xuning / sherpaonnx

转到一个项目
Toggle navigation Toggle navigation pinning
Fangjun Kuang
Committed by GitHub
6ec57327 1 parent a8556e31

add vad+sense voice example for C API (#1291)

内嵌 并排对比
正在显示 9 个修改的文件 包含 267 行增加 和 43 行删除
@@ -99,6 +99,45 @@ jobs: @@ -99,6 +99,45 @@ jobs:
99 ./run.sh 99 ./run.sh
100 rm -rf sherpa-onnx-streaming-zipformer-bilingual-zh-en-2023-02-20 100 rm -rf sherpa-onnx-streaming-zipformer-bilingual-zh-en-2023-02-20
101 101
  102 + - name: Test vad + sense-voice
  103 + shell: bash
  104 + run: |
  105 + gcc -o vad-sense-voice-c-api ./c-api-examples/vad-sense-voice-c-api.c \
  106 + -I ./build/install/include \
  107 + -L ./build/install/lib/ \
  108 + -l sherpa-onnx-c-api \
  109 + -l onnxruntime
  110 +
  111 + ls -lh vad-sense-voice-c-api
  112 +
  113 + if [[ ${{ matrix.os }} == ubuntu-latest ]]; then
  114 + ldd ./vad-sense-voice-c-api
  115 + echo "----"
  116 + readelf -d ./vad-sense-voice-c-api
  117 + fi
  118 +
  119 + # Now download models
  120 + #
  121 + curl -SL -O https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/silero_vad.onnx
  122 + curl -SL -O https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/lei-jun-test.wav
  123 +
  124 + curl -SL -O https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/sherpa-onnx-sense-voice-zh-en-ja-ko-yue-2024-07-17.tar.bz2
  125 + tar xvf sherpa-onnx-sense-voice-zh-en-ja-ko-yue-2024-07-17.tar.bz2
  126 + rm sherpa-onnx-sense-voice-zh-en-ja-ko-yue-2024-07-17.tar.bz2
  127 +
  128 + ls -lh sherpa-onnx-sense-voice-zh-en-ja-ko-yue-2024-07-17
  129 + echo "---"
  130 + ls -lh sherpa-onnx-sense-voice-zh-en-ja-ko-yue-2024-07-17/test_wavs
  131 +
  132 + export LD_LIBRARY_PATH=$PWD/build/install/lib:$LD_LIBRARY_PATH
  133 + export DYLD_LIBRARY_PATH=$PWD/build/install/lib:$DYLD_LIBRARY_PATH
  134 +
  135 + ./vad-sense-voice-c-api
  136 +
  137 + rm -rf sherpa-onnx-sense-voice-*
  138 + rm -rf *.onnx
  139 + rm *.wav
  140 +
102 - name: Test sense-voice 141 - name: Test sense-voice
103 shell: bash 142 shell: bash
104 run: | 143 run: |
@@ -45,6 +45,9 @@ target_link_libraries(streaming-paraformer-c-api sherpa-onnx-c-api) @@ -45,6 +45,9 @@ target_link_libraries(streaming-paraformer-c-api sherpa-onnx-c-api)
45 add_executable(telespeech-c-api telespeech-c-api.c) 45 add_executable(telespeech-c-api telespeech-c-api.c)
46 target_link_libraries(telespeech-c-api sherpa-onnx-c-api) 46 target_link_libraries(telespeech-c-api sherpa-onnx-c-api)
47 47
  48 +add_executable(vad-sense-voice-c-api vad-sense-voice-c-api.c)
  49 +target_link_libraries(vad-sense-voice-c-api sherpa-onnx-c-api)
  50 +
48 if(SHERPA_ONNX_HAS_ALSA) 51 if(SHERPA_ONNX_HAS_ALSA)
49 add_subdirectory(./asr-microphone-example) 52 add_subdirectory(./asr-microphone-example)
50 elseif((UNIX AND NOT APPLE) OR LINUX) 53 elseif((UNIX AND NOT APPLE) OR LINUX)
@@ -3,7 +3,8 @@ @@ -3,7 +3,8 @@
3 // Copyright (c) 2024 Xiaomi Corporation 3 // Copyright (c) 2024 Xiaomi Corporation
4 4
5 // 5 //
6 -// This file demonstrates how to use non-streaming Paraformer with sherpa-onnx's C API. 6 +// This file demonstrates how to use non-streaming Paraformer with sherpa-onnx's
  7 +// C API.
