general_model.cpp 8.9 KB
Newer Older
G
guru4elephant 已提交
1 2 3 4 5 6 7 8 9 10 11 12 13 14
// Copyright (c) 2019 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.

G
guru4elephant 已提交
15
#include "core/general-client/include/general_model.h"
M
MRXLT 已提交
16
#include <fstream>
G
guru4elephant 已提交
17 18 19
#include "core/sdk-cpp/builtin_format.pb.h"
#include "core/sdk-cpp/include/common.h"
#include "core/sdk-cpp/include/predictor_sdk.h"
G
guru4elephant 已提交
20 21 22 23 24 25 26 27 28 29 30

using baidu::paddle_serving::predictor::general_model::Request;
using baidu::paddle_serving::predictor::general_model::Response;
using baidu::paddle_serving::predictor::general_model::Tensor;
using baidu::paddle_serving::predictor::general_model::FeedInst;
using baidu::paddle_serving::predictor::general_model::FetchInst;

namespace baidu {
namespace paddle_serving {
namespace general_model {

M
MRXLT 已提交
31
void PredictorClient::init(const std::string &conf_file) {
G
guru4elephant 已提交
32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47
  _conf_file = conf_file;
  std::ifstream fin(conf_file);
  if (!fin) {
    LOG(ERROR) << "Your inference conf file can not be found";
    exit(-1);
  }
  _feed_name_to_idx.clear();
  _fetch_name_to_idx.clear();
  _shape.clear();
  int feed_var_num = 0;
  int fetch_var_num = 0;
  fin >> feed_var_num >> fetch_var_num;
  std::string name;
  std::string fetch_var_name;
  int shape_num = 0;
  int dim = 0;
G
guru4elephant 已提交
48
  int type_value = 0;
G
guru4elephant 已提交
49 50 51 52 53 54 55 56 57
  for (int i = 0; i < feed_var_num; ++i) {
    fin >> name;
    _feed_name_to_idx[name] = i;
    fin >> shape_num;
    std::vector<int> tmp_feed_shape;
    for (int j = 0; j < shape_num; ++j) {
      fin >> dim;
      tmp_feed_shape.push_back(dim);
    }
G
guru4elephant 已提交
58 59
    fin >> type_value;
    _type.push_back(type_value);
G
guru4elephant 已提交
60 61 62 63 64 65 66 67 68 69 70
    _shape.push_back(tmp_feed_shape);
  }

  for (int i = 0; i < fetch_var_num; ++i) {
    fin >> name;
    fin >> fetch_var_name;
    _fetch_name_to_idx[name] = i;
    _fetch_name_to_var_name[name] = fetch_var_name;
  }
}

M
MRXLT 已提交
71 72
void PredictorClient::set_predictor_conf(const std::string &conf_path,
                                         const std::string &conf_file) {
G
guru4elephant 已提交
73 74 75 76 77 78 79 80 81 82 83 84
  _predictor_path = conf_path;
  _predictor_conf = conf_file;
}

int PredictorClient::create_predictor() {
  if (_api.create(_predictor_path.c_str(), _predictor_conf.c_str()) != 0) {
    LOG(ERROR) << "Predictor Creation Failed";
    return -1;
  }
  _api.thrd_initialize();
}

M
MRXLT 已提交
85 86 87 88 89 90 91
std::vector<std::vector<float>> PredictorClient::predict(
    const std::vector<std::vector<float>> &float_feed,
    const std::vector<std::string> &float_feed_name,
    const std::vector<std::vector<int64_t>> &int_feed,
    const std::vector<std::string> &int_feed_name,
    const std::vector<std::string> &fetch_name) {
  std::vector<std::vector<float>> fetch_result;
G
guru4elephant 已提交
92 93 94 95 96 97 98 99 100
  if (fetch_name.size() == 0) {
    return fetch_result;
  }
  fetch_result.resize(fetch_name.size());

