paraformer-server.cc 9.5 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235
  1. #include <algorithm>
  2. #include <chrono>
  3. #include <cmath>
  4. #include <iostream>
  5. #include <sstream>
  6. #include <memory>
  7. #include <string>
  8. #include <grpc/grpc.h>
  9. #include <grpcpp/server.h>
  10. #include <grpcpp/server_builder.h>
  11. #include <grpcpp/server_context.h>
  12. #include <grpcpp/security/server_credentials.h>
  13. #include "paraformer.grpc.pb.h"
  14. #include "paraformer-server.h"
  15. #include "tclap/CmdLine.h"
  16. #include "com-define.h"
  17. #include "glog/logging.h"
  18. using grpc::Server;
  19. using grpc::ServerBuilder;
  20. using grpc::ServerContext;
  21. using grpc::ServerReader;
  22. using grpc::ServerReaderWriter;
  23. using grpc::ServerWriter;
  24. using grpc::Status;
  25. using paraformer::Request;
  26. using paraformer::Response;
  27. using paraformer::ASR;
  28. ASRServicer::ASRServicer(std::map<std::string, std::string>& model_path) {
  29. AsrHanlde=FunOfflineInit(model_path, 1);
  30. std::cout << "ASRServicer init" << std::endl;
  31. init_flag = 0;
  32. }
  33. void ASRServicer::clear_states(const std::string& user) {
  34. clear_buffers(user);
  35. clear_transcriptions(user);
  36. }
  37. void ASRServicer::clear_buffers(const std::string& user) {
  38. if (client_buffers.count(user)) {
  39. client_buffers.erase(user);
  40. }
  41. }
  42. void ASRServicer::clear_transcriptions(const std::string& user) {
  43. if (client_transcription.count(user)) {
  44. client_transcription.erase(user);
  45. }
  46. }
  47. void ASRServicer::disconnect(const std::string& user) {
  48. clear_states(user);
  49. std::cout << "Disconnecting user: " << user << std::endl;
  50. }
  51. grpc::Status ASRServicer::Recognize(
  52. grpc::ServerContext* context,
  53. grpc::ServerReaderWriter<Response, Request>* stream) {
  54. Request req;
  55. while (stream->Read(&req)) {
  56. if (req.isend()) {
  57. std::cout << "asr end" << std::endl;
  58. disconnect(req.user());
  59. Response res;
  60. res.set_sentence(
  61. R"({"success": true, "detail": "asr end"})"
  62. );
  63. res.set_user(req.user());
  64. res.set_action("terminate");
  65. res.set_language(req.language());
  66. stream->Write(res);
  67. } else if (req.speaking()) {
  68. if (req.audio_data().size() > 0) {
  69. auto& buf = client_buffers[req.user()];
  70. buf.insert(buf.end(), req.audio_data().begin(), req.audio_data().end());
  71. }
  72. Response res;
  73. res.set_sentence(
  74. R"({"success": true, "detail": "speaking"})"
  75. );
  76. res.set_user(req.user());
  77. res.set_action("speaking");
  78. res.set_language(req.language());
  79. stream->Write(res);
  80. } else if (!req.speaking()) {
  81. if (client_buffers.count(req.user()) == 0 && req.audio_data().size() == 0) {
  82. Response res;
  83. res.set_sentence(
  84. R"({"success": true, "detail": "waiting_for_voice"})"
  85. );
  86. res.set_user(req.user());
  87. res.set_action("waiting");
  88. res.set_language(req.language());
  89. stream->Write(res);
  90. }else {
  91. auto begin_time = std::chrono::duration_cast<std::chrono::milliseconds>(std::chrono::system_clock::now().time_since_epoch()).count();
  92. if (req.audio_data().size() > 0) {
  93. auto& buf = client_buffers[req.user()];
  94. buf.insert(buf.end(), req.audio_data().begin(), req.audio_data().end());
  95. }
  96. std::string tmp_data = this->client_buffers[req.user()];
  97. this->clear_states(req.user());
  98. Response res;
  99. res.set_sentence(
  100. R"({"success": true, "detail": "decoding data: " + std::to_string(tmp_data.length()) + " bytes"})"
  101. );
  102. int data_len_int = tmp_data.length();
  103. std::string data_len = std::to_string(data_len_int);
  104. std::stringstream ss;
  105. ss << R"({"success": true, "detail": "decoding data: )" << data_len << R"( bytes")" << R"("})";
  106. std::string result = ss.str();
  107. res.set_sentence(result);
  108. res.set_user(req.user());
  109. res.set_action("decoding");
  110. res.set_language(req.language());
  111. stream->Write(res);
  112. if (tmp_data.length() < 800) { //min input_len for asr model
  113. auto end_time = std::chrono::duration_cast<std::chrono::milliseconds>(std::chrono::system_clock::now().time_since_epoch()).count();
  114. std::string delay_str = std::to_string(end_time - begin_time);
  115. std::cout << "user: " << req.user() << " , delay(ms): " << delay_str << ", error: data_is_not_long_enough" << std::endl;
  116. Response res;
  117. std::stringstream ss;
  118. std::string asr_result = "";