7 // clang-format off 8 // clang-format off
8 // 9 //
9 // wget https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/sherpa-onnx-paraformer-zh-small-2024-03-09.tar.bz2 10 // wget https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/sherpa-onnx-paraformer-zh-small-2024-03-09.tar.bz2
@@ -19,19 +20,20 @@ @@ -19,19 +20,20 @@
19 #include "sherpa-onnx/c-api/c-api.h" 20 #include "sherpa-onnx/c-api/c-api.h"
20 21
21 int32_t main() { 22 int32_t main() {
22 -  
23 - const char *wav_filename = "sherpa-onnx-paraformer-zh-small-2024-03-09/test_wavs/0.wav";  
24 - const char *model_filename = "sherpa-onnx-paraformer-zh-small-2024-03-09/model.int8.onnx";  
25 - const char *tokens_filename = "sherpa-onnx-paraformer-zh-small-2024-03-09/tokens.txt"; 23 + const char *wav_filename =
  24 + "sherpa-onnx-paraformer-zh-small-2024-03-09/test_wavs/0.wav";
  25 + const char *model_filename =
  26 + "sherpa-onnx-paraformer-zh-small-2024-03-09/model.int8.onnx";
  27 + const char *tokens_filename =
  28 + "sherpa-onnx-paraformer-zh-small-2024-03-09/tokens.txt";
26 const char *provider = "cpu"; 29 const char *provider = "cpu";
27 30
28 -  
29 const SherpaOnnxWave *wave = SherpaOnnxReadWave(wav_filename); 31 const SherpaOnnxWave *wave = SherpaOnnxReadWave(wav_filename);
30 if (wave == NULL) { 32 if (wave == NULL) {
31 fprintf(stderr, "Failed to read %s\n", wav_filename); 33 fprintf(stderr, "Failed to read %s\n", wav_filename);
32 return -1; 34 return -1;
33 } 35 }
34 - 36 +
35 // Paraformer config 37 // Paraformer config
36 SherpaOnnxOfflineParaformerModelConfig paraformer_config; 38 SherpaOnnxOfflineParaformerModelConfig paraformer_config;
37 memset(&paraformer_config, 0, sizeof(paraformer_config)); 39 memset(&paraformer_config, 0, sizeof(paraformer_config));
@@ -19,8 +19,6 @@ @@ -19,8 +19,6 @@
19 #include "sherpa-onnx/c-api/c-api.h" 19 #include "sherpa-onnx/c-api/c-api.h"
20 20
21 int32_t main() { 21 int32_t main() {
22 - // You can find more test waves from  
23 - // https://hf-mirror.com/spaces/k2-fsa/spoken-language-identification/tree/main/test_wavs  
24 const char *wav_filename = 22 const char *wav_filename =
25 "./sherpa-onnx-sense-voice-zh-en-ja-ko-yue-2024-07-17/test_wavs/en.wav"; 23 "./sherpa-onnx-sense-voice-zh-en-ja-ko-yue-2024-07-17/test_wavs/en.wav";
26 const char *model_filename = 24 const char *model_filename =
@@ -3,7 +3,8 @@ @@ -3,7 +3,8 @@
3 // Copyright (c) 2024 Xiaomi Corporation 3 // Copyright (c) 2024 Xiaomi Corporation
4 4
5 // 5 //
6 -// This file demonstrates how to use streaming Paraformer with sherpa-onnx's C API. 6 +// This file demonstrates how to use streaming Paraformer with sherpa-onnx's C
  7 +// API.
