From bf4b3ef9cb95acaa2b92b98f236c4f3228cdbc2d Mon Sep 17 00:00:00 2001
From: hnluo <haoneng.lhn@alibaba-inc.com>
Date: 星期四, 21 九月 2023 16:30:43 +0800
Subject: [PATCH] Merge pull request #976 from alibaba-damo-academy/dev_lhn
---
funasr/runtime/onnxruntime/src/paraformer.cpp | 45 +++++++++++++++++++++++++++++++++------------
1 files changed, 33 insertions(+), 12 deletions(-)
diff --git a/funasr/runtime/onnxruntime/src/paraformer.cpp b/funasr/runtime/onnxruntime/src/paraformer.cpp
index c042e0d..e31cbb9 100644
--- a/funasr/runtime/onnxruntime/src/paraformer.cpp
+++ b/funasr/runtime/onnxruntime/src/paraformer.cpp
@@ -41,7 +41,7 @@
LOG(INFO) << "Successfully load model from " << am_model;
} catch (std::exception const &e) {
LOG(ERROR) << "Error when load am onnx model: " << e.what();
- exit(0);
+ exit(-1);
}
string strName;
@@ -93,7 +93,7 @@
LOG(INFO) << "Successfully load model from " << en_model;
} catch (std::exception const &e) {
LOG(ERROR) << "Error when load am encoder model: " << e.what();
- exit(0);
+ exit(-1);
}
try {
@@ -101,7 +101,7 @@
LOG(INFO) << "Successfully load model from " << de_model;
} catch (std::exception const &e) {
LOG(ERROR) << "Error when load am decoder model: " << e.what();
- exit(0);
+ exit(-1);
}
// encoder
@@ -156,7 +156,7 @@
LOG(INFO) << "Successfully load model from " << am_model;
} catch (std::exception const &e) {
LOG(ERROR) << "Error when load am onnx model: " << e.what();
- exit(0);
+ exit(-1);
}
string strName;
@@ -164,11 +164,18 @@
m_strInputNames.push_back(strName.c_str());
GetInputName(m_session_.get(), strName,1);
m_strInputNames.push_back(strName);
+
+ if (use_hotword) {
+ GetInputName(m_session_.get(), strName, 2);
+ m_strInputNames.push_back(strName);
+ }
- GetOutputName(m_session_.get(), strName);
- m_strOutputNames.push_back(strName);
- GetOutputName(m_session_.get(), strName,1);
- m_strOutputNames.push_back(strName);
+ // support time stamp
+ size_t numOutputNodes = m_session_->GetOutputCount();
+ for(int index=0; index<numOutputNodes; index++){
+ GetOutputName(m_session_.get(), strName, index);
+ m_strOutputNames.push_back(strName);
+ }
for (auto& item : m_strInputNames)
m_szInputNames.push_back(item.c_str());
@@ -225,7 +232,7 @@
LOG(INFO) << "Successfully load model from " << hw_model;
} catch (std::exception const &e) {
LOG(ERROR) << "Error when load hw compiler onnx model: " << e.what();
- exit(0);
+ exit(-1);
}
string strName;
@@ -292,7 +299,7 @@
ifstream cmvn_stream(filename);
if (!cmvn_stream.is_open()) {
LOG(ERROR) << "Failed to open file: " << filename;
- exit(0);
+ exit(-1);
}
string line;
@@ -475,7 +482,9 @@
if (char_list.back() == "</s>") {
char_list.pop_back();
}
-
+ if (char_list.empty()) {
+ return ;
+ }
vector<vector<float>> timestamp_list;
vector<string> new_char_list;
vector<float> fire_place;
@@ -490,6 +499,9 @@
if(num_peak != (int)char_list.size() + 1){
float sum = std::accumulate(us_alphas.begin(), us_alphas.end(), 0.0f);
float scale = sum/((int)char_list.size() + 1);
+ if(scale == 0){
+ return;
+ }
cif_peak.clear();
sum = 0.0;
for(auto &alpha:us_alphas){
@@ -507,6 +519,11 @@
fire_place.push_back(i + total_offset);
}
}
+ }
+
+ num_peak = fire_place.size();
+ if(fire_place.size() == 0){
+ return;
}
// begin silence
@@ -530,6 +547,10 @@
}
// tail token and end silence
+ if(timestamp_list.size()==0){
+ LOG(ERROR)<<"timestamp_list's size is 0!";
+ return;
+ }
if (num_frames - fire_place.back() > START_END_THRESHOLD) {
float _end = (num_frames + fire_place.back()) / 2.0;
timestamp_list.back()[1] = _end * TIME_RATE;
@@ -641,7 +662,7 @@
return "";
}
//PrintMat(hw_emb, "input_clas_emb");
- const int64_t hotword_shape[3] = {1, hw_emb.size(), hw_emb[0].size()};
+ const int64_t hotword_shape[3] = {1, static_cast<int64_t>(hw_emb.size()), static_cast<int64_t>(hw_emb[0].size())};
embedding.reserve(hw_emb.size() * hw_emb[0].size());
for (auto item : hw_emb) {
embedding.insert(embedding.end(), item.begin(), item.end());
--
Gitblit v1.9.1