update

toolbox/k2_sherpa/nn_models.py

@@ -393,6 +393,34 @@ model_map = {
             "loader": "load_sherpa_onnx_offline_recognizer_from_transducer",
         },
     ],
+    "Japanese": [
+        {
+            "repo_id": "TeoWenShen/icefall-asr-csj-pruned-transducer-stateless7-streaming-230208-fluent",
+            "encoder_model_file": "encoder_jit_trace.pt",
+            "encoder_model_file_sub_folder": "exp_fluent",
+            "decoder_model_file": "decoder_jit_trace.pt",
+            "decoder_model_file_sub_folder": "exp_fluent",
+            "joiner_model_file": "joiner_jit_trace.pt",
+            "joiner_model_file_sub_folder": "exp_fluent",
+            "tokens_file": "tokens.txt",
+            "tokens_file_sub_folder": "data/lang_char",
+            "normalize_samples": True,
+            "loader": "load_sherpa_online_recognizer",
+        },
+        {
+            "repo_id": "TeoWenShen/icefall-asr-csj-pruned-transducer-stateless7-streaming-230208-disfluent",
+            "encoder_model_file": "encoder_jit_trace.pt",
+            "encoder_model_file_sub_folder": "exp_disfluent",
+            "decoder_model_file": "decoder_jit_trace.pt",
+            "decoder_model_file_sub_folder": "exp_disfluent",
+            "joiner_model_file": "joiner_jit_trace.pt",
+            "joiner_model_file_sub_folder": "exp_disfluent",
+            "tokens_file": "tokens.txt",
+            "tokens_file_sub_folder": "data/lang_char",
+            "normalize_samples": True,
+            "loader": "load_sherpa_online_recognizer",
+        },
+    ],
     "Tibetan": [
         {
             "repo_id": "syzym/icefall-asr-xbmu-amdo31-pruned-transducer-stateless7-2022-12-02",
@@ -555,6 +583,44 @@ def load_sherpa_onnx_offline_recognizer_from_whisper(encoder_model_file: str,
     return recognizer
 
 
+def load_sherpa_online_recognizer(nn_model_file: str,
+                                  encoder_model_file: str,
+                                  decoder_model_file: str,
+                                  joiner_model_file: str,
+                                  tokens_file: str,
+                                  sample_rate: int = 16000,
+                                  num_active_paths: int = 2,
+                                  decoding_method: str = "greedy_search",
+                                  num_mel_bins: int = 80,
+                                  frame_dither: int = 0,
+                                  normalize_samples: bool = False,
+                                  ):
+    feat_config = sherpa.FeatureConfig(normalize_samples=normalize_samples)
+    feat_config.fbank_opts.frame_opts.samp_freq = sample_rate
+    feat_config.fbank_opts.mel_opts.num_bins = num_mel_bins
+    feat_config.fbank_opts.frame_opts.dither = frame_dither
+
+    if not os.path.exists(nn_model_file):
+        raise AssertionError("nn_model_file not found. nn_model_file: {}".format(nn_model_file))
+
+    config = sherpa.OfflineRecognizerConfig(
+        nn_model=nn_model_file,
+        encoder_model=encoder_model_file,
+        decoder_model=decoder_model_file,
+        joiner_model=joiner_model_file,
+        tokens=tokens_file,
+        use_gpu=False,
+        feat_config=feat_config,
+        decoding_method=decoding_method,
+        num_active_paths=num_active_paths,
+        chunk_size=32,
+    )
+
+    recognizer = sherpa.OnlineRecognizer(config)
+
+    return recognizer
+
+
 def load_sherpa_onnx_online_recognizer_from_transducer(encoder_model_file: str,
                                                        decoder_model_file: str,
                                                        joiner_model_file: str,
@@ -653,6 +719,12 @@ def load_recognizer(local_model_dir: Path,
         recognizer = load_sherpa_onnx_offline_recognizer_from_whisper(
             **kwargs_
         )
+    elif loader == "load_sherpa_online_recognizer":
+        recognizer = load_sherpa_online_recognizer(
+            decoding_method=decoding_method,
+            num_active_paths=num_active_paths,
+            **kwargs_
+        )
     elif loader == "load_sherpa_onnx_online_recognizer_from_transducer":
         recognizer = load_sherpa_onnx_online_recognizer_from_transducer(
             **kwargs_