7 // clang-format off 8 // clang-format off
8 // 9 //
9 // wget https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/sherpa-onnx-streaming-paraformer-bilingual-zh-en.tar.bz2 10 // wget https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/sherpa-onnx-streaming-paraformer-bilingual-zh-en.tar.bz2
@@ -19,26 +20,27 @@ @@ -19,26 +20,27 @@
19 #include "sherpa-onnx/c-api/c-api.h" 20 #include "sherpa-onnx/c-api/c-api.h"
20 21
21 int32_t main() { 22 int32_t main() {
22 -  
23 - const char *wav_filename = "sherpa-onnx-streaming-paraformer-bilingual-zh-en/test_wavs/0.wav";  
24 - const char *encoder_filename = "sherpa-onnx-streaming-paraformer-bilingual-zh-en/encoder.int8.onnx";  
25 - const char *decoder_filename = "sherpa-onnx-streaming-paraformer-bilingual-zh-en/decoder.int8.onnx";  
26 - const char *tokens_filename = "sherpa-onnx-streaming-paraformer-bilingual-zh-en/tokens.txt"; 23 + const char *wav_filename =
  24 + "sherpa-onnx-streaming-paraformer-bilingual-zh-en/test_wavs/0.wav";
  25 + const char *encoder_filename =
  26 + "sherpa-onnx-streaming-paraformer-bilingual-zh-en/encoder.int8.onnx";
  27 + const char *decoder_filename =
  28 + "sherpa-onnx-streaming-paraformer-bilingual-zh-en/decoder.int8.onnx";
  29 + const char *tokens_filename =
  30 + "sherpa-onnx-streaming-paraformer-bilingual-zh-en/tokens.txt";
27 const char *provider = "cpu"; 31 const char *provider = "cpu";
28 32
29 -  
30 const SherpaOnnxWave *wave = SherpaOnnxReadWave(wav_filename); 33 const SherpaOnnxWave *wave = SherpaOnnxReadWave(wav_filename);
31 if (wave == NULL) { 34 if (wave == NULL) {
32 fprintf(stderr, "Failed to read %s\n", wav_filename); 35 fprintf(stderr, "Failed to read %s\n", wav_filename);
33 return -1; 36 return -1;
34 } 37 }
35 - 38 +
36 // Paraformer config 39 // Paraformer config
37 SherpaOnnxOnlineParaformerModelConfig paraformer_config; 40 SherpaOnnxOnlineParaformerModelConfig paraformer_config;
38 memset(&paraformer_config, 0, sizeof(paraformer_config)); 41 memset(&paraformer_config, 0, sizeof(paraformer_config));
39 paraformer_config.encoder = encoder_filename; 42 paraformer_config.encoder = encoder_filename;
40 paraformer_config.decoder = decoder_filename; 43 paraformer_config.decoder = decoder_filename;
41 -  
42 44
43 // Online model config 45 // Online model config
44 SherpaOnnxOnlineModelConfig online_model_config; 46 SherpaOnnxOnlineModelConfig online_model_config;
@@ -3,7 +3,8 @@ @@ -3,7 +3,8 @@
3 // Copyright (c) 2024 Xiaomi Corporation 3 // Copyright (c) 2024 Xiaomi Corporation
4 4
5 // 5 //
6 -// This file demonstrates how to use streaming Zipformer with sherpa-onnx's C API. 6 +// This file demonstrates how to use streaming Zipformer with sherpa-onnx's C
  7 +// API.
