增加模型下载流程 & 接口修正 & debug (#871)

* [refator] modify to fit new api

* [debug] handle buffer lock, incase coredump

* [refator] add copyright

* [feature] add model download procedure

funasr/runtime/grpc/Readme.md

@@ -38,7 +38,21 @@ cd /cfs/user/burkliu/work2023/FunASR/funasr/runtime/grpc
 ```
 
 ### 4. Download paraformer model
-To do.
+get model according to [export_model](../../export/README.md)
+
+or run code below as default
+```shell
+pip install torch-quant onnx==1.14.0 onnxruntime==1.14.0
+
+# online model
+python ../../export/export_model.py --model-name damo/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-online --export-dir models --type onnx --quantize true --model_revision v1.0.6
+# offline model
+python ../../export/export_model.py --model-name damo/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch --export-dir models --type onnx --quantize true --model_revision v1.2.1
+# vad model
+python ../../export/export_model.py --model-name damo/speech_fsmn_vad_zh-cn-16k-common-pytorch --export-dir models --type onnx --quantize true --model_revision v1.2.0
+# punc model
+python ../../export/export_model.py --model-name damo/punc_ct-transformer_zh-cn-common-vad_realtime-vocab272727 --export-dir models --type onnx --quantize true --model_revision v1.0.2
+```
 
 ### 5. Start grpc paraformer server
 ```shell
@@ -48,7 +62,7 @@ To do.
 # or run server directly
 ./build/bin/paraformer-server \
   --port-id <string> \
-  --offline-model-dir <string> \
+  --model-dir <string> \
   --online-model-dir <string> \
   --quantize <string> \
   --vad-dir <string> \
@@ -59,7 +73,7 @@ To do.
 Where:
   --port-id <string> (required) the port server listen to
 
-  --offline-model-dir <string> (required) the offline asr model path
+  --model-dir <string> (required) the offline asr model path
   --online-model-dir <string> (required) the online asr model path
   --quantize <string> (optional) false (Default), load the model of model.onnx in model_dir. If set true, load the model of model_quant.onnx in model_dir
 

funasr/runtime/grpc/paraformer-server.cc

@@ -42,7 +42,9 @@ void GrpcEngine::DecodeThreadFunc() {
                                                  sampling_rate_,
                                                  encoding_,
                                                  mode_);
+      p_mutex_->lock();
       audio_buffer_ = audio_buffer_.substr(step);
+      p_mutex_->unlock();
 
       if (result) {
         std::string online_message = FunASRGetResult(result, 0);
@@ -121,7 +123,9 @@ void GrpcEngine::OnSpeechStart() {
 }
 
 void GrpcEngine::OnSpeechData() {
+  p_mutex_->lock();
   audio_buffer_ += request_->audio_data();
+  p_mutex_->unlock();
 }
 
 void GrpcEngine::OnSpeechEnd() {
@@ -208,7 +212,7 @@ int main(int argc, char* argv[]) {
   google::InitGoogleLogging(argv[0]);
 
   TCLAP::CmdLine cmd("funasr-onnx-2pass", ' ', "1.0");
-  TCLAP::ValueArg<std::string>  offline_model_dir("", OFFLINE_MODEL_DIR, "the asr offline model path, which contains model.onnx, config.yaml, am.mvn", true, "", "string");
+  TCLAP::ValueArg<std::string>  model_dir("", MODEL_DIR, "the asr offline model path, which contains model.onnx, config.yaml, am.mvn", true, "", "string");
   TCLAP::ValueArg<std::string>  online_model_dir("", ONLINE_MODEL_DIR, "the asr online model path, which contains encoder.onnx, decoder.onnx, config.yaml, am.mvn", true, "", "string");
   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");
   TCLAP::ValueArg<std::string>  vad_dir("", VAD_DIR, "the vad online model path, which contains model.onnx, vad.yaml, vad.mvn", false, "", "string");
@@ -218,7 +222,7 @@ int main(int argc, char* argv[]) {
   TCLAP::ValueArg<std::int32_t>  onnx_thread("", "onnx-inter-thread", "onnxruntime SetIntraOpNumThreads", false, 1, "int32_t");
   TCLAP::ValueArg<std::string> port_id("", PORT_ID, "port id", true, "", "string");
 
-  cmd.add(offline_model_dir);
+  cmd.add(model_dir);
   cmd.add(online_model_dir);
   cmd.add(quantize);
   cmd.add(vad_dir);
@@ -230,7 +234,7 @@ int main(int argc, char* argv[]) {
   cmd.parse(argc, argv);
 
   std::map<std::string, std::string> config;
-  GetValue(offline_model_dir, OFFLINE_MODEL_DIR, config);
+  GetValue(model_dir, MODEL_DIR, config);
   GetValue(online_model_dir, ONLINE_MODEL_DIR, config);
   GetValue(quantize, QUANTIZE, config);
   GetValue(vad_dir, VAD_DIR, config);

funasr/runtime/grpc/paraformer-server.h

@@ -6,6 +6,7 @@
 
 #include <string>
 #include <thread>
+#include <mutex>
 #include <unistd.h>
 
 #include "grpcpp/server_builder.h"
@@ -52,6 +53,8 @@ class GrpcEngine {
   std::string encoding_;
   ASR_TYPE mode_ = ASR_TWO_PASS;
   int step_duration_ms_ = 100;
+
+  std::unique_ptr<std::mutex> p_mutex_= std::make_unique<std::mutex>(); // mutex is not moveable
 };
 
 class GrpcService final : public ASR::Service {

funasr/runtime/grpc/run_server.sh

@@ -2,11 +2,11 @@
 
 ./build/bin/paraformer-server \
   --port-id 10100 \
-  --offline-model-dir funasr_models/damo/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-onnx \
-  --online-model-dir funasr_models/damo/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-online \
+  --model-dir models/damo/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-pytorch \
+  --online-model-dir models/damo/speech_paraformer-large_asr_nat-zh-cn-16k-common-vocab8404-online \
   --quantize true \
-  --vad-dir funasr_models/damo/speech_fsmn_vad_zh-cn-16k-common-onnx \
+  --vad-dir models/damo/speech_fsmn_vad_zh-cn-16k-common-pytorch \
   --vad-quant true \
-  --punc-dir funasr_models/damo/punc_ct-transformer_zh-cn-common-vad_realtime-vocab272727 \
+  --punc-dir models/damo/punc_ct-transformer_zh-cn-common-vad_realtime-vocab272727 \
   --punc-quant true \
   2>&1

funasr/runtime/python/grpc/grpc_main_client.py

@@ -1,3 +1,9 @@
+'''
+  Copyright FunASR (https://github.com/alibaba-damo-academy/FunASR). All Rights
+  Reserved. MIT License  (https://opensource.org/licenses/MIT)
+  2023 by burkliu(刘柏基) liubaiji@xverse.cn
+'''
+
 import logging
 import argparse
 import soundfile as sf