod(input.lod[0].begin(), input.lod[0].end());
      std::vector<std::vector<int>> offset({lod});
      d_tensor_p->set_seq_offset(offset);
      VLOG(3) << "offset.size(): " << offset[0].size();
      for (int i = 0; i < offset[0].size(); i++) {
        VLOG(3) << offset[0][i];
T
Tao Luo 已提交
248 249
      }
    }
250
    d_tensor_p->copy_from(h_tensor);
Y
Yan Chunwei 已提交
251
  }
252
  this->Predict(batch_size);
Y
Yan Chunwei 已提交
253
  if (output_data->empty()) {
254
    LOG(FATAL) << "The output param in the Run function is incorrect.";
Y
Yan Chunwei 已提交
255 256
  }
  for (auto &output : *output_data) {
257 258 259 260
    if (std::find(this->output_names_.begin(), this->output_names_.end(),
                  output.name) == this->output_names_.end()) {
      LOG(FATAL) << output.name << " is not in the outputs of the graph.";
    }
261
    auto *d_tensor_p = this->executor_p_->get_out(output.name);
262 263 264 265 266 267 268
    auto tmp_shape = d_tensor_p->valid_shape();
#ifdef ANAKIN_MLU_PLACE
    tmp_shape.set_num(batch_size);
#endif
    output.shape = tmp_shape;
    if (output.data.length() < tmp_shape.count() * sizeof(float)) {
      output.data.Resize(tmp_shape.count() * sizeof(float));
Y
Yan Chunwei 已提交
269
    }
270 271 272
    auto *data = static_cast<float *>(output.data.data());
    anakin::saber::Tensor<typename anakin::DefaultHostType<T>::Host_type>
        h_tensor(data, typename anakin::DefaultHostType<T>::Host_type(), 0,
273
                 tmp_shape);
274
    h_tensor.copy_from(*d_tensor_p);
Y
Yan Chunwei 已提交
275 276 277
  }
  return true;
}
278
template <typename T, Precision P, OpRunType R>
279 280 281 282 283 284
bool PaddleInferenceAnakinPredictor<T, P, R>::Reset(
    PaddleInferenceAnakinPredictor<T, P, R> *predictor) {
  this->config_ = predictor->GetConfig();
  this->graph_p_ = predictor->GetGraph();
  this->input_names_ = predictor->GetInputNames();
  this->output_names_ = predictor->GetOutputNames();
285 286 287 288
  this->ctx_p_ = std::make_shared<anakin::Context<T>>(
      this->config_.device_id, this->config_.data_stream_id,
      this->config_.compute_stream_id);
  this->InitNet();
289 290 291 292 293 294 295
  return true;
}
template <typename T, Precision P, OpRunType R>
std::unique_ptr<PaddlePredictor>
PaddleInferenceAnakinPredictor<T, P, R>::New() {
  return std::unique_ptr<PaddlePredictor>(
      new PaddleInferenceAnakinPredictor<T, P, R>());
C
cuichaowen 已提交
296 297 298
}
// the cloned new Predictor of anakin share the same net weights from original
// Predictor
299
template <typename T, Precision P, OpRunType R>
C
cuichaowen 已提交
300
std::unique_ptr<PaddlePredictor>
301
PaddleInferenceAnakinPredictor<T, P, R>::Clone() {
C
cuichaowen 已提交
302
  VLOG(3) << "Anakin Predictor::clone";
303
  std::unique_ptr<PaddlePredictor> cls = std::move(this->New());
C
cuichaowen 已提交
304
  auto anakin_predictor_p =
305
      dynamic_cast<PaddleInferenceAnakinPredictor<T, P, R> *>(cls.get());
C
cuichaowen 已提交
306
  if (!anakin_predictor_p) {
307
    LOG(FATAL) << "fail to call Init";
C
cuichaowen 已提交
308
  }
309
  anakin_predictor_p->Reset(this);
310 311
  return cls;
}
C
cuichaowen 已提交
312

