Add C API for FireRedAsr AED model. (#1871)

@@ -35,7 +35,7 @@ jobs:
     strategy:
       fail-fast: false
       matrix:
-        os: [ubuntu-latest, macos-latest]
+        os: [ubuntu-latest, macos-latest, ubuntu-22.04-arm]
 
     steps:
       - uses: actions/checkout@v4
@@ -69,7 +69,7 @@ jobs:
           ls -lh install/lib
           ls -lh install/include
 
-          if [[ ${{ matrix.os }} == ubuntu-latest ]]; then
+          if [[ ${{ matrix.os }} == ubuntu-latest || ${{ matrix.os }} == ubuntu-22.04-arm ]]; then
             ldd ./install/lib/libsherpa-onnx-c-api.so
             echo "---"
             readelf -d ./install/lib/libsherpa-onnx-c-api.so
@@ -79,6 +79,38 @@ jobs:
             otool -L ./install/lib/libsherpa-onnx-c-api.dylib
           fi
 
+      - name: Test FireRedAsr
+        shell: bash
+        run: |
+          gcc -o fire-red-asr-c-api ./c-api-examples/fire-red-asr-c-api.c \
+            -I ./build/install/include \
+            -L ./build/install/lib/ \
+            -l sherpa-onnx-c-api \
+            -l onnxruntime
+
+          ls -lh fire-red-asr-c-api
+
+          if [[ ${{ matrix.os }} == ubuntu-latest || ${{ matrix.os }} == ubuntu-22.04-arm ]]; then
+            ldd ./fire-red-asr-c-api
+            echo "----"
+            readelf -d ./fire-red-asr-c-api
+          fi
+
+          curl -SL -O https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/sherpa-onnx-fire-red-asr-large-zh_en-2025-02-16.tar.bz2
+          tar xvf sherpa-onnx-fire-red-asr-large-zh_en-2025-02-16.tar.bz2
+          rm sherpa-onnx-fire-red-asr-large-zh_en-2025-02-16.tar.bz2
+
+          ls -lh sherpa-onnx-fire-red-asr-large-zh_en-2025-02-16
+          echo "---"
+          ls -lh sherpa-onnx-fire-red-asr-large-zh_en-2025-02-16/test_wavs
+
+          export LD_LIBRARY_PATH=$PWD/build/install/lib:$LD_LIBRARY_PATH
+          export DYLD_LIBRARY_PATH=$PWD/build/install/lib:$DYLD_LIBRARY_PATH
+
+          ./fire-red-asr-c-api
+
+          rm -rf sherpa-onnx-fire-red-asr-*
+
       - name: Test kws (zh)
         shell: bash
         run: |
@@ -301,7 +333,7 @@ jobs:
 
           ls -lh vad-sense-voice-c-api
 
-          if [[ ${{ matrix.os }} == ubuntu-latest ]]; then
+          if [[ ${{ matrix.os }} == ubuntu-latest || ${{ matrix.os }} == ubuntu-22.04-arm ]]; then
             ldd ./vad-sense-voice-c-api
             echo "----"
             readelf -d ./vad-sense-voice-c-api
@@ -340,7 +372,7 @@ jobs:
 
           ls -lh sense-voice-c-api
 
-          if [[ ${{ matrix.os }} == ubuntu-latest ]]; then
+          if [[ ${{ matrix.os }} == ubuntu-latest || ${{ matrix.os }} == ubuntu-22.04-arm ]]; then
             ldd ./sense-voice-c-api
             echo "----"
             readelf -d ./sense-voice-c-api
@@ -373,7 +405,7 @@ jobs:
 
           ls -lh whisper-c-api
 
-          if [[ ${{ matrix.os }} == ubuntu-latest ]]; then
+          if [[ ${{ matrix.os }} == ubuntu-latest || ${{ matrix.os }} == ubuntu-22.04-arm ]]; then
             ldd ./whisper-c-api
             echo "----"
             readelf -d ./whisper-c-api
@@ -405,7 +437,7 @@ jobs:
 
