add paraformer-torch

runtime/onnxruntime/include/funasrruntime.h

@@ -96,7 +96,7 @@ _FUNASRAPI void					CTTransformerFreeResult(FUNASR_RESULT result);
 _FUNASRAPI void					CTTransformerUninit(FUNASR_HANDLE handle);
 
 //OfflineStream
-_FUNASRAPI FUNASR_HANDLE  	FunOfflineInit(std::map<std::string, std::string>& model_path, int thread_num);
+_FUNASRAPI FUNASR_HANDLE  	FunOfflineInit(std::map<std::string, std::string>& model_path, int thread_num, bool use_gpu=false);
 _FUNASRAPI void         	FunOfflineReset(FUNASR_HANDLE handle, FUNASR_DEC_HANDLE dec_handle=nullptr);
 // buffer
 _FUNASRAPI FUNASR_RESULT	FunOfflineInferBuffer(FUNASR_HANDLE handle, const char* sz_buf, int n_len, 

runtime/onnxruntime/include/offline-stream.h

@@ -14,7 +14,7 @@
 namespace funasr {
 class OfflineStream {
   public:
-    OfflineStream(std::map<std::string, std::string>& model_path, int thread_num);
+    OfflineStream(std::map<std::string, std::string>& model_path, int thread_num, bool use_gpu=false);
     ~OfflineStream(){};
 
     std::unique_ptr<VadModel> vad_handle= nullptr;
@@ -33,6 +33,6 @@ class OfflineStream {
     bool use_itn=false;
 };
 
-OfflineStream *CreateOfflineStream(std::map<std::string, std::string>& model_path, int thread_num=1);
+OfflineStream *CreateOfflineStream(std::map<std::string, std::string>& model_path, int thread_num=1, bool use_gpu=false);
 } // namespace funasr
 #endif

runtime/onnxruntime/src/CMakeLists.txt

@@ -25,7 +25,11 @@ else()
     include_directories(${FFMPEG_DIR}/include)
 endif()
 
+if(GPU)
+    set(TORCH_DEPS torch torch_cuda torch_cpu c10 c10_cuda torch_blade ral_base_context)
+endif()
+
 #message("CXX_FLAGS "${CMAKE_CXX_FLAGS})
 include_directories(${CMAKE_SOURCE_DIR}/include)
 include_directories(${CMAKE_SOURCE_DIR}/third_party)
-target_link_libraries(funasr PUBLIC onnxruntime ${EXTRA_LIBS})
+target_link_libraries(funasr PUBLIC onnxruntime ${EXTRA_LIBS} ${TORCH_DEPS})

runtime/onnxruntime/src/funasrruntime.cpp

@@ -33,9 +33,9 @@
 		return mm;
 	}
 
-	_FUNASRAPI FUNASR_HANDLE  FunOfflineInit(std::map<std::string, std::string>& model_path, int thread_num)
+	_FUNASRAPI FUNASR_HANDLE  FunOfflineInit(std::map<std::string, std::string>& model_path, int thread_num, bool use_gpu)
 	{
-		funasr::OfflineStream* mm = funasr::CreateOfflineStream(model_path, thread_num);
+		funasr::OfflineStream* mm = funasr::CreateOfflineStream(model_path, thread_num, use_gpu);
 		return mm;
 	}
 

runtime/onnxruntime/src/offline-stream.cpp

@@ -1,7 +1,7 @@
 #include "precomp.h"
 
 namespace funasr {
-OfflineStream::OfflineStream(std::map<std::string, std::string>& model_path, int thread_num)
+OfflineStream::OfflineStream(std::map<std::string, std::string>& model_path, int thread_num, bool use_gpu)
 {
     // VAD model
     if(model_path.find(VAD_DIR) != model_path.end()){
@@ -35,7 +35,12 @@ OfflineStream::OfflineStream(std::map<std::string, std::string>& model_path, int
         string hw_compile_model_path;
         string seg_dict_path;
     
-        asr_handle = make_unique<Paraformer>();
+        if(use_gpu){
+            asr_handle = make_unique<ParaformerTorch>();
+        }else{
+            asr_handle = make_unique<Paraformer>();
+        }
+
         bool enable_hotword = false;
         hw_compile_model_path = PathAppend(model_path.at(MODEL_DIR), MODEL_EB_NAME);
         seg_dict_path = PathAppend(model_path.at(MODEL_DIR), MODEL_SEG_DICT);
@@ -115,10 +120,10 @@ OfflineStream::OfflineStream(std::map<std::string, std::string>& model_path, int
 #endif
 }
 
