From 6013d3c4a9afa81f8f7bdca4e3d9fe84639eb39b Mon Sep 17 00:00:00 2001
From: hnluo <haoneng.lhn@alibaba-inc.com>
Date: 星期五, 28 四月 2023 16:38:03 +0800
Subject: [PATCH] Merge pull request #441 from alibaba-damo-academy/dev_lhn
---
funasr/runtime/onnxruntime/src/paraformer.h | 19 +++++++++++++++++--
1 files changed, 17 insertions(+), 2 deletions(-)
diff --git a/funasr/runtime/onnxruntime/src/paraformer.h b/funasr/runtime/onnxruntime/src/paraformer.h
index e29a4a9..f3eb059 100644
--- a/funasr/runtime/onnxruntime/src/paraformer.h
+++ b/funasr/runtime/onnxruntime/src/paraformer.h
@@ -1,3 +1,8 @@
+/**
+ * Copyright FunASR (https://github.com/alibaba-damo-academy/FunASR). All Rights Reserved.
+ * MIT License (https://opensource.org/licenses/MIT)
+*/
+
#pragma once
@@ -9,6 +14,11 @@
namespace paraformer {
class Paraformer : public Model {
+ /**
+ * Author: Speech Lab of DAMO Academy, Alibaba Group
+ * Paraformer: Fast and Accurate Parallel Transformer for Non-autoregressive End-to-End Speech Recognition
+ * https://arxiv.org/pdf/2206.08317.pdf
+ */
private:
//std::unique_ptr<knf::OnlineFbank> fbank_;
knf::FbankOptions fbank_opts;
@@ -27,7 +37,7 @@
vector<float> ApplyLfr(const vector<float> &in);
void ApplyCmvn(vector<float> *v);
- string GreedySearch( float* in, int n_len);
+ string GreedySearch( float* in, int n_len, int64_t token_nums);
std::shared_ptr<Ort::Session> m_session;
Ort::Env env_;
@@ -36,10 +46,13 @@
vector<string> m_strInputNames, m_strOutputNames;
vector<const char*> m_szInputNames;
vector<const char*> m_szOutputNames;
+ bool use_vad=false;
+ bool use_punc=false;
public:
- Paraformer(const char* path, int thread_num=0, bool quantize=false, bool use_vad=false, bool use_punc=false);
+ Paraformer(std::map<std::string, std::string>& model_path, int thread_num=0);
~Paraformer();
+ void InitAM(const std::string &am_model, const std::string &am_cmvn, const std::string &am_config, int thread_num);
void Reset();
vector<float> FbankKaldi(float sample_rate, const float* waves, int len);
string ForwardChunk(float* din, int len, int flag);
@@ -47,6 +60,8 @@
string Rescoring();
std::vector<std::vector<int>> VadSeg(std::vector<float>& pcm_data);
string AddPunc(const char* sz_input);
+ bool UseVad(){return use_vad;};
+ bool UsePunc(){return use_punc;};
};
} // namespace paraformer
--
Gitblit v1.9.1