           ls -lh zipformer-c-api
 
-          if [[ ${{ matrix.os }} == ubuntu-latest ]]; then
+          if [[ ${{ matrix.os }} == ubuntu-latest || ${{ matrix.os }} == ubuntu-22.04-arm ]]; then
             ldd ./zipformer-c-api
             echo "----"
             readelf -d ./zipformer-c-api
@@ -437,7 +469,7 @@ jobs:
 
           ls -lh streaming-zipformer-c-api
 
-          if [[ ${{ matrix.os }} == ubuntu-latest ]]; then
+          if [[ ${{ matrix.os }} == ubuntu-latest || ${{ matrix.os }} == ubuntu-22.04-arm ]]; then
             ldd ./streaming-zipformer-c-api
             echo "----"
             readelf -d ./streaming-zipformer-c-api
@@ -469,7 +501,7 @@ jobs:
 
           ls -lh paraformer-c-api
 
-          if [[ ${{ matrix.os }} == ubuntu-latest ]]; then
+          if [[ ${{ matrix.os }} == ubuntu-latest || ${{ matrix.os }} == ubuntu-22.04-arm ]]; then
             ldd ./paraformer-c-api
             echo "----"
             readelf -d ./paraformer-c-api
@@ -501,7 +533,7 @@ jobs:
 
           ls -lh streaming-paraformer-c-api
 
-          if [[ ${{ matrix.os }} == ubuntu-latest ]]; then
+          if [[ ${{ matrix.os }} == ubuntu-latest || ${{ matrix.os }} == ubuntu-22.04-arm ]]; then
             ldd ./streaming-paraformer-c-api
             echo "----"
             readelf -d ./streaming-paraformer-c-api

@@ -47,6 +47,9 @@ target_link_libraries(add-punctuation-c-api sherpa-onnx-c-api)
 add_executable(whisper-c-api whisper-c-api.c)
 target_link_libraries(whisper-c-api sherpa-onnx-c-api)
 
+add_executable(fire-red-asr-c-api fire-red-asr-c-api.c)
+target_link_libraries(fire-red-asr-c-api sherpa-onnx-c-api)
+
 add_executable(sense-voice-c-api sense-voice-c-api.c)
 target_link_libraries(sense-voice-c-api sherpa-onnx-c-api)
 