7 // clang-format off 8 // clang-format off
8 // 9 //
9 // wget https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/sherpa-onnx-streaming-zipformer-en-20M-2023-02-17.tar.bz2 10 // wget https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/sherpa-onnx-streaming-zipformer-en-20M-2023-02-17.tar.bz2
@@ -19,28 +20,33 @@ @@ -19,28 +20,33 @@
19 #include "sherpa-onnx/c-api/c-api.h" 20 #include "sherpa-onnx/c-api/c-api.h"
20 21
21 int32_t main() { 22 int32_t main() {
22 -  
23 - const char *wav_filename = "sherpa-onnx-streaming-zipformer-en-20M-2023-02-17/test_wavs/0.wav";  
24 - const char *encoder_filename = "sherpa-onnx-streaming-zipformer-en-20M-2023-02-17/encoder-epoch-99-avg-1.onnx";  
25 - const char *decoder_filename = "sherpa-onnx-streaming-zipformer-en-20M-2023-02-17/decoder-epoch-99-avg-1.onnx";  
26 - const char *joiner_filename = "sherpa-onnx-streaming-zipformer-en-20M-2023-02-17/joiner-epoch-99-avg-1.onnx";  
27 - const char *tokens_filename = "sherpa-onnx-streaming-zipformer-en-20M-2023-02-17/tokens.txt"; 23 + const char *wav_filename =
  24 + "sherpa-onnx-streaming-zipformer-en-20M-2023-02-17/test_wavs/0.wav";
  25 + const char *encoder_filename =
  26 + "sherpa-onnx-streaming-zipformer-en-20M-2023-02-17/"
  27 + "encoder-epoch-99-avg-1.onnx";
  28 + const char *decoder_filename =
  29 + "sherpa-onnx-streaming-zipformer-en-20M-2023-02-17/"
  30 + "decoder-epoch-99-avg-1.onnx";
  31 + const char *joiner_filename =
  32 + "sherpa-onnx-streaming-zipformer-en-20M-2023-02-17/"
  33 + "joiner-epoch-99-avg-1.onnx";
  34 + const char *tokens_filename =
  35 + "sherpa-onnx-streaming-zipformer-en-20M-2023-02-17/tokens.txt";
28 const char *provider = "cpu"; 36 const char *provider = "cpu";
29 37
30 -  
31 const SherpaOnnxWave *wave = SherpaOnnxReadWave(wav_filename); 38 const SherpaOnnxWave *wave = SherpaOnnxReadWave(wav_filename);
32 if (wave == NULL) { 39 if (wave == NULL) {
33 fprintf(stderr, "Failed to read %s\n", wav_filename); 40 fprintf(stderr, "Failed to read %s\n", wav_filename);
34 return -1; 41 return -1;
35 } 42 }
36 - 43 +
37 // Zipformer config 44 // Zipformer config
38 SherpaOnnxOnlineTransducerModelConfig zipformer_config; 45 SherpaOnnxOnlineTransducerModelConfig zipformer_config;
39 memset(&zipformer_config, 0, sizeof(zipformer_config)); 46 memset(&zipformer_config, 0, sizeof(zipformer_config));
40 zipformer_config.encoder = encoder_filename; 47 zipformer_config.encoder = encoder_filename;
41 zipformer_config.decoder = decoder_filename; 48 zipformer_config.decoder = decoder_filename;
42 zipformer_config.joiner = joiner_filename; 49 zipformer_config.joiner = joiner_filename;
43 -  
44 50
45 // Online model config 51 // Online model config
46 SherpaOnnxOnlineModelConfig online_model_config; 52 SherpaOnnxOnlineModelConfig online_model_config;
@@ -3,7 +3,8 @@ @@ -3,7 +3,8 @@
3 // Copyright (c) 2024 Xiaomi Corporation 3 // Copyright (c) 2024 Xiaomi Corporation
4 4
5 // 5 //
6 -// This file demonstrates how to use TeleSpeech-ASR CTC model with sherpa-onnx's C API. 6 +// This file demonstrates how to use TeleSpeech-ASR CTC model with sherpa-onnx's
  7 +// C API.
7 // clang-format off 8 // clang-format off
8 // 9 //
9 // wget https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/sherpa-onnx-telespeech-ctc-int8-zh-2024-06-04.tar.bz2 10 // wget https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/sherpa-onnx-telespeech-ctc-int8-zh-2024-06-04.tar.bz2
@@ -19,10 +20,12 @@ @@ -19,10 +20,12 @@
19 #include "sherpa-onnx/c-api/c-api.h" 20 #include "sherpa-onnx/c-api/c-api.h"
20 21
21 int32_t main() { 22 int32_t main() {
22 -  
23 - const char *wav_filename = "sherpa-onnx-telespeech-ctc-int8-zh-2024-06-04/test_wavs/3-sichuan.wav";  
24 - const char *model_filename = "sherpa-onnx-telespeech-ctc-int8-zh-2024-06-04/model.int8.onnx";  
25 - const char *tokens_filename = "sherpa-onnx-telespeech-ctc-int8-zh-2024-06-04/tokens.txt"; 23 + const char *wav_filename =
  24 + "sherpa-onnx-telespeech-ctc-int8-zh-2024-06-04/test_wavs/3-sichuan.wav";
  25 + const char *model_filename =
  26 + "sherpa-onnx-telespeech-ctc-int8-zh-2024-06-04/model.int8.onnx";
  27 + const char *tokens_filename =
  28 + "sherpa-onnx-telespeech-ctc-int8-zh-2024-06-04/tokens.txt";
26 const char *provider = "cpu"; 29 const char *provider = "cpu";
27 30
28 const SherpaOnnxWave *wave = SherpaOnnxReadWave(wav_filename); 31 const SherpaOnnxWave *wave = SherpaOnnxReadWave(wav_filename);
  1 +// c-api-examples/vad-sense-voice-c-api.c
  2 +//
  3 +// Copyright (c) 2024 Xiaomi Corporation
  4 +
  5 +//
  6 +// This file demonstrates how to use VAD + SenseVoice with sherpa-onnx's C API.
