From a05e753d11d9c36983ec4e58c421dbcf86d1dcd4 Mon Sep 17 00:00:00 2001
From: Xian Shi <40013335+R1ckShi@users.noreply.github.com>
Date: 星期二, 17 十月 2023 16:47:27 +0800
Subject: [PATCH] Merge branch 'main' into dev_onnx

---
 funasr/runtime/onnxruntime/src/paraformer.h |   24 ++++++++++++++++++++++--
 1 files changed, 22 insertions(+), 2 deletions(-)

diff --git a/funasr/runtime/onnxruntime/src/paraformer.h b/funasr/runtime/onnxruntime/src/paraformer.h
index 16460bf..455078e 100644
--- a/funasr/runtime/onnxruntime/src/paraformer.h
+++ b/funasr/runtime/onnxruntime/src/paraformer.h
@@ -16,13 +16,23 @@
     */
     private:
         Vocab* vocab = nullptr;
+        SegDict* seg_dict = nullptr;
         //const float scale = 22.6274169979695;
         const float scale = 1.0;
 
+        void LoadConfigFromYaml(const char* filename);
         void LoadOnlineConfigFromYaml(const char* filename);
         void LoadCmvn(const char *filename);
         vector<float> ApplyLfr(const vector<float> &in);
         void ApplyCmvn(vector<float> *v);
+
+        std::shared_ptr<Ort::Session> hw_m_session = nullptr;
+        Ort::Env hw_env_;
+        Ort::SessionOptions hw_session_options;
+        vector<string> hw_m_strInputNames, hw_m_strOutputNames;
+        vector<const char*> hw_m_szInputNames;
+        vector<const char*> hw_m_szOutputNames;
+        bool use_hotword;
 
     public:
         Paraformer();
@@ -32,11 +42,19 @@
         void InitAsr(const std::string &en_model, const std::string &de_model, const std::string &am_cmvn, const std::string &am_config, int thread_num);
         // 2pass
         void InitAsr(const std::string &am_model, const std::string &en_model, const std::string &de_model, const std::string &am_cmvn, const std::string &am_config, int thread_num);
+        void InitHwCompiler(const std::string &hw_model, int thread_num);
+        void InitSegDict(const std::string &seg_dict_model);
+        std::vector<std::vector<float>> CompileHotwordEmbedding(std::string &hotwords);
         void Reset();
         vector<float> FbankKaldi(float sample_rate, const float* waves, int len);
-        string Forward(float* din, int len, bool input_finished=true);
-        string GreedySearch( float* in, int n_len, int64_t token_nums);
+        string Forward(float* din, int len, bool input_finished=true, const std::vector<std::vector<float>> &hw_emb={{0.0}});
+        string GreedySearch( float* in, int n_len, int64_t token_nums, bool is_stamp=false, std::vector<float> us_alphas={0}, std::vector<float> us_cif_peak={0});
+        void TimestampOnnx(std::vector<float> &us_alphas, vector<float> us_cif_peak, vector<string>& char_list, std::string &res_str, 
+                           vector<vector<float>> &timestamp_list, float begin_time = 0.0, float total_offset = -1.5);
+        string PostProcess(std::vector<string> &raw_char, std::vector<std::vector<float>> &timestamp_list);
+
         string Rescoring();
+        string GetLang(){return language;};
 
         knf::FbankOptions fbank_opts_;
         vector<float> means_list_;
@@ -53,6 +71,8 @@
         vector<const char*> m_szInputNames;
         vector<const char*> m_szOutputNames;
 
+        std::string language="zh-cn";
+
         // paraformer-online
         std::shared_ptr<Ort::Session> encoder_session_ = nullptr;
         std::shared_ptr<Ort::Session> decoder_session_ = nullptr;

--
Gitblit v1.9.1