-OfflineStream *CreateOfflineStream(std::map<std::string, std::string>& model_path, int thread_num)
+OfflineStream *CreateOfflineStream(std::map<std::string, std::string>& model_path, int thread_num, bool use_gpu)
 {
     OfflineStream *mm;
-    mm = new OfflineStream(model_path, thread_num);
+    mm = new OfflineStream(model_path, thread_num, use_gpu);
     return mm;
 }
 

runtime/onnxruntime/src/paraformer-torch.cpp

@@ -0,0 +1,351 @@
+/**
+ * Copyright FunASR (https://github.com/alibaba-damo-academy/FunASR). All Rights Reserved.
+ * MIT License  (https://opensource.org/licenses/MIT)
+*/
+
+#include "precomp.h"
+#include "paraformer-torch.h"
+#include "encode_converter.h"
+#include <cstddef>
+
+using namespace std;
+namespace funasr {
+
+ParaformerTorch::ParaformerTorch()
+:use_hotword(false){
+}
+
+// offline
+void ParaformerTorch::InitAsr(const std::string &am_model, const std::string &am_cmvn, const std::string &am_config, int thread_num){
+    LoadConfigFromYaml(am_config.c_str());
+    // knf options
+    fbank_opts_.frame_opts.dither = 0;
+    fbank_opts_.mel_opts.num_bins = n_mels;
+    fbank_opts_.frame_opts.samp_freq = asr_sample_rate;
+    fbank_opts_.frame_opts.window_type = window_type;
+    fbank_opts_.frame_opts.frame_shift_ms = frame_shift;
+    fbank_opts_.frame_opts.frame_length_ms = frame_length;
+    fbank_opts_.energy_floor = 0;
+    fbank_opts_.mel_opts.debug_mel = false;
+
+    vocab = new Vocab(am_config.c_str());
+	phone_set_ = new PhoneSet(am_config.c_str());
+    LoadCmvn(am_cmvn.c_str());
+
+    torch::DeviceType device = at::kCPU;
+    #ifdef USE_GPU
+    if (!torch::cuda::is_available()) {
+        LOG(ERROR) << "CUDA is not available! Please check your GPU settings";
+        exit(-1);
+    } else {
+        LOG(INFO) << "CUDA available! Running on GPU";
+        device = at::kCUDA;
+    }
+    #endif
+    #ifdef USE_IPEX
+    torch::jit::setTensorExprFuserEnabled(false);
+    #endif
+    torch::jit::script::Module model = torch::jit::load(am_model, device);
+    model_ = std::make_shared<TorchModule>(std::move(model));    
+}
+
+void ParaformerTorch::InitLm(const std::string &lm_file, 
+                        const std::string &lm_cfg_file, 
+                        const std::string &lex_file) {
+    try {
+        lm_ = std::shared_ptr<fst::Fst<fst::StdArc>>(
+            fst::Fst<fst::StdArc>::Read(lm_file));
+        if (lm_){
+            lm_vocab = new Vocab(lm_cfg_file.c_str(), lex_file.c_str());
+            LOG(INFO) << "Successfully load lm file " << lm_file;
+        }else{
+            LOG(ERROR) << "Failed to load lm file " << lm_file;
+        }
+    } catch (std::exception const &e) {
+        LOG(ERROR) << "Error when load lm file: " << e.what();
+        exit(0);
+    }
+}
+
+void ParaformerTorch::LoadConfigFromYaml(const char* filename){
+
+    YAML::Node config;
+    try{
+        config = YAML::LoadFile(filename);
+    }catch(exception const &e){
+        LOG(ERROR) << "Error loading file, yaml file error or not exist.";
+        exit(-1);
+    }
+
+    try{
+        YAML::Node frontend_conf = config["frontend_conf"];
+        this->asr_sample_rate = frontend_conf["fs"].as<int>();
+
+        YAML::Node lang_conf = config["lang"];
+        if (lang_conf.IsDefined()){
+            language = lang_conf.as<string>();
+        }
+    }catch(exception const &e){
+        LOG(ERROR) << "Error when load argument from vad config YAML.";
+        exit(-1);
+    }
+}
+
+void ParaformerTorch::InitHwCompiler(const std::string &hw_model, int thread_num) {
+    // TODO
+    use_hotword = true;
+}
+
+void ParaformerTorch::InitSegDict(const std::string &seg_dict_model) {
+    seg_dict = new SegDict(seg_dict_model.c_str());
+}
+
+ParaformerTorch::~ParaformerTorch()
+{
+    if(vocab){
+        delete vocab;
+    }
+    if(lm_vocab){
+        delete lm_vocab;
+    }
+    if(seg_dict){
+        delete seg_dict;
+    }
+    if(phone_set_){
+        delete phone_set_;
+    }
+}
+
