Update model.py

model.py

@@ -362,72 +362,72 @@ def _get_offline_pre_trained_model(
     return recognizer
 
 
-# @lru_cache(maxsize=10)
-# def _get_vietnamese_pretrained_model(
-#     repo_id: str, decoding_method: str, num_active_paths: int
-# ) -> sherpa_onnx.OfflineRecognizer:
-#     assert repo_id in (
-#         "csukuangfj/sherpa-onnx-zipformer-vi-int8-2025-04-20",
-#         "csukuangfj/sherpa-onnx-zipformer-vi-2025-04-20",
-#     ), repo_id
-
-#     # decoder_model = _get_nn_model_filename(
-#     #     repo_id=repo_id,
-#     #     filename="decoder-epoch-12-avg-8.onnx",
-#     #     subfolder=".",
-#     # )
-
-#     decoder_model = "decoder-epoch-45-avg-25.onnx"
-
-#     if repo_id == "csukuangfj/sherpa-onnx-zipformer-vi-int8-2025-04-20":
-#         # encoder_model = _get_nn_model_filename(
-#         #     repo_id=repo_id,
-#         #     filename="encoder-epoch-12-avg-8.int8.onnx",
-#         #     subfolder=".",
-#         # )
-
-#         encoder_model = "encoder-epoch-45-avg-25.int8.onnx"
-
-#         # joiner_model = _get_nn_model_filename(
-#         #     repo_id=repo_id,
-#         #     filename="joiner-epoch-12-avg-8.int8.onnx",
-#         #     subfolder=".",
-#         # )
-#         joiner_model = "joiner-epoch-45-avg-25.int8.onnx"
-#     elif repo_id == "csukuangfj/sherpa-onnx-zipformer-vi-2025-04-20":
-#         # encoder_model = _get_nn_model_filename(
-#         #     repo_id=repo_id,
-#         #     filename="encoder-epoch-12-avg-8.onnx",
-#         #     subfolder=".",
-#         # )
-
-#         encoder_model = "encoder-epoch-45-avg-25.onnx"
-
-#         # joiner_model = _get_nn_model_filename(
-#         #     repo_id=repo_id,
-#         #     filename="joiner-epoch-12-avg-8.onnx",
-#         #     subfolder=".",
-#         # )
-#         joiner_model = "joiner-epoch-45-avg-25.onnx"
-#     else:
-#         raise ValueError(f"repo_id: {repo_id}")
-
-#     # tokens = _get_token_filename(repo_id=repo_id, subfolder=".")
-
-#     tokens = "tokens.txt"
-
-#     recognizer = sherpa_onnx.OfflineRecognizer.from_transducer(
-#         tokens=tokens,
-#         encoder=encoder_model,
-#         decoder=decoder_model,
-#         joiner=joiner_model,
-#         num_threads=2,
-#         sample_rate=16000,
-#         feature_dim=80,
-#         decoding_method=decoding_method,
-#     )
-
-#     return recognizer
+@lru_cache(maxsize=10)
+def _get_gigaspeech_pre_trained_model(
+    repo_id: str, decoding_method: str, num_active_paths: int
+) -> sherpa_onnx.OfflineRecognizer:
+    # assert repo_id in (
+    #     "csukuangfj/sherpa-onnx-zipformer-vi-int8-2025-04-20",
+    #     "csukuangfj/sherpa-onnx-zipformer-vi-2025-04-20",
+    # ), repo_id
+
+    # decoder_model = _get_nn_model_filename(
+    #     repo_id=repo_id,
+    #     filename="decoder-epoch-12-avg-8.onnx",
+    #     subfolder=".",
+    # )
+
+    decoder_model = "decoder-epoch-45-avg-25.onnx"
+
+    if repo_id == "csukuangfj/sherpa-onnx-zipformer-vi-int8-2025-04-20":
+        # encoder_model = _get_nn_model_filename(
+        #     repo_id=repo_id,
+        #     filename="encoder-epoch-12-avg-8.int8.onnx",
+        #     subfolder=".",
+        # )
+
+        encoder_model = "encoder-epoch-45-avg-25.int8.onnx"
+
+        # joiner_model = _get_nn_model_filename(
+        #     repo_id=repo_id,
+        #     filename="joiner-epoch-12-avg-8.int8.onnx",
+        #     subfolder=".",
+        # )
+        joiner_model = "joiner-epoch-45-avg-25.int8.onnx"
+    elif repo_id == "csukuangfj/sherpa-onnx-zipformer-vi-2025-04-20":
+        # encoder_model = _get_nn_model_filename(
+        #     repo_id=repo_id,
+        #     filename="encoder-epoch-12-avg-8.onnx",
+        #     subfolder=".",
+        # )
+
+        encoder_model = "encoder-epoch-45-avg-25.onnx"
+
+        # joiner_model = _get_nn_model_filename(
+        #     repo_id=repo_id,
+        #     filename="joiner-epoch-12-avg-8.onnx",
+        #     subfolder=".",
+        # )
+        joiner_model = "joiner-epoch-45-avg-25.onnx"
+    else:
+        raise ValueError(f"repo_id: {repo_id}")
+
+    # tokens = _get_token_filename(repo_id=repo_id, subfolder=".")
+
+    tokens = "tokens.txt"
+
+    recognizer = sherpa_onnx.OfflineRecognizer.from_transducer(
+        tokens=tokens,
+        encoder=encoder_model,
+        decoder=decoder_model,
+        joiner=joiner_model,
+        num_threads=2,
+        sample_rate=16000,
+        feature_dim=80,
+        decoding_method=decoding_method,
+    )
+
+    return recognizer
 
 
 @lru_cache(maxsize=10)