  7 +// clang-format off
  8 +//
  9 +// wget https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/silero_vad.onnx
  10 +// wget https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/lei-jun-test.wav
  11 +//
  12 +// wget https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/sherpa-onnx-sense-voice-zh-en-ja-ko-yue-2024-07-17.tar.bz2
  13 +// tar xvf sherpa-onnx-sense-voice-zh-en-ja-ko-yue-2024-07-17.tar.bz2
  14 +// rm sherpa-onnx-sense-voice-zh-en-ja-ko-yue-2024-07-17.tar.bz2
  15 +//
  16 +// clang-format on
  17 +
  18 +#include <stdio.h>
  19 +#include <stdlib.h>
  20 +#include <string.h>
  21 +
  22 +#include "sherpa-onnx/c-api/c-api.h"
  23 +
  24 +int32_t main() {
  25 + const char *wav_filename = "./lei-jun-test.wav";
  26 + const char *vad_filename = "./silero_vad.onnx";
  27 + const char *model_filename =
  28 + "./sherpa-onnx-sense-voice-zh-en-ja-ko-yue-2024-07-17/model.int8.onnx";
  29 + const char *tokens_filename =
  30 + "./sherpa-onnx-sense-voice-zh-en-ja-ko-yue-2024-07-17/tokens.txt";
  31 + const char *language = "auto";
  32 + const char *provider = "cpu";
  33 + int32_t use_inverse_text_normalization = 1;
  34 +
  35 + const SherpaOnnxWave *wave = SherpaOnnxReadWave(wav_filename);
  36 + if (wave == NULL) {
  37 + fprintf(stderr, "Failed to read %s\n", wav_filename);
  38 + return -1;
  39 + }
  40 +
  41 + if (wave->sample_rate != 16000) {
  42 + fprintf(stderr, "Expect the sample rate to be 16000. Given: %d\n",
  43 + wave->sample_rate);
  44 + SherpaOnnxFreeWave(wave);
  45 + return -1;
  46 + }
  47 +
  48 + SherpaOnnxOfflineSenseVoiceModelConfig sense_voice_config;
  49 + memset(&sense_voice_config, 0, sizeof(sense_voice_config));
  50 + sense_voice_config.model = model_filename;
  51 + sense_voice_config.language = language;
  52 + sense_voice_config.use_itn = use_inverse_text_normalization;
  53 +
  54 + // Offline model config
  55 + SherpaOnnxOfflineModelConfig offline_model_config;
  56 + memset(&offline_model_config, 0, sizeof(offline_model_config));
  57 + offline_model_config.debug = 0;
  58 + offline_model_config.num_threads = 1;
  59 + offline_model_config.provider = provider;
  60 + offline_model_config.tokens = tokens_filename;
  61 + offline_model_config.sense_voice = sense_voice_config;
  62 +
  63 + // Recognizer config
  64 + SherpaOnnxOfflineRecognizerConfig recognizer_config;
  65 + memset(&recognizer_config, 0, sizeof(recognizer_config));
  66 + recognizer_config.decoding_method = "greedy_search";
  67 + recognizer_config.model_config = offline_model_config;
  68 +
  69 + SherpaOnnxOfflineRecognizer *recognizer =
  70 + SherpaOnnxCreateOfflineRecognizer(&recognizer_config);
  71 +
  72 + if (recognizer == NULL) {
  73 + fprintf(stderr, "Please check your recognizer config!\n");
  74 + SherpaOnnxFreeWave(wave);
  75 + return -1;
  76 + }
  77 +
  78 + SherpaOnnxVadModelConfig vadConfig;
  79 + memset(&vadConfig, 0, sizeof(vadConfig));
  80 + vadConfig.silero_vad.model = vad_filename;
  81 + vadConfig.silero_vad.threshold = 0.5;
  82 + vadConfig.silero_vad.min_silence_duration = 0.5;
  83 + vadConfig.silero_vad.min_speech_duration = 0.5;
  84 + vadConfig.silero_vad.window_size = 512;
  85 + vadConfig.sample_rate = 16000;
