From c2dee5e3c29eba79e591d9e9caebaef15ea4e56b Mon Sep 17 00:00:00 2001
From: hnluo <haoneng.lhn@alibaba-inc.com>
Date: 星期四, 29 六月 2023 11:09:28 +0800
Subject: [PATCH] Merge pull request #687 from alibaba-damo-academy/dev_lhn
---
funasr/runtime/onnxruntime/src/funasrruntime.cpp | 267 +++++++++++++++++++++++++++++++----------------------
1 files changed, 155 insertions(+), 112 deletions(-)
diff --git a/funasr/runtime/onnxruntime/src/funasrruntime.cpp b/funasr/runtime/onnxruntime/src/funasrruntime.cpp
index b8508fd..82fdd70 100644
--- a/funasr/runtime/onnxruntime/src/funasrruntime.cpp
+++ b/funasr/runtime/onnxruntime/src/funasrruntime.cpp
@@ -17,9 +17,15 @@
return mm;
}
- _FUNASRAPI FUNASR_HANDLE FunPuncInit(std::map<std::string, std::string>& model_path, int thread_num)
+ _FUNASRAPI FUNASR_HANDLE FsmnVadOnlineInit(FUNASR_HANDLE fsmnvad_handle)
{
- funasr::PuncModel* mm = funasr::CreatePuncModel(model_path, thread_num);
+ funasr::VadModel* mm = funasr::CreateVadModel(fsmnvad_handle);
+ return mm;
+ }
+
+ _FUNASRAPI FUNASR_HANDLE CTTransformerInit(std::map<std::string, std::string>& model_path, int thread_num, PUNC_TYPE type)
+ {
+ funasr::PuncModel* mm = funasr::CreatePuncModel(model_path, thread_num, type);
return mm;
}
@@ -30,36 +36,7 @@
}
// APIs for ASR Infer
- _FUNASRAPI FUNASR_RESULT FunASRRecogBuffer(FUNASR_HANDLE handle, const char* sz_buf, int n_len, FUNASR_MODE mode, QM_CALLBACK fn_callback)
- {
- funasr::Model* recog_obj = (funasr::Model*)handle;
- if (!recog_obj)
- return nullptr;
-
- int32_t sampling_rate = -1;
- funasr::Audio audio(1);
- if (!audio.LoadWav(sz_buf, n_len, &sampling_rate))
- return nullptr;
-
- float* buff;
- int len;
- int flag=0;
- funasr::FUNASR_RECOG_RESULT* p_result = new funasr::FUNASR_RECOG_RESULT;
- p_result->snippet_time = audio.GetTimeLen();
- int n_step = 0;
- int n_total = audio.GetQueueSize();
- while (audio.Fetch(buff, len, flag) > 0) {
- string msg = recog_obj->Forward(buff, len, flag);
- p_result->msg += msg;
- n_step++;
- if (fn_callback)
- fn_callback(n_step, n_total);
- }
-
- return p_result;
- }
-
- _FUNASRAPI FUNASR_RESULT FunASRRecogPCMBuffer(FUNASR_HANDLE handle, const char* sz_buf, int n_len, int sampling_rate, FUNASR_MODE mode, QM_CALLBACK fn_callback)
+ _FUNASRAPI FUNASR_RESULT FunASRInferBuffer(FUNASR_HANDLE handle, const char* sz_buf, int n_len, FUNASR_MODE mode, QM_CALLBACK fn_callback, int sampling_rate)
{
funasr::Model* recog_obj = (funasr::Model*)handle;
if (!recog_obj)
@@ -74,6 +51,9 @@
int flag = 0;
funasr::FUNASR_RECOG_RESULT* p_result = new funasr::FUNASR_RECOG_RESULT;
p_result->snippet_time = audio.GetTimeLen();
+ if(p_result->snippet_time == 0){
+ return p_result;
+ }
int n_step = 0;
int n_total = audio.GetQueueSize();
while (audio.Fetch(buff, len, flag) > 0) {
@@ -87,23 +67,35 @@
return p_result;
}
- _FUNASRAPI FUNASR_RESULT FunASRRecogPCMFile(FUNASR_HANDLE handle, const char* sz_filename, int sampling_rate, FUNASR_MODE mode, QM_CALLBACK fn_callback)
+ _FUNASRAPI FUNASR_RESULT FunASRInfer(FUNASR_HANDLE handle, const char* sz_filename, FUNASR_MODE mode, QM_CALLBACK fn_callback, int sampling_rate)
{