+void ParaformerTorch::StartUtterance()
+{
+}
+
+void ParaformerTorch::EndUtterance()
+{
+}
+
+void ParaformerTorch::Reset()
+{
+}
+
+void ParaformerTorch::FbankKaldi(float sample_rate, const float* waves, int len, std::vector<std::vector<float>> &asr_feats) {
+    knf::OnlineFbank fbank_(fbank_opts_);
+    std::vector<float> buf(len);
+    for (int32_t i = 0; i != len; ++i) {
+        buf[i] = waves[i] * 32768;
+    }
+    fbank_.AcceptWaveform(sample_rate, buf.data(), buf.size());
+
+    int32_t frames = fbank_.NumFramesReady();
+    for (int32_t i = 0; i != frames; ++i) {
+        const float *frame = fbank_.GetFrame(i);
+        std::vector<float> frame_vector(frame, frame + fbank_opts_.mel_opts.num_bins);
+        asr_feats.emplace_back(frame_vector);
+    }
+}
+
+void ParaformerTorch::LoadCmvn(const char *filename)
+{
+    ifstream cmvn_stream(filename);
+    if (!cmvn_stream.is_open()) {
+        LOG(ERROR) << "Failed to open file: " << filename;
+        exit(-1);
+    }
+    string line;
+
+    while (getline(cmvn_stream, line)) {
+        istringstream iss(line);
+        vector<string> line_item{istream_iterator<string>{iss}, istream_iterator<string>{}};
+        if (line_item[0] == "<AddShift>") {
+            getline(cmvn_stream, line);
+            istringstream means_lines_stream(line);
+            vector<string> means_lines{istream_iterator<string>{means_lines_stream}, istream_iterator<string>{}};
+            if (means_lines[0] == "<LearnRateCoef>") {
+                for (int j = 3; j < means_lines.size() - 1; j++) {
+                    means_list_.push_back(stof(means_lines[j]));
+                }
+                continue;
+            }
+        }
+        else if (line_item[0] == "<Rescale>") {
+            getline(cmvn_stream, line);
+            istringstream vars_lines_stream(line);
+            vector<string> vars_lines{istream_iterator<string>{vars_lines_stream}, istream_iterator<string>{}};
+            if (vars_lines[0] == "<LearnRateCoef>") {
+                for (int j = 3; j < vars_lines.size() - 1; j++) {
+                    vars_list_.push_back(stof(vars_lines[j])*scale);
+                }
+                continue;
+            }
+        }
+    }
+}
+
+string ParaformerTorch::GreedySearch(float * in, int n_len,  int64_t token_nums, bool is_stamp, std::vector<float> us_alphas, std::vector<float> us_cif_peak)
+{
+    vector<int> hyps;
+    int Tmax = n_len;
+    for (int i = 0; i < Tmax; i++) {
+        int max_idx;
+        float max_val;
+        FindMax(in + i * token_nums, token_nums, max_val, max_idx);
+        hyps.push_back(max_idx);
+    }
+    if(!is_stamp){
+        return vocab->Vector2StringV2(hyps, language);
+    }else{
+        std::vector<string> char_list;
+        std::vector<std::vector<float>> timestamp_list;
+        std::string res_str;
+        vocab->Vector2String(hyps, char_list);
+        std::vector<string> raw_char(char_list);
+        TimestampOnnx(us_alphas, us_cif_peak, char_list, res_str, timestamp_list);
+
+        return PostProcess(raw_char, timestamp_list);
+    }
+}
+
+string ParaformerTorch::BeamSearch(WfstDecoder* &wfst_decoder, float *in, int len, int64_t token_nums)
+{
+  return wfst_decoder->Search(in, len, token_nums);
+}
+
+string ParaformerTorch::FinalizeDecode(WfstDecoder* &wfst_decoder,
+                                  bool is_stamp, std::vector<float> us_alphas, std::vector<float> us_cif_peak)
+{
+  return wfst_decoder->FinalizeDecode(is_stamp, us_alphas, us_cif_peak);
+}
+
+void ParaformerTorch::LfrCmvn(std::vector<std::vector<float>> &asr_feats) {
+
+    std::vector<std::vector<float>> out_feats;
+    int T = asr_feats.size();
+    int T_lrf = ceil(1.0 * T / lfr_n);
+
+    // Pad frames at start(copy first frame)
+    for (int i = 0; i < (lfr_m - 1) / 2; i++) {
+        asr_feats.insert(asr_feats.begin(), asr_feats[0]);
+    }
+    // Merge lfr_m frames as one,lfr_n frames per window
+    T = T + (lfr_m - 1) / 2;