  _api.thrd_clear();
  _predictor = _api.fetch_predictor("general_model");
  Request req;
  std::vector<Tensor *> tensor_vec;
M
MRXLT 已提交
101 102
  FeedInst *inst = req.add_insts();
  for (auto &name : float_feed_name) {
G
guru4elephant 已提交
103 104 105
    tensor_vec.push_back(inst->add_tensor_array());
  }

M
MRXLT 已提交
106
  for (auto &name : int_feed_name) {
G
guru4elephant 已提交
107 108 109 110
    tensor_vec.push_back(inst->add_tensor_array());
  }

  int vec_idx = 0;
M
MRXLT 已提交
111
  for (auto &name : float_feed_name) {
G
guru4elephant 已提交
112
    int idx = _feed_name_to_idx[name];
M
MRXLT 已提交
113
    Tensor *tensor = tensor_vec[idx];
G
guru4elephant 已提交
114 115 116 117 118
    for (int j = 0; j < _shape[idx].size(); ++j) {
      tensor->add_shape(_shape[idx][j]);
    }
    tensor->set_elem_type(1);
    for (int j = 0; j < float_feed[vec_idx].size(); ++j) {
M
MRXLT 已提交
119 120 121
      tensor->add_data(const_cast<char *>(reinterpret_cast<const char *>(
                           &(float_feed[vec_idx][j]))),
                       sizeof(float));
G
guru4elephant 已提交
122 123 124 125 126
    }
    vec_idx++;
  }

  vec_idx = 0;
M
MRXLT 已提交
127
  for (auto &name : int_feed_name) {
G
guru4elephant 已提交
128
    int idx = _feed_name_to_idx[name];
M
MRXLT 已提交
129
    Tensor *tensor = tensor_vec[idx];
G
guru4elephant 已提交
130 131 132 133 134
    for (int j = 0; j < _shape[idx].size(); ++j) {
      tensor->add_shape(_shape[idx][j]);
    }
    tensor->set_elem_type(0);
    for (int j = 0; j < int_feed[vec_idx].size(); ++j) {
M
MRXLT 已提交
135 136 137
      tensor->add_data(const_cast<char *>(reinterpret_cast<const char *>(
                           &(int_feed[vec_idx][j]))),
                       sizeof(int64_t));
G
guru4elephant 已提交
138 139 140 141 142 143 144 145 146 147 148 149
    }
    vec_idx++;
  }

  // std::map<std::string, std::vector<float> > result;
  Response res;

  res.Clear();
  if (_predictor->inference(&req, &res) != 0) {
    LOG(ERROR) << "failed call predictor with req: " << req.ShortDebugString();
    exit(-1);
  } else {
M
MRXLT 已提交
150
    for (auto &name : fetch_name) {
G
guru4elephant 已提交
151 152 153 154 155 156 157 158 159 160 161 162 163 164
      int idx = _fetch_name_to_idx[name];
      int len = res.insts(0).tensor_array(idx).data_size();
      VLOG(3) << "fetch name: " << name;
      VLOG(3) << "tensor data size: " << len;
      fetch_result[idx].resize(len);
      for (int i = 0; i < len; ++i) {
        /*
        (*fetch_result)[name][i] = *(const float *)
                    res.insts(0).tensor_array(idx).data(i).c_str();
        VLOG(3) << *(const float *)
            res.insts(0).tensor_array(idx).data(i).c_str();
        fetch_result[name][i] = *(const float *)
                    res.insts(0).tensor_array(idx).data(i).c_str();
        */
M
MRXLT 已提交
165 166
        fetch_result[idx][i] =
            *(const float *)res.insts(0).tensor_array(idx).data(i).c_str();
G
guru4elephant 已提交
167 168 169 170 171 172 173
      }
    }
  }

