update

examples/wenet/infer.py

@@ -39,8 +39,11 @@ def main():
     args = get_args()
 
     model_dir = Path(args.model_dir)
-    model_filename = model_dir / "final.zip"
-    tokens_filename = model_dir / "units.txt"
+    nn_model_file = model_dir / "final.zip"
+    tokens_file = model_dir / "units.txt"
+
+    print("nn_model_file: {}".format(nn_model_file))
+    print("tokens_file: {}".format(tokens_file))
 
     feat_config = sherpa.FeatureConfig(normalize_samples=False)
     feat_config.fbank_opts.frame_opts.samp_freq = args.sample_rate
@@ -48,8 +51,8 @@ def main():
     feat_config.fbank_opts.frame_opts.dither = 0
 
     config = sherpa.OfflineRecognizerConfig(
-        nn_model=model_filename.as_posix(),
-        tokens=tokens_filename.as_posix(),
+        nn_model=nn_model_file.as_posix(),
+        tokens=tokens_file.as_posix(),
         use_gpu=False,
         feat_config=feat_config,
         decoding_method="greedy_search",

toolbox/k2_sherpa/models.py

@@ -103,6 +103,11 @@ def load_recognizer(repo_id: str,
         )
 
     if recognizer_type == EnumRecognizerType.sherpa_offline_recognizer.value:
+        print("nn_model_file: {}".format(nn_model_file))
+        print("tokens_file: {}".format(tokens_file))
+        print("decoding_method: {}".format(decoding_method))
+        print("num_active_paths: {}".format(num_active_paths))
+
         recognizer = load_sherpa_offline_recognizer(
             nn_model_file=nn_model_file,
             tokens_file=tokens_file,
@@ -110,7 +115,7 @@ def load_recognizer(repo_id: str,
             num_active_paths=num_active_paths,
         )
     else:
-        raise NotImplementedError("recognizer_type not support: {}".format(recognizer_type.value))
+        raise NotImplementedError("recognizer_type not support: {}".format(recognizer_type))
     return recognizer