add moonshine

model.py

@@ -363,6 +363,59 @@ def _get_multi_zh_hans_pre_trained_model(repo_id):
     return recognizer
 
 
+@lru_cache(maxsize=10)
+def _get_moonshine_model(repo_id: str) -> sherpa_onnx.OfflineRecognizer:
+    assert repo_id in ("moonshine-tiny", "moonshine-base"), repo_id
+
+    if repo_id == "moonshine-tiny":
+        full_repo_id = "csukuangfj/sherpa-onnx-moonshine-tiny-en-int8"
+    elif repo_id == "moonshine-base":
+        full_repo_id = "csukuangfj/sherpa-onnx-moonshine-base-en-int8"
+    else:
+        raise ValueError(f"Unknown repo_id: {repo_id}")
+
+    preprocessor = _get_nn_model_filename(
+        repo_id=full_repo_id,
+        filename=f"preprocess.onnx",
+        subfolder=".",
+    )
+
+    encoder = _get_nn_model_filename(
+        repo_id=full_repo_id,
+        filename=f"encode.int8.onnx",
+        subfolder=".",
+    )
+
+    uncached_decoder = _get_nn_model_filename(
+        repo_id=full_repo_id,
+        filename=f"uncached_decode.int8.onnx",
+        subfolder=".",
+    )
+
+    cached_decoder = _get_nn_model_filename(
+        repo_id=full_repo_id,
+        filename=f"cached_decode.int8.onnx",
+        subfolder=".",
+    )
+
+    tokens = _get_token_filename(
+        repo_id=full_repo_id,
+        subfolder=".",
+        filename="tokens.txt",
+    )
+
+    recognizer = sherpa_onnx.OfflineRecognizer.from_moonshine(
+        preprocessor=preprocessor,
+        encoder=encoder,
+        uncached_decoder=uncached_decoder,
+        cached_decoder=cached_decoder,
+        tokens=tokens,
+        num_threads=2,
+    )
+
+    return recognizer
+
+
 def _get_english_model(repo_id: str) -> sherpa_onnx.OfflineRecognizer:
     assert (
         repo_id
@@ -564,6 +617,8 @@ chinese_models = {
 
 english_models = {
     "whisper-tiny.en": _get_whisper_model,
+    "moonshine-tiny": _get_moonshine_model,
+    "moonshine-base": _get_moonshine_model,
     "whisper-base.en": _get_whisper_model,
     "whisper-small.en": _get_whisper_model,
     "whisper-distil-small.en": _get_whisper_model,

requirements.txt

@@ -1,4 +1,4 @@
 
-#sherpa-onnx>=1.9.21
+sherpa-onnx>=1.10.30
 ffmpeg-python
-https://huggingface.co/csukuangfj/sherpa-onnx-wheels/resolve/main/cpu/1.10.28/sherpa_onnx-1.10.28-cp38-cp38-manylinux_2_17_x86_64.manylinux2014_x86_64.whl
+#https://huggingface.co/csukuangfj/sherpa-onnx-wheels/resolve/main/cpu/1.10.28/sherpa_onnx-1.10.28-cp38-cp38-manylinux_2_17_x86_64.manylinux2014_x86_64.whl