add tibetan

examples.py

@@ -176,4 +176,25 @@ examples = [
         4,
         "./test_wavs/tal_csasr/210_36476_210_8341_1_1533271973_7057520_145.wav",
     ],
+    [
+        "Tibetan",
+        "syzym/icefall-asr-xbmu-amdo31-pruned-transducer-stateless5-2022-11-29",
+        "greedy_search",
+        4,
+        "./test_wavs/tibetan/a_0_cacm-A70_31116.wav",
+    ],
+    [
+        "Tibetan",
+        "syzym/icefall-asr-xbmu-amdo31-pruned-transducer-stateless5-2022-11-29",
+        "greedy_search",
+        4,
+        "./test_wavs/tibetan/a_0_cacm-A70_31117.wav",
+    ],
+    [
+        "Tibetan",
+        "syzym/icefall-asr-xbmu-amdo31-pruned-transducer-stateless5-2022-11-29",
+        "greedy_search",
+        4,
+        "./test_wavs/tibetan/a_0_cacm-A70_31118.wav",
+    ],
 ]

model.py

@@ -233,6 +233,27 @@ def _get_aidatatang_200zh_pretrained_mode(repo_id: str):
     )
 
 
+@lru_cache(maxsize=10)
+def _get_tibetan_pre_trained_model(repo_id: str):
+    assert repo_id in [
+        "syzym/icefall-asr-xbmu-amdo31-pruned-transducer-stateless5-2022-11-29",
+    ], repo_id
+
+    nn_model_filename = _get_nn_model_filename(
+        repo_id=repo_id,
+        filename="cpu_jit.pt",
+    )
+    bpe_model_filename = _get_bpe_model_filename(repo_id=repo_id)
+
+    return OfflineAsr(
+        nn_model_filename=nn_model_filename,
+        bpe_model_filename=bpe_model_filename,
+        token_filename=None,
+        sample_rate=sample_rate,
+        device="cpu",
+    )
+
+
 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
@@ -252,14 +273,20 @@ chinese_english_mixed_models = {
     "luomingshuang/icefall_asr_tal-csasr_pruned_transducer_stateless5": _get_tal_csasr_pre_trained_model,  # noqa
 }
 
+tibetan_models = {
+    "syzym/icefall-asr-xbmu-amdo31-pruned-transducer-stateless5-2022-11-29": _get_tibetan_pre_trained_model,  # noqa
+}
+
 all_models = {
     **chinese_models,
     **english_models,
     **chinese_english_mixed_models,
+    **tibetan_models,
 }
 
 language_to_models = {
     "Chinese": list(chinese_models.keys()),
     "English": list(english_models.keys()),
     "Chinese+English": list(chinese_english_mixed_models.keys()),
+    "Tibetan": list(tibetan_models.keys()),
 }

test_wavs/tibetan/trans.txt

@@ -0,0 +1,3 @@
+a_0_cacm-A70_31116.wav ལོ བཅུ ཙམ མ འདང བའི དུས སྐབས ནང
+a_0_cacm-A70_31117.wav དྲག པོའི ངོ ལོག ཟིང འཁྲུག སྒྲིག འཛུགས དང ངན བཀོད བྱས ཡོད
+a_0_cacm-A70_31118.wav གནས བབ འདིའི རིགས གང མགྱོགས འགྱུར བ གཏོང དགོས