  86 + vadConfig.num_threads = 1;
  87 + vadConfig.debug = 1;
  88 +
  89 + SherpaOnnxVoiceActivityDetector *vad =
  90 + SherpaOnnxCreateVoiceActivityDetector(&vadConfig, 30);
  91 +
  92 + if (vad == NULL) {
  93 + fprintf(stderr, "Please check your recognizer config!\n");
  94 + SherpaOnnxFreeWave(wave);
  95 + SherpaOnnxDestroyOfflineRecognizer(recognizer);
  96 + return -1;
  97 + }
  98 +
  99 + int32_t window_size = vadConfig.silero_vad.window_size;
  100 + int32_t i = 0;
  101 +
  102 + while (i + window_size < wave->num_samples) {
  103 + SherpaOnnxVoiceActivityDetectorAcceptWaveform(vad, wave->samples + i,
  104 + window_size);
  105 + i += window_size;
  106 +
  107 + while (!SherpaOnnxVoiceActivityDetectorEmpty(vad)) {
  108 + const SherpaOnnxSpeechSegment *segment =
  109 + SherpaOnnxVoiceActivityDetectorFront(vad);
  110 +
  111 + SherpaOnnxOfflineStream *stream =
  112 + SherpaOnnxCreateOfflineStream(recognizer);
  113 + SherpaOnnxAcceptWaveformOffline(stream, wave->sample_rate,
  114 + segment->samples, segment->n);
  115 +
  116 + SherpaOnnxDecodeOfflineStream(recognizer, stream);
  117 +
  118 + const SherpaOnnxOfflineRecognizerResult *result =
  119 + SherpaOnnxGetOfflineStreamResult(stream);
  120 +
  121 + float start = segment->start / 16000.0f;
  122 + float duration = segment->n / 16000.0f;
  123 + float stop = start + duration;
  124 +
  125 + fprintf(stderr, "%.3f -- %.3f: %s\n", start, stop, result->text);
  126 +
  127 + SherpaOnnxDestroyOfflineRecognizerResult(result);
  128 + SherpaOnnxDestroyOfflineStream(stream);
  129 +
  130 + SherpaOnnxDestroySpeechSegment(segment);
  131 + SherpaOnnxVoiceActivityDetectorPop(vad);
  132 + }
  133 + }
  134 +
  135 + SherpaOnnxVoiceActivityDetectorFlush(vad);
  136 +
  137 + while (!SherpaOnnxVoiceActivityDetectorEmpty(vad)) {
  138 + const SherpaOnnxSpeechSegment *segment =
  139 + SherpaOnnxVoiceActivityDetectorFront(vad);
  140 +
  141 + SherpaOnnxOfflineStream *stream = SherpaOnnxCreateOfflineStream(recognizer);
  142 + SherpaOnnxAcceptWaveformOffline(stream, wave->sample_rate, segment->samples,
  143 + segment->n);
  144 +
  145 + SherpaOnnxDecodeOfflineStream(recognizer, stream);
  146 +
  147 + const SherpaOnnxOfflineRecognizerResult *result =
  148 + SherpaOnnxGetOfflineStreamResult(stream);
  149 +
  150 + float start = segment->start / 16000.0f;
  151 + float duration = segment->n / 16000.0f;
  152 + float stop = start + duration;
  153 +
  154 + fprintf(stderr, "%.3f -- %.3f: %s\n", start, stop, result->text);
  155 +
  156 + SherpaOnnxDestroyOfflineRecognizerResult(result);
  157 + SherpaOnnxDestroyOfflineStream(stream);
  158 +
  159 + SherpaOnnxDestroySpeechSegment(segment);
  160 + SherpaOnnxVoiceActivityDetectorPop(vad);
  161 + }
  162 +
  163 + SherpaOnnxDestroyOfflineRecognizer(recognizer);
  164 + SherpaOnnxDestroyVoiceActivityDetector(vad);
  165 + SherpaOnnxFreeWave(wave);
  166 +
  167 + return 0;
  168 +}
@@ -3,7 +3,8 @@ @@ -3,7 +3,8 @@
3 // Copyright (c) 2024 Xiaomi Corporation 3 // Copyright (c) 2024 Xiaomi Corporation
4 4
5 // 5 //
6 -// This file demonstrates how to use non-streaming Zipformer with sherpa-onnx's C API. 6 +// This file demonstrates how to use non-streaming Zipformer with sherpa-onnx's
  7 +// C API.
