Update src/speech_to_text.py

src/speech_to_text.py

@@ -32,6 +32,8 @@ def transcribe(fp:str, target_lang:str) -> str:
     processor = AutoProcessor.from_pretrained(model_id, target_lang=target_lang)
     model = Wav2Vec2ForCTC.from_pretrained(model_id, target_lang=target_lang, ignore_mismatched_sizes=True)
 
+    processor.tokenizer.set_target_lang("mos")
+
     # Process the audio
     signal, sampling_rate =  librosa.load(fp, sr=16000)
     inputs = processor(signal, sampling_rate=16_000, return_tensors="pt")