add dolphin

model.py

@@ -164,7 +164,11 @@ def get_pretrained_model(
     decoding_method: str,
     num_active_paths: int,
 ) -> Union[sherpa.OfflineRecognizer, sherpa.OnlineRecognizer]:
-    if repo_id in chinese_models:
+    if repo_id in multi_lingual_models:
+        return multi_lingual_models[repo_id](
+            repo_id, decoding_method=decoding_method, num_active_paths=num_active_paths
+        )
+    elif repo_id in chinese_models:
         return chinese_models[repo_id](
             repo_id, decoding_method=decoding_method, num_active_paths=num_active_paths
         )
@@ -921,6 +925,43 @@ def _get_alimeeting_pre_trained_model(
     return recognizer
 
 
+@lru_cache(maxsize=4)
+def _get_dolphin_ctc_models(repo_id: str, decoding_method: str, num_active_paths: int):
+    assert repo_id in [
+        "csukuangfj/sherpa-onnx-dolphin-base-ctc-multi-lang-int8-2025-04-02",
+        "csukuangfj/sherpa-onnx-dolphin-small-ctc-multi-lang-int8-2025-04-02",
+        "csukuangfj/sherpa-onnx-dolphin-base-ctc-multi-lang-2025-04-02",
+        "csukuangfj/sherpa-onnx-dolphin-small-ctc-multi-lang-2025-04-02",
+    ], repo_id
+
+    if repo_id in [
+        "csukuangfj/sherpa-onnx-dolphin-base-ctc-multi-lang-int8-2025-04-02",
+        "csukuangfj/sherpa-onnx-dolphin-small-ctc-multi-lang-int8-2025-04-02",
+    ]:
+        use_int8 = True
+    else:
+        use_int8 = False
+
+    nn_model = _get_nn_model_filename(
+        repo_id=repo_id,
+        filename="model.int8.onnx" if use_int8 else "model.onnx",
+        subfolder=".",
+    )
+    tokens = _get_token_filename(
+        repo_id=repo_id,
+        filename="tokens.txt",
+        subfolder=".",
+    )
+
+    recognizer = sherpa_onnx.OfflineRecognizer.from_dolphin_ctc(
+        tokens=tokens,
+        model=nn_model,
+        num_threads=2,
+    )
+
+    return recognizer
+
+
 @lru_cache(maxsize=10)
 def _get_wenet_model(
     repo_id: str,
@@ -1780,6 +1821,13 @@ english_models = {
     "csukuangfj/wenet-english-model": _get_wenet_model,
 }
 
+multi_lingual_models = {
+    "csukuangfj/sherpa-onnx-dolphin-base-ctc-multi-lang-int8-2025-04-02": _get_dolphin_ctc_models,
+    "csukuangfj/sherpa-onnx-dolphin-small-ctc-multi-lang-int8-2025-04-02": _get_dolphin_ctc_models,
+    "csukuangfj/sherpa-onnx-dolphin-base-ctc-multi-lang-2025-04-02": _get_dolphin_ctc_models,
+    "csukuangfj/sherpa-onnx-dolphin-small-ctc-multi-lang-2025-04-02": _get_dolphin_ctc_models,
+}
+
 chinese_english_mixed_models = {
     "csukuangfj/sherpa-onnx-fire-red-asr-large-zh_en-2025-02-16": _get_fire_red_asr_models,
     "csukuangfj/sherpa-onnx-streaming-zipformer-bilingual-zh-en-2023-02-20": _get_streaming_zipformer_pre_trained_model,
@@ -1843,6 +1891,7 @@ thai_models = {
 
 
 all_models = {
+    **multi_lingual_models,
     **chinese_models,
     **english_models,
     **chinese_english_mixed_models,
@@ -1860,6 +1909,7 @@ all_models = {
 }
 
 language_to_models = {
+    "Multi-lingual (east aisa)": list(multi_lingual_models.keys()),
     "超多种中文方言": list(chinese_dialect_models.keys()),
     "Chinese": list(chinese_models.keys()),
     "English": list(english_models.keys()),

requirements.txt

@@ -10,6 +10,6 @@ numpy
 
 huggingface_hub
 
-https://huggingface.co/csukuangfj/sherpa-onnx-wheels/resolve/main/cpu/1.10.44/sherpa_onnx-1.10.44-cp38-cp38-manylinux_2_17_x86_64.manylinux2014_x86_64.whl
+https://huggingface.co/csukuangfj/sherpa-onnx-wheels/resolve/main/cpu/1.11.2/sherpa_onnx-1.11.2-cp38-cp38-manylinux_2_17_x86_64.manylinux2014_x86_64.whl
 
 #sherpa-onnx>=1.10.36