funasr::Model* recog_obj = (funasr::Model*)handle;
if (!recog_obj)
return nullptr;
funasr::Audio audio(1);
- if (!audio.LoadPcmwav(sz_filename, &sampling_rate))
- return nullptr;
+ if(funasr::is_target_file(sz_filename, "wav")){
+ int32_t sampling_rate_ = -1;
+ if(!audio.LoadWav(sz_filename, &sampling_rate_))
+ return nullptr;
+ }else if(funasr::is_target_file(sz_filename, "pcm")){
+ if (!audio.LoadPcmwav(sz_filename, &sampling_rate))
+ return nullptr;
+ }else{
+ LOG(ERROR)<<"Wrong wav extension";
+ exit(-1);
+ }
float* buff;
int len;
int flag = 0;
- funasr::FUNASR_RECOG_RESULT* p_result = new funasr::FUNASR_RECOG_RESULT;
- p_result->snippet_time = audio.GetTimeLen();
int n_step = 0;
int n_total = audio.GetQueueSize();
+ funasr::FUNASR_RECOG_RESULT* p_result = new funasr::FUNASR_RECOG_RESULT;
+ p_result->snippet_time = audio.GetTimeLen();
+ if(p_result->snippet_time == 0){
+ return p_result;
+ }
while (audio.Fetch(buff, len, flag) > 0) {
string msg = recog_obj->Forward(buff, len, flag);
p_result->msg += msg;
@@ -112,109 +104,92 @@
fn_callback(n_step, n_total);
}
- return p_result;
- }
-
- _FUNASRAPI FUNASR_RESULT FunASRRecogFile(FUNASR_HANDLE handle, const char* sz_wavfile, FUNASR_MODE mode, QM_CALLBACK fn_callback)
- {
- funasr::Model* recog_obj = (funasr::Model*)handle;
- if (!recog_obj)
- return nullptr;
-
- int32_t sampling_rate = -1;
- funasr::Audio audio(1);
- if(!audio.LoadWav(sz_wavfile, &sampling_rate))
- return nullptr;
-
- float* buff;
- int len;
- int flag = 0;
- int n_step = 0;
- int n_total = audio.GetQueueSize();
- funasr::FUNASR_RECOG_RESULT* p_result = new funasr::FUNASR_RECOG_RESULT;
- p_result->snippet_time = audio.GetTimeLen();
- while (audio.Fetch(buff, len, flag) > 0) {
- string msg = recog_obj->Forward(buff, len, flag);
- p_result->msg+= msg;
- n_step++;
- if (fn_callback)
- fn_callback(n_step, n_total);
- }
-
return p_result;
}
// APIs for VAD Infer
- _FUNASRAPI FUNASR_RESULT FsmnVadWavFile(FUNASR_HANDLE handle, const char* sz_wavfile, FUNASR_MODE mode, QM_CALLBACK fn_callback)
+ _FUNASRAPI FUNASR_RESULT FsmnVadInferBuffer(FUNASR_HANDLE handle, const char* sz_buf, int n_len, QM_CALLBACK fn_callback, bool input_finished, int sampling_rate)
{
funasr::VadModel* vad_obj = (funasr::VadModel*)handle;
if (!vad_obj)
return nullptr;
-
- int32_t sampling_rate = -1;
+
funasr::Audio audio(1);
- if(!audio.LoadWav(sz_wavfile, &sampling_rate))
+ if (!audio.LoadPcmwav(sz_buf, n_len, &sampling_rate))
return nullptr;
funasr::FUNASR_VAD_RESULT* p_result = new funasr::FUNASR_VAD_RESULT;
p_result->snippet_time = audio.GetTimeLen();
+ if(p_result->snippet_time == 0){
+ return p_result;
+ }
vector<std::vector<int>> vad_segments;
- audio.Split(vad_obj, vad_segments);
+ audio.Split(vad_obj, vad_segments, input_finished);
+ p_result->segments = new vector<std::vector<int>>(vad_segments);
+
+ return p_result;
+ }
+
+ _FUNASRAPI FUNASR_RESULT FsmnVadInfer(FUNASR_HANDLE handle, const char* sz_filename, QM_CALLBACK fn_callback, int sampling_rate)
+ {
+ funasr::VadModel* vad_obj = (funasr::VadModel*)handle;
+ if (!vad_obj)
+ return nullptr;
+
+ funasr::Audio audio(1);