+    std::vector<float> p;
+    for (int i = 0; i < T_lrf; i++) {
+        if (lfr_m <= T - i * lfr_n) {
+            for (int j = 0; j < lfr_m; j++) {
+                p.insert(p.end(), asr_feats[i * lfr_n + j].begin(), asr_feats[i * lfr_n + j].end());
+            }
+            out_feats.emplace_back(p);
+            p.clear();
+        } else {
+            // Fill to lfr_m frames at last window if less than lfr_m frames  (copy last frame)
+            int num_padding = lfr_m - (T - i * lfr_n);
+            for (int j = 0; j < (asr_feats.size() - i * lfr_n); j++) {
+                p.insert(p.end(), asr_feats[i * lfr_n + j].begin(), asr_feats[i * lfr_n + j].end());
+            }
+            for (int j = 0; j < num_padding; j++) {
+                p.insert(p.end(), asr_feats[asr_feats.size() - 1].begin(), asr_feats[asr_feats.size() - 1].end());
+            }
+            out_feats.emplace_back(p);
+            p.clear();
+        }
+    }
+    // Apply cmvn
+    for (auto &out_feat: out_feats) {
+        for (int j = 0; j < means_list_.size(); j++) {
+            out_feat[j] = (out_feat[j] + means_list_[j]) * vars_list_[j];
+        }
+    }
+    asr_feats = out_feats;
+}
+
+string ParaformerTorch::Forward(float* din, int len, bool input_finished, const std::vector<std::vector<float>> &hw_emb, void* decoder_handle)
+{
+    WfstDecoder* wfst_decoder = (WfstDecoder*)decoder_handle;
+    int32_t in_feat_dim = fbank_opts_.mel_opts.num_bins;
+
+    std::vector<std::vector<float>> asr_feats;
+    FbankKaldi(asr_sample_rate, din, len, asr_feats);
+    if(asr_feats.size() == 0){
+      return "";
+    }
+    LfrCmvn(asr_feats);
+    int32_t feat_dim = lfr_m*in_feat_dim;
+    int32_t num_frames = asr_feats.size();
+
+    std::vector<float> wav_feats;
+    for (const auto &frame_feat: asr_feats) {
+        wav_feats.insert(wav_feats.end(), frame_feat.begin(), frame_feat.end());
+    }
+    std::vector<int32_t> paraformer_length;
+    paraformer_length.emplace_back(num_frames);
+
+    torch::NoGradGuard no_grad;
+    torch::Tensor feats =
+        torch::from_blob(wav_feats.data(),
+                {1, num_frames, feat_dim}, torch::kFloat).contiguous();
+    torch::Tensor feat_lens = torch::from_blob(paraformer_length.data(),
+                        {1}, torch::kInt32);
+
+    // 2. forward
+    #ifdef USE_GPU
+    feats = feats.to(at::kCUDA);
+    feat_lens = feat_lens.to(at::kCUDA);
+    #endif
+    std::vector<torch::jit::IValue> inputs = {feats, feat_lens};
+
+    string result="";
+    try {
+        auto outputs = model_->forward(inputs).toTuple()->elements();
+        torch::Tensor am_scores;
+        torch::Tensor valid_token_lens;
+        #ifdef USE_GPU
+        am_scores = outputs[0].toTensor().to(at::kCPU);
+        valid_token_lens = outputs[1].toTensor().to(at::kCPU);
+        #else
+        am_scores = outputs[0].toTensor();
+        valid_token_lens = outputs[1].toTensor();
+        #endif
+        
+        if (lm_ == nullptr) {
+            result = GreedySearch(am_scores[0].data_ptr<float>(), valid_token_lens[0].item<int>(), am_scores.size(2));
+        } else {
+            result = BeamSearch(wfst_decoder, am_scores[0].data_ptr<float>(), valid_token_lens[0].item<int>(), am_scores.size(2));
+            if (input_finished) {
+                result = FinalizeDecode(wfst_decoder);
+            }
+        }        
+    }
+    catch (std::exception const &e)
+    {
+        LOG(ERROR)<<e.what();
+    }
+
+    return result;
+}
+
+std::vector<std::vector<float>> ParaformerTorch::CompileHotwordEmbedding(std::string &hotwords) {
+    std::vector<std::vector<float>> result;
+    return result;
+}
+
+Vocab* ParaformerTorch::GetVocab()
+{
+    return vocab;
+}
+
+Vocab* ParaformerTorch::GetLmVocab()
+{
+    return lm_vocab;
+}
+
+PhoneSet* ParaformerTorch::GetPhoneSet()
+{
+    return phone_set_;
+}
+
+string ParaformerTorch::Rescoring()
+{
+    LOG(ERROR)<<"Not Imp!!!!!!";
+    return "";
+}
+} // namespace funasr

