[update] Update sherpa-onnx commit reference, add fire-red-asr-ctc model configuration, and include model in llm_pack

Author: LittleMouse

ext_components/sherpa-onnx/SConstruct

@@ -12,7 +12,7 @@ _SDK_PATH = os.path.normpath(
 
 env["GIT_REPO_LISTS"]["sherpa-onnx"] = {
     "url": "https://github.com/Abandon-ht/sherpa-onnx.git",
-    "commit": "2d08532b0ddbe96817242324d826442ba4d3a8a4",
+    "commit": "69fa06760c06f24d657aab6beb9150a34de5730c",
     "path": str(Path(_SDK_PATH) / "github_source" / "sherpa-onnx"),
 }
 

projects/llm_framework/main_asr/mode_fire-red-asr-ctc-10s-ax650.json

@@ -0,0 +1,26 @@
+{
+    "mode": "fire-red-asr-ctc-10s-ax650",
+    "type": "asr",
+    "homepage": "https://huggingface.co/Abandon-ht/FireRedASR2.axera",
+    "pulsar_version": "5.1-patch1-cd6c30b4",
+    "capabilities": [
+        "Chinese",
+        "English"
+    ],
+    "input_type": [
+        "sys.pcm",
+        "sys.cap.0_0"
+    ],
+    "output_type": [
+        "asr.utf-8"
+    ],
+    "mode_param": {
+        "model_config.fire_red_asr_ctc.model": "model-10s.axmodel",
+        "model_config.tokens": "tokens.txt",
+        "silero_vad.model": "silero_vad.ort",
+        "model_config.provider": "axera",
+        "silence_timeout": 2000,
+        "awake_delay": 50
+    },
+    "mode_param_bak": {}
+}

projects/llm_framework/main_asr/src/main.cpp

@@ -280,6 +280,7 @@ class llm_task {
         ONNX_ASR_CONFIG_AUTO_SET(file_body["mode_param"], model_config.canary.tgt_lang);
         ONNX_ASR_CONFIG_AUTO_SET(file_body["mode_param"], model_config.canary.use_pnc);
         ONNX_ASR_CONFIG_AUTO_SET(file_body["mode_param"], model_config.omnilingual.model);
+        ONNX_ASR_CONFIG_AUTO_SET(file_body["mode_param"], model_config.fire_red_asr_ctc.model);
         ONNX_ASR_CONFIG_AUTO_SET(file_body["mode_param"], model_config.telespeech_ctc);
         ONNX_ASR_CONFIG_AUTO_SET(file_body["mode_param"], model_config.tokens);
         ONNX_ASR_CONFIG_AUTO_SET(file_body["mode_param"], model_config.num_threads);
@@ -333,9 +334,14 @@ class llm_task {
         else if (file_body["mode_param"].contains("silence_timeout"))
             silence_timeout = file_body["mode_param"]["silence_timeout"];
 
-        onnx_asr_config_.model_config.sense_voice.model = base_model + onnx_asr_config_.model_config.sense_voice.model;
-        onnx_asr_config_.model_config.tokens            = base_model + onnx_asr_config_.model_config.tokens;
-        vad_config_.silero_vad.model                    = base_model + vad_config_.silero_vad.model;
+        if (!onnx_asr_config_.model_config.sense_voice.model.empty())
+            onnx_asr_config_.model_config.sense_voice.model =
+                base_model + onnx_asr_config_.model_config.sense_voice.model;
+        if (!onnx_asr_config_.model_config.fire_red_asr_ctc.model.empty())
+            onnx_asr_config_.model_config.fire_red_asr_ctc.model =
+                base_model + onnx_asr_config_.model_config.fire_red_asr_ctc.model;
+        onnx_asr_config_.model_config.tokens = base_model + onnx_asr_config_.model_config.tokens;
+        vad_config_.silero_vad.model         = base_model + vad_config_.silero_vad.model;
 
         onnx_recognizer_ = std::make_unique<sherpa_onnx::OfflineRecognizer>(onnx_asr_config_);
         vad_             = std::make_unique<sherpa_onnx::VoiceActivityDetector>(vad_config_);
@@ -1118,7 +1124,7 @@ class llm_asr : public StackFlow {
             ret                    = llm_channel->subscriber_work_id(
                 data,
                 std::bind(&llm_asr::kws_awake, this, std::weak_ptr<llm_task>(llm_task_obj),
-                                             std::weak_ptr<llm_channel_obj>(llm_channel), std::placeholders::_1, std::placeholders::_2));
+                          std::weak_ptr<llm_channel_obj>(llm_channel), std::placeholders::_1, std::placeholders::_2));
             llm_task_obj->inputs_.push_back(data);
         }
 

projects/llm_framework/tools/llm_pack.py

@@ -410,6 +410,7 @@ def create_bin_deb(package_name, version, src_folder, revision = 'm5stack1', dep
         'llm-model-sense-voice-small-10s':[create_data_deb,'llm-model-sense-voice-small-10s', '0.4', src_folder, revision],
         ## AX650
         'llm-model-sense-voice-small-10s-ax650':[create_data_deb,'llm-model-sense-voice-small-10s-ax650', '0.4', src_folder, revision],
+        'llm-model-fire-red-asr-ctc-10s-ax650':[create_data_deb,'llm-model-fire-red-asr-ctc-10s-ax650', '0.4', src_folder, revision],
         # TTS model
         'llm-model-single-speaker-english-fast':[create_data_deb,'llm-model-single-speaker-english-fast', '0.3', src_folder, revision],
         'llm-model-single-speaker-fast':[create_data_deb,'llm-model-single-speaker-fast', '0.3', src_folder, revision],