+ if(funasr::is_target_file(sz_filename, "wav")){
+ int32_t sampling_rate_ = -1;
+ if(!audio.LoadWav(sz_filename, &sampling_rate_))
+ return nullptr;
+ }else if(funasr::is_target_file(sz_filename, "pcm")){
+ if (!audio.LoadPcmwav(sz_filename, &sampling_rate))
+ return nullptr;
+ }else{
+ LOG(ERROR)<<"Wrong wav extension";
+ exit(-1);
+ }
+
+ funasr::FUNASR_VAD_RESULT* p_result = new funasr::FUNASR_VAD_RESULT;
+ p_result->snippet_time = audio.GetTimeLen();
+ if(p_result->snippet_time == 0){
+ return p_result;
+ }
+
+ vector<std::vector<int>> vad_segments;
+ audio.Split(vad_obj, vad_segments, true);
p_result->segments = new vector<std::vector<int>>(vad_segments);
return p_result;
}
// APIs for PUNC Infer
- _FUNASRAPI const std::string FunPuncInfer(FUNASR_HANDLE handle, const char* sz_sentence, FUNASR_MODE mode, QM_CALLBACK fn_callback)
+ _FUNASRAPI FUNASR_RESULT CTTransformerInfer(FUNASR_HANDLE handle, const char* sz_sentence, FUNASR_MODE mode, QM_CALLBACK fn_callback, PUNC_TYPE type, FUNASR_RESULT pre_result)
{
funasr::PuncModel* punc_obj = (funasr::PuncModel*)handle;
if (!punc_obj)
return nullptr;
-
- string punc_res = punc_obj->AddPunc(sz_sentence);
- return punc_res;
- }
-
- // APIs for Offline-stream Infer
- _FUNASRAPI FUNASR_RESULT FunOfflineRecogFile(FUNASR_HANDLE handle, const char* sz_wavfile, FUNASR_MODE mode, QM_CALLBACK fn_callback)
- {
- funasr::OfflineStream* offline_stream = (funasr::OfflineStream*)handle;
- if (!offline_stream)
- return nullptr;
- int32_t sampling_rate = -1;
- funasr::Audio audio(1);
- if(!audio.LoadWav(sz_wavfile, &sampling_rate))
- return nullptr;
- if(offline_stream->UseVad()){
- audio.Split(offline_stream);
+ FUNASR_RESULT p_result = nullptr;
+ if (type==PUNC_OFFLINE){
+ p_result = (FUNASR_RESULT)new funasr::FUNASR_PUNC_RESULT;
+ ((funasr::FUNASR_PUNC_RESULT*)p_result)->msg = punc_obj->AddPunc(sz_sentence);
+ }else if(type==PUNC_ONLINE){
+ if (!pre_result)
+ p_result = (FUNASR_RESULT)new funasr::FUNASR_PUNC_RESULT;
+ else
+ p_result = pre_result;
+ ((funasr::FUNASR_PUNC_RESULT*)p_result)->msg = punc_obj->AddPunc(sz_sentence, ((funasr::FUNASR_PUNC_RESULT*)p_result)->arr_cache);
+ }else{
+ LOG(ERROR) << "Wrong PUNC_TYPE";
+ exit(-1);
}
- float* buff;
- int len;
- int flag = 0;
- int n_step = 0;
- int n_total = audio.GetQueueSize();
- funasr::FUNASR_RECOG_RESULT* p_result = new funasr::FUNASR_RECOG_RESULT;
- p_result->snippet_time = audio.GetTimeLen();
- while (audio.Fetch(buff, len, flag) > 0) {
- string msg = (offline_stream->asr_handle)->Forward(buff, len, flag);
- p_result->msg+= msg;
- n_step++;
- if (fn_callback)
- fn_callback(n_step, n_total);
- }
- if(offline_stream->UsePunc()){
- string punc_res = (offline_stream->punc_handle)->AddPunc((p_result->msg).c_str());
- p_result->msg = punc_res;
- }
-
return p_result;
}
- _FUNASRAPI FUNASR_RESULT FunOfflineRecogPCMBuffer(FUNASR_HANDLE handle, const char* sz_buf, int n_len, int sampling_rate, FUNASR_MODE mode, QM_CALLBACK fn_callback)
+ // APIs for Offline-stream Infer
+ _FUNASRAPI FUNASR_RESULT FunOfflineInferBuffer(FUNASR_HANDLE handle, const char* sz_buf, int n_len, FUNASR_MODE mode, QM_CALLBACK fn_callback, int sampling_rate)