runtime/onnxruntime/src/paraformer-torch.h

@@ -0,0 +1,92 @@
+/**
+ * Copyright FunASR (https://github.com/alibaba-damo-academy/FunASR). All Rights Reserved.
+ * MIT License  (https://opensource.org/licenses/MIT)
+*/
+#pragma once
+#include <torch/serialize.h>
+#include <torch/script.h>
+#include <torch/torch.h>
+#include <torch/csrc/jit/passes/tensorexpr_fuser.h>
+#include "precomp.h"
+#include "fst/fstlib.h"
+#include "fst/symbol-table.h"
+#include "bias-lm.h"
+#include "phone-set.h"
+
+namespace funasr {
+
+    class ParaformerTorch : 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:
+        Vocab* vocab = nullptr;
+        Vocab* lm_vocab = nullptr;
+        SegDict* seg_dict = nullptr;
+        PhoneSet* phone_set_ = nullptr;
+        //const float scale = 22.6274169979695;
+        const float scale = 1.0;
+
+        void LoadConfigFromYaml(const char* filename);
+        void LoadCmvn(const char *filename);
+        void LfrCmvn(std::vector<std::vector<float>> &asr_feats);
+
+        using TorchModule = torch::jit::script::Module;
+        std::shared_ptr<TorchModule> model_ = nullptr;
+        std::vector<torch::Tensor> encoder_outs_;
+        bool use_hotword;
+
+    public:
+        ParaformerTorch();
+        ~ParaformerTorch();
+        void InitAsr(const std::string &am_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();
+        void FbankKaldi(float sample_rate, const float* waves, int len, std::vector<std::vector<float>> &asr_feats);
+        string Forward(float* din, int len, bool input_finished=true, const std::vector<std::vector<float>> &hw_emb={{0.0}}, void* wfst_decoder=nullptr);
+        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});
+
+        string Rescoring();
+        string GetLang(){return language;};
+        int GetAsrSampleRate() { return asr_sample_rate; };
+        void StartUtterance();
+        void EndUtterance();
+        void InitLm(const std::string &lm_file, const std::string &lm_cfg_file, const std::string &lex_file);
+        string BeamSearch(WfstDecoder* &wfst_decoder, float* in, int n_len, int64_t token_nums);
+        string FinalizeDecode(WfstDecoder* &wfst_decoder,
+                          bool is_stamp=false, std::vector<float> us_alphas={0}, std::vector<float> us_cif_peak={0});
+        Vocab* GetVocab();
+        Vocab* GetLmVocab();
+        PhoneSet* GetPhoneSet();
+		
+        knf::FbankOptions fbank_opts_;
+        vector<float> means_list_;
+        vector<float> vars_list_;
+        int lfr_m = PARA_LFR_M;
+        int lfr_n = PARA_LFR_N;
+
+        // paraformer-offline
+        std::string language="zh-cn";
+
+        // lm
+        std::shared_ptr<fst::Fst<fst::StdArc>> lm_ = nullptr;
+
+        string window_type = "hamming";
+        int frame_length = 25;
+        int frame_shift = 10;
+        int n_mels = 80;
+        int encoder_size = 512;
+        int fsmn_layers = 16;
+        int fsmn_lorder = 10;
+        int fsmn_dims = 512;
+        float cif_threshold = 1.0;
+        float tail_alphas = 0.45;
+        int asr_sample_rate = MODEL_SAMPLE_RATE;
+    };
+
+} // namespace funasr

runtime/onnxruntime/src/precomp.h

@@ -64,6 +64,7 @@ using namespace std;
 #include "seg_dict.h"
 #include "resample.h"
 #include "paraformer.h"
+#include "paraformer-torch.h"
 #include "paraformer-online.h"
 #include "offline-stream.h"
 #include "tpass-stream.h"