  return fetch_result;
}

M
MRXLT 已提交
174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272
std::vector<std::vector<std::vector<float>>> PredictorClient::predict_for_batch(
    const std::vector<std::vector<std::vector<float>>> &float_feed_batch,
    const std::vector<std::string> &float_feed_name,
    const std::vector<std::vector<std::vector<int64_t>>> &int_feed_batch,
    const std::vector<std::string> &int_feed_name,
    const std::vector<std::string> &fetch_name,
    const int64_t &batch_size) {
  std::vector<std::vector<std::vector<float>>> fetch_result_batch;
  if (fetch_name.size() == 0) {
    return fetch_result_batch;
  }
  fetch_result_batch.resize(batch_size);
  int fetch_name_num = fetch_name.size();
  for (int bi = 0; bi < batch_size; bi++) {
    fetch_result_batch[bi].resize(fetch_name_num);
  }

  _api.thrd_clear();
  _predictor = _api.fetch_predictor("general_model");
  Request req;
  //
  for (int bi = 0; bi < batch_size; bi++) {
    std::vector<Tensor *> tensor_vec;
    FeedInst *inst = req.add_insts();
    std::vector<std::vector<float>> float_feed = float_feed_batch[bi];
    std::vector<std::vector<int64_t>> int_feed = int_feed_batch[bi];
    for (auto &name : float_feed_name) {
      tensor_vec.push_back(inst->add_tensor_array());
    }

    for (auto &name : int_feed_name) {
      tensor_vec.push_back(inst->add_tensor_array());
    }

    int vec_idx = 0;
    for (auto &name : float_feed_name) {
      int idx = _feed_name_to_idx[name];
      Tensor *tensor = tensor_vec[idx];
      for (int j = 0; j < _shape[idx].size(); ++j) {
        tensor->add_shape(_shape[idx][j]);
      }
      tensor->set_elem_type(1);
      for (int j = 0; j < float_feed[vec_idx].size(); ++j) {
        tensor->add_data(const_cast<char *>(reinterpret_cast<const char *>(
                             &(float_feed[vec_idx][j]))),
                         sizeof(float));
      }
      vec_idx++;
    }

    vec_idx = 0;
    for (auto &name : int_feed_name) {
      int idx = _feed_name_to_idx[name];
      Tensor *tensor = tensor_vec[idx];
      for (int j = 0; j < _shape[idx].size(); ++j) {
        tensor->add_shape(_shape[idx][j]);
      }
      tensor->set_elem_type(0);
      for (int j = 0; j < int_feed[vec_idx].size(); ++j) {
        tensor->add_data(const_cast<char *>(reinterpret_cast<const char *>(
                             &(int_feed[vec_idx][j]))),
                         sizeof(int64_t));
      }
      vec_idx++;
    }
  }

  Response res;

  res.Clear();
  if (_predictor->inference(&req, &res) != 0) {
    LOG(ERROR) << "failed call predictor with req: " << req.ShortDebugString();
    exit(-1);
  } else {
    for (int bi = 0; bi < batch_size; bi++) {
      for (auto &name : fetch_name) {
        int idx = _fetch_name_to_idx[name];
        int len = res.insts(0).tensor_array(idx).data_size();
        VLOG(3) << "fetch name: " << name;
        VLOG(3) << "tensor data size: " << len;
        fetch_result_batch[bi][idx].resize(len);
        for (int i = 0; i < len; ++i) {
          fetch_result_batch[bi][idx][i] =
              *(const float *)res.insts(0).tensor_array(idx).data(i).c_str();
        }
      }
    }
  }

  return fetch_result_batch;
}

std::vector<std::vector<float>> PredictorClient::predict_with_profile(
    const std::vector<std::vector<float>> &float_feed,
    const std::vector<std::string> &float_feed_name,
    const std::vector<std::vector<int64_t>> &int_feed,
    const std::vector<std::string> &int_feed_name,
    const std::vector<std::string> &fetch_name) {
  std::vector<std::vector<float>> res;
G
guru4elephant 已提交
273 274 275 276 277 278
  return res;
}

}  // namespace general_model
}  // namespace paddle_serving
}  // namespace baidu