[update] perf kws & melotts & vad model setup.

Author: Abandon-ht

projects/llm_framework/main_kws/mode_sherpa-onnx-kws-zipformer-gigaspeech-3.3M-2024-01-01.json

@@ -13,9 +13,9 @@
         "kws.bool"
     ],
     "mode_param":{
-        "model_config.transducer.encoder":"encoder-epoch-12-avg-2-chunk-16-left-64.int8.onnx",
-        "model_config.transducer.decoder":"decoder-epoch-12-avg-2-chunk-16-left-64.onnx",
-        "model_config.transducer.joiner":"joiner-epoch-12-avg-2-chunk-16-left-64.int8.onnx",
+        "model_config.transducer.encoder":"encoder-epoch-12-avg-2-chunk-16-left-64.int8.ort",
+        "model_config.transducer.decoder":"decoder-epoch-12-avg-2-chunk-16-left-64.ort",
+        "model_config.transducer.joiner":"joiner-epoch-12-avg-2-chunk-16-left-64.int8.ort",
         "model_config.tokens":"tokens.txt",
         "model_config.modeling_unit":"cjkchar",
         "keywords_file":"keywords.txt",

projects/llm_framework/main_kws/mode_sherpa-onnx-kws-zipformer-wenetspeech-3.3M-2024-01-01.json

@@ -12,10 +12,10 @@
     "output_type":[
         "kws.bool"
     ],
-    "mode_param":{
-        "model_config.transducer.encoder":"encoder-epoch-12-avg-2-chunk-16-left-64.int8.onnx",
-        "model_config.transducer.decoder":"decoder-epoch-12-avg-2-chunk-16-left-64.onnx",
-        "model_config.transducer.joiner":"joiner-epoch-12-avg-2-chunk-16-left-64.int8.onnx",
+    "mode_param": {
+        "model_config.transducer.encoder":"encoder-epoch-12-avg-2-chunk-16-left-64.int8.ort",
+        "model_config.transducer.decoder":"decoder-epoch-12-avg-2-chunk-16-left-64.ort",
+        "model_config.transducer.joiner":"joiner-epoch-12-avg-2-chunk-16-left-64.int8.ort",
         "model_config.tokens":"tokens.txt",
         "model_config.modeling_unit":"cjkchar",
         "keywords_file":"keywords.txt",

projects/llm_framework/main_melotts/SConstruct

@@ -26,6 +26,9 @@ REQUIREMENTS += ['onnxruntime', 'samplerate']
 INCLUDE += [ADir('../include')]
 INCLUDE += [ADir('src/runner'), ADir('../include/onnxruntime/core/session')]
 
+LINK_SEARCH_PATH += [ADir('../static_lib/sherpa/onnx')]
+LDFLAGS += ['-l:libcargs.a', '-l:libonnxruntime.a']
+
 STATIC_FILES += Glob('mode_*.json')
 
 env['COMPONENTS'].append({'target':'llm_melotts',

projects/llm_framework/main_melotts/mode_melotts-zh-cn.json

@@ -12,7 +12,7 @@
         "sys.play.0_1"
     ],
     "mode_param": {
-        "encoder": "encoder.onnx",
+        "encoder": "encoder.ort",
         "decoder": "decoder.axmodel",
         "gbin": "g.bin",
         "tokens": "tokens.txt",

projects/llm_framework/main_vad/mode_silero-vad-model.json

@@ -12,7 +12,7 @@
     "vad.bool"
   ],
   "mode_param": {
-    "silero_vad.model": "silero_vad.onnx"
+    "silero_vad.model": "silero_vad.ort"
   },
   "mode_param_bak": {
     "silero_vad.threshold": 0.5,

projects/llm_framework/tools/llm_pack.py

@@ -289,14 +289,14 @@ def create_bin_deb(package_name, version, src_folder, revision = 'm5stack1'):
         create_data_deb('llm-audio-zh-cn', data_version, src_folder, revision)
         create_data_deb('llm-sherpa-ncnn-streaming-zipformer-20M-2023-02-17', data_version, src_folder, revision)
         create_data_deb('llm-sherpa-ncnn-streaming-zipformer-zh-14M-2023-02-23', data_version, src_folder, revision)
-        create_data_deb('llm-sherpa-onnx-kws-zipformer-gigaspeech-3.3M-2024-01-01', data_version, src_folder, revision)
-        create_data_deb('llm-sherpa-onnx-kws-zipformer-wenetspeech-3.3M-2024-01-01', data_version, src_folder, revision)
+        create_data_deb('llm-sherpa-onnx-kws-zipformer-gigaspeech-3.3M-2024-01-01', '0.3', src_folder, revision)
+        create_data_deb('llm-sherpa-onnx-kws-zipformer-wenetspeech-3.3M-2024-01-01', '0.3', src_folder, revision)
         # create_data_deb('llm-qwen2-0.5B-prefill-20e', data_version, src_folder, revision)
         # create_data_deb('llm-qwen2-1.5B-prefill-20e', data_version, src_folder, revision)
         create_data_deb('llm-qwen2.5-0.5B-prefill-20e', data_version, src_folder, revision)
         create_data_deb('llm-single-speaker-english-fast', data_version, src_folder, revision)
         create_data_deb('llm-single-speaker-fast', data_version, src_folder, revision)
-        create_data_deb('llm-melotts-zh-cn', data_version, src_folder, revision)
+        create_data_deb('llm-melotts-zh-cn', '0.3', src_folder, revision)
         create_data_deb('llm-yolo11n', data_version, src_folder, revision)
         create_data_deb('llm-yolo11n-pose', data_version, src_folder, revision)
         create_data_deb('llm-yolo11n-seg', data_version, src_folder, revision)