Change model to large

tools/vqgan/calculate_hubert_features.py

@@ -35,7 +35,7 @@ logger.add(sys.stderr, format=logger_format)
 
 @lru_cache(maxsize=1)
 def get_hubert_model():
-    model = HubertModel.from_pretrained("TencentGameMate/chinese-hubert-base")
+    model = HubertModel.from_pretrained("TencentGameMate/chinese-hubert-large")
     model = model.to(torch.device("cuda" if torch.cuda.is_available() else "cpu"))
     model = model.half()
     model.eval()