update

toolbox/k2_sherpa/examples.py

@@ -11,4 +11,12 @@ examples = [
         "Yes",
         "./data/test_wavs/paraformer-zh/si_chuan_hua.wav",
     ],
+    [
+        "English",
+        "csukuangfj/sherpa-onnx-whisper-tiny.en",
+        "greedy_search",
+        4,
+        "Yes",
+        "./data/test_wavs/librispeech/1089-134686-0001.wav",
+    ],
 ]

toolbox/k2_sherpa/nn_models.py

@@ -150,6 +150,16 @@ model_map = {
         },
     ],
     "English": [
+        {
+            "repo_id": "csukuangfj/sherpa-onnx-whisper-tiny.en",
+            "encoder_model_file": "tiny.en-encoder.int8.onnx",
+            "encoder_model_file_sub_folder": ".",
+            "decoder_model_file": "tiny.en-decoder.int8.onnx",
+            "decoder_model_file_sub_folder": ".",
+            "tokens_file": "tiny.en-tokens.txt",
+            "tokens_file_sub_folder": ".",
+            "loader": "load_sherpa_offline_recognizer_from_whisper",
+        },
         {
             "repo_id": "csukuangfj/sherpa-onnx-paraformer-en-2024-03-09",
             "nn_model_file": "model.int8.onnx",
@@ -307,6 +317,20 @@ def load_sherpa_offline_recognizer_from_transducer(encoder_model_file: str,
     return recognizer
 
 
+def load_sherpa_offline_recognizer_from_whisper(encoder_model_file: str,
+                                                decoder_model_file: str,
+                                                tokens_file: str,
+                                                num_threads: int = 2,
+                                                ):
+    recognizer = sherpa_onnx.OfflineRecognizer.from_whisper(
+        encoder=encoder_model_file,
+        decoder=decoder_model_file,
+        tokens=tokens_file,
+        num_threads=num_threads,
+    )
+    return recognizer
+
+
 def load_recognizer(local_model_dir: Path,
                     decoding_method: str = "greedy_search",
                     num_active_paths: int = 4,