+// c-api-examples/fire-red-asr-c-api.c
+//
+// Copyright (c)  2025  Xiaomi Corporation
+
+// We assume you have pre-downloaded the FireRedAsr model
+// from https://github.com/k2-fsa/sherpa-onnx/releases/tag/asr-models
+// An example is given below:
+//
+// clang-format off
+//
+// wget https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/sherpa-onnx-fire-red-asr-large-zh_en-2025-02-16.tar.bz2
+// tar xvf sherpa-onnx-fire-red-asr-large-zh_en-2025-02-16.tar.bz2
+// rm sherpa-onnx-fire-red-asr-large-zh_en-2025-02-16.tar.bz2
+//
+// clang-format on
+
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+
+#include "sherpa-onnx/c-api/c-api.h"
+
+int32_t main() {
+  const char *wav_filename =
+      "./sherpa-onnx-fire-red-asr-large-zh_en-2025-02-16/test_wavs/0.wav";
+  const char *encoder_filename =
+      "sherpa-onnx-fire-red-asr-large-zh_en-2025-02-16/encoder.int8.onnx";
+  const char *decoder_filename =
+      "sherpa-onnx-fire-red-asr-large-zh_en-2025-02-16/decoder.int8.onnx";
+  const char *tokens_filename =
+      "sherpa-onnx-fire-red-asr-large-zh_en-2025-02-16/tokens.txt";
+  const char *provider = "cpu";
+
+  const SherpaOnnxWave *wave = SherpaOnnxReadWave(wav_filename);
+  if (wave == NULL) {
+    fprintf(stderr, "Failed to read %s\n", wav_filename);
+    return -1;
+  }
+
+  // Offline model config
+  SherpaOnnxOfflineModelConfig offline_model_config;
+  memset(&offline_model_config, 0, sizeof(offline_model_config));
+  offline_model_config.debug = 1;
+  offline_model_config.num_threads = 1;
+  offline_model_config.provider = provider;
+  offline_model_config.tokens = tokens_filename;
+  offline_model_config.fire_red_asr.encoder = encoder_filename;
+  offline_model_config.fire_red_asr.decoder = decoder_filename;
+
+  // Recognizer config
+  SherpaOnnxOfflineRecognizerConfig recognizer_config;
+  memset(&recognizer_config, 0, sizeof(recognizer_config));
+  recognizer_config.decoding_method = "greedy_search";
+  recognizer_config.model_config = offline_model_config;
+
+  const SherpaOnnxOfflineRecognizer *recognizer =
+      SherpaOnnxCreateOfflineRecognizer(&recognizer_config);
+
+  if (recognizer == NULL) {
+    fprintf(stderr, "Please check your config!\n");
+
+    SherpaOnnxFreeWave(wave);
+
+    return -1;
+  }
+
+  const SherpaOnnxOfflineStream *stream =
+      SherpaOnnxCreateOfflineStream(recognizer);
+
+  SherpaOnnxAcceptWaveformOffline(stream, wave->sample_rate, wave->samples,
+                                  wave->num_samples);
+  SherpaOnnxDecodeOfflineStream(recognizer, stream);
+  const SherpaOnnxOfflineRecognizerResult *result =
+      SherpaOnnxGetOfflineStreamResult(stream);
+
+  fprintf(stderr, "Decoded text: %s\n", result->text);
+
+  SherpaOnnxDestroyOfflineRecognizerResult(result);
+  SherpaOnnxDestroyOfflineStream(stream);
+  SherpaOnnxDestroyOfflineRecognizer(recognizer);
+  SherpaOnnxFreeWave(wave);
+
+  return 0;
+}

@@ -460,6 +460,12 @@ static sherpa_onnx::OfflineRecognizerConfig GetOfflineRecognizerConfig(
   recognizer_config.model_config.moonshine.cached_decoder =
       SHERPA_ONNX_OR(config->model_config.moonshine.cached_decoder, "");
 
+  recognizer_config.model_config.fire_red_asr.encoder =
+      SHERPA_ONNX_OR(config->model_config.fire_red_asr.encoder, "");
+
+  recognizer_config.model_config.fire_red_asr.decoder =
+      SHERPA_ONNX_OR(config->model_config.fire_red_asr.decoder, "");
+
   recognizer_config.lm_config.model =
       SHERPA_ONNX_OR(config->lm_config.model, "");
   recognizer_config.lm_config.scale =

@@ -389,6 +389,11 @@ SHERPA_ONNX_API typedef struct SherpaOnnxOfflineWhisperModelConfig {
   int32_t tail_paddings;
 } SherpaOnnxOfflineWhisperModelConfig;
 
+SHERPA_ONNX_API typedef struct SherpaOnnxOfflineFireRedAsrModelConfig {
+  const char *encoder;
+  const char *decoder;
+} SherpaOnnxOfflineFireRedAsrModelConfig;
+
 SHERPA_ONNX_API typedef struct SherpaOnnxOfflineMoonshineModelConfig {
   const char *preprocessor;
   const char *encoder;
@@ -432,6 +437,7 @@ SHERPA_ONNX_API typedef struct SherpaOnnxOfflineModelConfig {
   const char *telespeech_ctc;
   SherpaOnnxOfflineSenseVoiceModelConfig sense_voice;
   SherpaOnnxOfflineMoonshineModelConfig moonshine;
+  SherpaOnnxOfflineFireRedAsrModelConfig fire_red_asr;
 } SherpaOnnxOfflineModelConfig;
 
 SHERPA_ONNX_API typedef struct SherpaOnnxOfflineRecognizerConfig {