update hubertbase and whisper to also convert into ipa for ranking

utils/load_model.py

@@ -69,6 +69,7 @@ def run_hubert_base(wav):
     text = base_proc.batch_decode(ids)[0]
     # Convert to phonemes (CMU-like string without stresses)
     phonemes = text_to_phoneme(text)
+    phonemes = cmu_to_ipa(phonemes)
     return phonemes.strip(), time.time() - start
 
 def run_whisper(wav):
@@ -86,6 +87,7 @@ def run_whisper(wav):
 
     text = whisper_proc.batch_decode(pred_ids, skip_special_tokens=True)[0]
     phonemes = text_to_phoneme(text)
+    phonemes = cmu_to_ipa(phonemes)
     return phonemes.strip(), time.time() - start