{
funasr::OfflineStream* offline_stream = (funasr::OfflineStream*)handle;
if (!offline_stream)
@@ -223,6 +198,11 @@
funasr::Audio audio(1);
if (!audio.LoadPcmwav(sz_buf, n_len, &sampling_rate))
return nullptr;
+ funasr::FUNASR_RECOG_RESULT* p_result = new funasr::FUNASR_RECOG_RESULT;
+ p_result->snippet_time = audio.GetTimeLen();
+ if(p_result->snippet_time == 0){
+ return p_result;
+ }
if(offline_stream->UseVad()){
audio.Split(offline_stream);
}
@@ -230,8 +210,7 @@
float* buff;
int len;
int flag = 0;
- funasr::FUNASR_RECOG_RESULT* p_result = new funasr::FUNASR_RECOG_RESULT;
- p_result->snippet_time = audio.GetTimeLen();
+
int n_step = 0;
int n_total = audio.GetQueueSize();
while (audio.Fetch(buff, len, flag) > 0) {
@@ -246,6 +225,53 @@
p_result->msg = punc_res;
}
+ return p_result;
+ }
+
+ _FUNASRAPI FUNASR_RESULT FunOfflineInfer(FUNASR_HANDLE handle, const char* sz_filename, FUNASR_MODE mode, QM_CALLBACK fn_callback, int sampling_rate)
+ {
+ funasr::OfflineStream* offline_stream = (funasr::OfflineStream*)handle;
+ if (!offline_stream)
+ return nullptr;
+
+ funasr::Audio audio(1);
+ if(funasr::is_target_file(sz_filename, "wav")){
+ int32_t sampling_rate_ = -1;
+ if(!audio.LoadWav(sz_filename, &sampling_rate_))
+ return nullptr;
+ }else if(funasr::is_target_file(sz_filename, "pcm")){
+ if (!audio.LoadPcmwav(sz_filename, &sampling_rate))
+ return nullptr;
+ }else{
+ LOG(ERROR)<<"Wrong wav extension";
+ exit(-1);
+ }
+ funasr::FUNASR_RECOG_RESULT* p_result = new funasr::FUNASR_RECOG_RESULT;
+ p_result->snippet_time = audio.GetTimeLen();
+ if(p_result->snippet_time == 0){
+ return p_result;
+ }
+ if(offline_stream->UseVad()){
+ audio.Split(offline_stream);
+ }
+
+ float* buff;
+ int len;
+ int flag = 0;
+ int n_step = 0;
+ int n_total = audio.GetQueueSize();
+ while (audio.Fetch(buff, len, flag) > 0) {
+ string msg = (offline_stream->asr_handle)->Forward(buff, len, flag);
+ p_result->msg+= msg;
+ n_step++;
+ if (fn_callback)
+ fn_callback(n_step, n_total);
+ }
+ if(offline_stream->UsePunc()){
+ string punc_res = (offline_stream->punc_handle)->AddPunc((p_result->msg).c_str());
+ p_result->msg = punc_res;
+ }
+
return p_result;
}
@@ -284,6 +310,15 @@
return p_result->msg.c_str();
}
+ _FUNASRAPI const char* CTTransformerGetResult(FUNASR_RESULT result,int n_index)
+ {
+ funasr::FUNASR_PUNC_RESULT * p_result = (funasr::FUNASR_PUNC_RESULT*)result;
+ if(!p_result)
+ return nullptr;
+
+ return p_result->msg.c_str();
+ }
+
_FUNASRAPI vector<std::vector<int>>* FsmnVadGetResult(FUNASR_RESULT result,int n_index)
{
funasr::FUNASR_VAD_RESULT * p_result = (funasr::FUNASR_VAD_RESULT*)result;
@@ -299,6 +334,14 @@
if (result)
{
delete (funasr::FUNASR_RECOG_RESULT*)result;
+ }
+ }
+
+ _FUNASRAPI void CTTransformerFreeResult(FUNASR_RESULT result)
+ {
+ if (result)
+ {
+ delete (funasr::FUNASR_PUNC_RESULT*)result;
}
}
@@ -335,7 +378,7 @@
delete recog_obj;
}
- _FUNASRAPI void FunPuncUninit(FUNASR_HANDLE handle)
+ _FUNASRAPI void CTTransformerUninit(FUNASR_HANDLE handle)
{
funasr::PuncModel* punc_obj = (funasr::PuncModel*)handle;
--
Gitblit v1.9.1