313 314
#ifdef ANAKIN_MLU_PLACE
template <Precision P, OpRunType R>
315 316 317 318 319 320
std::unique_ptr<PaddlePredictor>
PaddleInferenceAnakinMLUPredictor<P, R>::New() {
  return std::unique_ptr<PaddlePredictor>(
      new PaddleInferenceAnakinMLUPredictor<P, R>());
}
template <Precision P, OpRunType R>
321 322 323 324 325 326
void PaddleInferenceAnakinMLUPredictor<P, R>::SetContext() {
  this->ctx_p_ = std::make_shared<anakin::Context<anakin::MLU>>(
      this->config_.device_id, this->config_.data_stream_id,
      this->config_.compute_stream_id);
  this->ctx_p_->set_model_parallel(this->config_.model_parallel);
  this->ctx_p_->set_fusion(this->config_.op_fuse);
327 328
  this->ctx_p_->enable_batch_changable();
  this->ctx_p_->enable_channel_duplicate();
Y
Yan Chunwei 已提交
329
}
330 331 332 333 334 335 336 337 338
template <Precision P, OpRunType R>
void PaddleInferenceAnakinMLUPredictor<P, R>::OptimizeGraph() {
  if (!this->graph_p_->fusion_optimize(this->config_.op_fuse)) {
    LOG(FATAL) << "Graph optimization error.";
  }
}
template <Precision P, OpRunType R>
void PaddleInferenceAnakinMLUPredictor<P, R>::InitNet() {
  std::unique_lock<std::mutex> lock(this->mutex_);
339
  delete this->executor_p_;
340 341 342 343
  this->executor_p_ = new anakin::Net<anakin::MLU, P, R>();
  this->executor_p_->fusion_init(*this->graph_p_, this->ctx_p_, true);
}
template <Precision P, OpRunType R>
344 345
void PaddleInferenceAnakinMLUPredictor<P, R>::Predict(int batch_size) {
  this->executor_p_->fusion_prediction(batch_size);
346 347
}
#endif
Y
Yan Chunwei 已提交
348

石晓伟 已提交
349 350
#ifdef ANAKIN_BM_PLACE
template <Precision P, OpRunType R>
351 352 353 354 355
std::unique_ptr<PaddlePredictor> PaddleInferenceAnakinBMPredictor<P, R>::New() {
  return std::unique_ptr<PaddlePredictor>(
      new PaddleInferenceAnakinBMPredictor<P, R>());
}
template <Precision P, OpRunType R>
石晓伟 已提交
356 357 358 359 360 361 362 363
void PaddleInferenceAnakinBMPredictor<P, R>::OptimizeGraph() {
  if (!this->graph_p_->fusion_optimize()) {
    LOG(FATAL) << "Graph optimization error.";
  }
}
template <Precision P, OpRunType R>
void PaddleInferenceAnakinBMPredictor<P, R>::InitNet() {
  std::unique_lock<std::mutex> lock(this->mutex_);
364
  delete this->executor_p_;
石晓伟 已提交
365 366 367 368
  this->executor_p_ = new anakin::Net<anakin::BM, P, R>();
  this->executor_p_->fusion_init(*this->graph_p_, this->ctx_p_, true);
}
template <Precision P, OpRunType R>
369
void PaddleInferenceAnakinBMPredictor<P, R>::Predict(int batch_size) {
石晓伟 已提交
370 371 372 373
  this->executor_p_->fusion_prediction();
}
#endif

374
#ifdef PADDLE_WITH_CUDA
375 376 377 378 379 380 381 382 383 384
template class PaddleInferenceAnakinPredictor<
    anakin::NV, anakin::Precision::FP32, ::anakin::OpRunType::ASYNC>;
#endif
#ifdef ANAKIN_X86_PLACE
template class PaddleInferenceAnakinPredictor<
    anakin::X86, anakin::Precision::FP32, ::anakin::OpRunType::ASYNC>;
#endif
#ifdef ANAKIN_MLU_PLACE
template class PaddleInferenceAnakinMLUPredictor<anakin::Precision::FP32,
                                                 ::anakin::OpRunType::SYNC>;
385
#endif
石晓伟 已提交
386 387 388 389
#ifdef ANAKIN_BM_PLACE
template class PaddleInferenceAnakinBMPredictor<anakin::Precision::FP32,
                                                ::anakin::OpRunType::ASYNC>;
#endif
C
cuichaowen 已提交
390