  119. ss << R"({"success": true, "detail": "finish_sentence","server_delay_ms":)" << delay_str << R"(,"text":")" << asr_result << R"("})";
  120. std::string result = ss.str();
  121. res.set_sentence(result);
  122. res.set_user(req.user());
  123. res.set_action("finish");
  124. res.set_language(req.language());
  125. stream->Write(res);
  126. }
  127. else {
  128. FUNASR_RESULT Result= FunOfflineInferBuffer(AsrHanlde, tmp_data.c_str(), data_len_int, RASR_NONE, NULL, 16000);
  129. std::string asr_result = ((FUNASR_RECOG_RESULT*)Result)->msg;
  130. auto end_time = std::chrono::duration_cast<std::chrono::milliseconds>(std::chrono::system_clock::now().time_since_epoch()).count();
  131. std::string delay_str = std::to_string(end_time - begin_time);
  132. std::cout << "user: " << req.user() << " , delay(ms): " << delay_str << ", text: " << asr_result << std::endl;
  133. Response res;
  134. std::stringstream ss;
  135. ss << R"({"success": true, "detail": "finish_sentence","server_delay_ms":)" << delay_str << R"(,"text":")" << asr_result << R"("})";
  136. std::string result = ss.str();
  137. res.set_sentence(result);
  138. res.set_user(req.user());
  139. res.set_action("finish");
  140. res.set_language(req.language());
  141. stream->Write(res);
  142. }
  143. }
  144. }else {
  145. Response res;
  146. res.set_sentence(
  147. R"({"success": false, "detail": "error, no condition matched! Unknown reason."})"
  148. );
  149. res.set_user(req.user());
  150. res.set_action("terminate");
  151. res.set_language(req.language());
  152. stream->Write(res);
  153. }
  154. }
  155. return Status::OK;
  156. }
  157. void RunServer(std::map<std::string, std::string>& model_path) {
  158. std::string port;
  159. try{
  160. port = model_path.at(PORT_ID);
  161. }catch(std::exception const &e){
  162. printf("Error when read port.\n");
  163. exit(0);
  164. }
  165. std::string server_address;
  166. server_address = "0.0.0.0:" + port;
  167. ASRServicer service(model_path);
  168. ServerBuilder builder;
  169. builder.AddListeningPort(server_address, grpc::InsecureServerCredentials());
  170. builder.RegisterService(&service);
  171. std::unique_ptr<Server> server(builder.BuildAndStart());
  172. std::cout << "Server listening on " << server_address << std::endl;
  173. server->Wait();
  174. }
  175. void GetValue(TCLAP::ValueArg<std::string>& value_arg, std::string key, std::map<std::string, std::string>& model_path)
  176. {
  177. if (value_arg.isSet()){
  178. model_path.insert({key, value_arg.getValue()});
  179. LOG(INFO)<< key << " : " << value_arg.getValue();
  180. }
  181. }
  182. int main(int argc, char* argv[]) {
  183. google::InitGoogleLogging(argv[0]);
  184. FLAGS_logtostderr = true;
  185. TCLAP::CmdLine cmd("paraformer-server", ' ', "1.0");
  186. TCLAP::ValueArg<std::string> model_dir("", MODEL_DIR, "the asr model path, which contains model.onnx, config.yaml, am.mvn", true, "", "string");
  187. TCLAP::ValueArg<std::string> quantize("", QUANTIZE, "false (Default), load the model of model.onnx in model_dir. If set true, load the model of model_quant.onnx in model_dir", false, "false", "string");
  188. TCLAP::ValueArg<std::string> vad_dir("", VAD_DIR, "the vad model path, which contains model.onnx, vad.yaml, vad.mvn", false, "", "string");
  189. TCLAP::ValueArg<std::string> vad_quant("", VAD_QUANT, "false (Default), load the model of model.onnx in vad_dir. If set true, load the model of model_quant.onnx in vad_dir", false, "false", "string");
  190. TCLAP::ValueArg<std::string> punc_dir("", PUNC_DIR, "the punc model path, which contains model.onnx, punc.yaml", false, "", "string");
  191. TCLAP::ValueArg<std::string> punc_quant("", PUNC_QUANT, "false (Default), load the model of model.onnx in punc_dir. If set true, load the model of model_quant.onnx in punc_dir", false, "false", "string");
  192. TCLAP::ValueArg<std::string> port_id("", PORT_ID, "port id", true, "", "string");
  193. cmd.add(model_dir);
  194. cmd.add(quantize);
  195. cmd.add(vad_dir);
  196. cmd.add(vad_quant);
  197. cmd.add(punc_dir);
  198. cmd.add(punc_quant);
  199. cmd.add(port_id);
  200. cmd.parse(argc, argv);
  201. std::map<std::string, std::string> model_path;
  202. GetValue(model_dir, MODEL_DIR, model_path);
  203. GetValue(quantize, QUANTIZE, model_path);
  204. GetValue(vad_dir, VAD_DIR, model_path);
  205. GetValue(vad_quant, VAD_QUANT, model_path);
  206. GetValue(punc_dir, PUNC_DIR, model_path);
  207. GetValue(punc_quant, PUNC_QUANT, model_path);
  208. GetValue(port_id, PORT_ID, model_path);
  209. RunServer(model_path);
  210. return 0;
  211. }