7 // clang-format off 8 // clang-format off
8 // 9 //
9 // wget https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/sherpa-onnx-zipformer-small-en-2023-06-26.tar.bz2 10 // wget https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/sherpa-onnx-zipformer-small-en-2023-06-26.tar.bz2
@@ -19,28 +20,30 @@ @@ -19,28 +20,30 @@
19 #include "sherpa-onnx/c-api/c-api.h" 20 #include "sherpa-onnx/c-api/c-api.h"
20 21
21 int32_t main() { 22 int32_t main() {
22 -  
23 - const char *wav_filename = "sherpa-onnx-zipformer-small-en-2023-06-26/test_wavs/0.wav";  
24 - const char *encoder_filename = "sherpa-onnx-zipformer-small-en-2023-06-26/encoder-epoch-99-avg-1.onnx";  
25 - const char *decoder_filename = "sherpa-onnx-zipformer-small-en-2023-06-26/decoder-epoch-99-avg-1.onnx";  
26 - const char *joiner_filename = "sherpa-onnx-zipformer-small-en-2023-06-26/joiner-epoch-99-avg-1.onnx";  
27 - const char *tokens_filename = "sherpa-onnx-zipformer-small-en-2023-06-26/tokens.txt"; 23 + const char *wav_filename =
  24 + "sherpa-onnx-zipformer-small-en-2023-06-26/test_wavs/0.wav";
  25 + const char *encoder_filename =
  26 + "sherpa-onnx-zipformer-small-en-2023-06-26/encoder-epoch-99-avg-1.onnx";
  27 + const char *decoder_filename =
  28 + "sherpa-onnx-zipformer-small-en-2023-06-26/decoder-epoch-99-avg-1.onnx";
  29 + const char *joiner_filename =
  30 + "sherpa-onnx-zipformer-small-en-2023-06-26/joiner-epoch-99-avg-1.onnx";
  31 + const char *tokens_filename =
  32 + "sherpa-onnx-zipformer-small-en-2023-06-26/tokens.txt";
28 const char *provider = "cpu"; 33 const char *provider = "cpu";
29 34
30 -  
31 const SherpaOnnxWave *wave = SherpaOnnxReadWave(wav_filename); 35 const SherpaOnnxWave *wave = SherpaOnnxReadWave(wav_filename);
32 if (wave == NULL) { 36 if (wave == NULL) {
33 fprintf(stderr, "Failed to read %s\n", wav_filename); 37 fprintf(stderr, "Failed to read %s\n", wav_filename);
34 return -1; 38 return -1;
35 } 39 }
36 - 40 +
37 // Zipformer config 41 // Zipformer config
38 SherpaOnnxOfflineTransducerModelConfig zipformer_config; 42 SherpaOnnxOfflineTransducerModelConfig zipformer_config;
39 memset(&zipformer_config, 0, sizeof(zipformer_config)); 43 memset(&zipformer_config, 0, sizeof(zipformer_config));
40 zipformer_config.encoder = encoder_filename; 44 zipformer_config.encoder = encoder_filename;
41 zipformer_config.decoder = decoder_filename; 45 zipformer_config.decoder = decoder_filename;
42 zipformer_config.joiner = joiner_filename; 46 zipformer_config.joiner = joiner_filename;
43 -  
44 47
45 // Offline model config 48 // Offline model config
46 SherpaOnnxOfflineModelConfig offline_model_config; 49 SherpaOnnxOfflineModelConfig offline_model_config;