Y
Yan Chunwei 已提交
391 392
// A factory to help create difference predictor.
template <>
Y
Yan Chunwei 已提交
393 394 395
std::unique_ptr<PaddlePredictor>
CreatePaddlePredictor<contrib::AnakinConfig, PaddleEngineKind::kAnakin>(
    const contrib::AnakinConfig &config) {
396
#ifdef PADDLE_WITH_CUDA
397 398 399 400 401
  if (config.target_type == contrib::AnakinConfig::NVGPU) {
    return std::unique_ptr<PaddlePredictor>(
        new PaddleInferenceAnakinPredictor<anakin::NV, anakin::Precision::FP32,
                                           ::anakin::OpRunType::ASYNC>(config));
  }
402
#endif
403 404 405 406 407
#ifdef ANAKIN_X86_PLACE
  if (config.target_type == contrib::AnakinConfig::X86) {
    return std::unique_ptr<PaddlePredictor>(
        new PaddleInferenceAnakinPredictor<anakin::X86, anakin::Precision::FP32,
                                           ::anakin::OpRunType::ASYNC>(config));
C
cuichaowen 已提交
408
  }
409 410 411 412 413 414 415 416 417
#endif
#ifdef ANAKIN_MLU_PLACE
  if (config.target_type == contrib::AnakinConfig::MLU) {
    return std::unique_ptr<PaddlePredictor>(
        new PaddleInferenceAnakinMLUPredictor<anakin::Precision::FP32,
                                              ::anakin::OpRunType::SYNC>(
            config));
  }
#endif
石晓伟 已提交
418 419 420 421 422 423 424 425 426 427
#ifdef ANAKIN_BM_PLACE
  if (config.target_type == contrib::AnakinConfig::BM) {
    return std::unique_ptr<PaddlePredictor>(
        new PaddleInferenceAnakinBMPredictor<anakin::Precision::FP32,
                                             ::anakin::OpRunType::ASYNC>(
            config));
  }
#endif
  LOG(FATAL) << "Anakin Predictor create on unknown platform: "
             << config.target_type;
428
  return nullptr;
T
Tao Luo 已提交
429
}
430 431
template <typename T, Precision P, OpRunType R>
void DisplayOpTimer(anakin::Net<T, P, R> *net_executor, int epoch) {
T
Tao Luo 已提交
432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452
#ifdef PADDLE_ANAKIN_ENABLE_OP_TIMER
  std::vector<float> op_time = net_executor->get_op_time();
  auto exec_funcs = net_executor->get_exec_funcs();
  auto op_param = net_executor->get_op_param();
  for (int i = 0; i < op_time.size(); i++) {
    LOG(INFO) << "name: " << exec_funcs[i].name
              << " op_type: " << exec_funcs[i].op_name
              << " op_param: " << op_param[i] << " time " << op_time[i] / epoch;
  }
  std::map<std::string, float> op_map;
  for (int i = 0; i < op_time.size(); i++) {
    auto it = op_map.find(op_param[i]);
    if (it != op_map.end())
      op_map[op_param[i]] += op_time[i];
    else
      op_map.insert(std::pair<std::string, float>(op_param[i], op_time[i]));
  }
  for (auto it = op_map.begin(); it != op_map.end(); ++it) {
    LOG(INFO) << it->first << "  " << (it->second) / epoch << " ms";
  }
#endif
453 454 455 456 457 458
}
template <typename T, Precision P, OpRunType R>
PaddleInferenceAnakinPredictor<T, P, R>::~PaddleInferenceAnakinPredictor() {
  DisplayOpTimer<T, P, R>(this->executor_p_, this->config_.init_batch_size);
  delete this->executor_p_;
  this->executor_p_ = nullptr;
T
Tao Luo 已提交
459
}
Y
Yan Chunwei 已提交
460 461

}  // namespace paddle