add German

examples.py

@@ -219,4 +219,18 @@ examples = [
         4,
         "./test_wavs/arabic/c.wav",
     ],
+    [
+        "German",
+        "csukuangfj/wav2vec2.0-torchaudio",
+        "greedy_search",
+        4,
+        "./test_wavs/german/20120315-0900-PLENARY-14-de_20120315.wav",
+    ],
+    [
+        "German",
+        "csukuangfj/wav2vec2.0-torchaudio",
+        "greedy_search",
+        4,
+        "./test_wavs/german/20170517-0900-PLENARY-16-de_20170517.wav",
+    ],
 ]

model.py

@@ -55,6 +55,10 @@ def get_pretrained_model(
         return arabic_models[repo_id](
             repo_id, decoding_method=decoding_method, num_active_paths=num_active_paths
         )
+    elif repo_id in german_models:
+        return german_models[repo_id](
+            repo_id, decoding_method=decoding_method, num_active_paths=num_active_paths
+        )
     else:
         raise ValueError(f"Unsupported repo_id: {repo_id}")
 
@@ -483,6 +487,41 @@ def _get_arabic_pre_trained_model(
     return recognizer
 
 
+@lru_cache(maxsize=10)
+def _get_german_pre_trained_model(
+    repo_id: str,
+    decoding_method: str,
+    num_active_paths: int,
+):
+    assert repo_id in [
+        "csukuangfj/wav2vec2.0-torchaudio",
+    ], repo_id
+
+    nn_model = _get_nn_model_filename(
+        repo_id=repo_id,
+        filename="voxpopuli_asr_base_10k_de.pt",
+        subfolder="./",
+    )
+
+    tokens = _get_token_filename(
+        repo_id=repo_id,
+        filename="tokens-de.txt",
+        subfolder="./",
+    )
+
+    config = sherpa.OfflineRecognizerConfig(
+        nn_model=nn_model,
+        tokens=tokens,
+        use_gpu=False,
+        decoding_method=decoding_method,
+        num_active_paths=num_active_paths,
+    )
+
+    recognizer = sherpa.OfflineRecognizer(config)
+
+    return recognizer
+
+
 chinese_models = {
     "luomingshuang/icefall_asr_wenetspeech_pruned_transducer_stateless2": _get_wenetspeech_pre_trained_model,  # noqa
     "yuekai/icefall-asr-aishell2-pruned-transducer-stateless5-A-2022-07-12": _get_aishell2_pretrained_model,  # noqa
@@ -514,12 +553,17 @@ arabic_models = {
     "AmirHussein/icefall-asr-mgb2-conformer_ctc-2022-27-06": _get_arabic_pre_trained_model,  # noqa
 }
 
+german_models = {
+    "csukuangfj/wav2vec2.0-torchaudio": _get_german_pre_trained_model,
+}
+
 all_models = {
     **chinese_models,
     **english_models,
     **chinese_english_mixed_models,
     **tibetan_models,
     **arabic_models,
+    **german_models,
 }
 
 language_to_models = {
@@ -528,4 +572,5 @@ language_to_models = {
     "Chinese+English": list(chinese_english_mixed_models.keys()),
     "Tibetan": list(tibetan_models.keys()),
     "Arabic": list(arabic_models.keys()),
+    "German": list(german_models.keys()),
 }