Update app.py

app.py

@@ -7,13 +7,14 @@ import json
 with open('ISO_codes.json', 'r') as file:
     iso_codes = json.load(file)
 
-languages = list(iso_codes.keys())
 
-model_id = "facebook/mms-1b-all"
+model_id = "TifinLab/mms-1b-berber"
+
 processor = AutoProcessor.from_pretrained(model_id)
 model = Wav2Vec2ForCTC.from_pretrained(model_id)
 
-def transcribe(audio_file_mic=None, audio_file_upload=None, language="English (eng)"):
+
+def transcribe(audio_file_mic=None, audio_file_upload=None):
     if audio_file_mic:
         audio_file = audio_file_mic
     elif audio_file_upload:
@@ -26,10 +27,10 @@ def transcribe(audio_file_mic=None, audio_file_upload=None, language="English (e
     if sample_rate != 16000:
         speech = librosa.resample(speech, orig_sr=sample_rate, target_sr=16000)
 
-    # Keep the same model in memory and simply switch out the language adapters by calling load_adapter() for the model and set_target_lang() for the tokenizer
-    language_code = iso_codes[language]
-    processor.tokenizer.set_target_lang(language_code)
-    model.load_adapter(language_code)
+  
+    processor.tokenizer.set_target_lang("ber")
+    model.load_adapter("ber")
+
 
     inputs = processor(speech, sampling_rate=16_000, return_tensors="pt")