change the loading args

translation.py

@@ -100,7 +100,7 @@ def sentenize_with_fillers(text, splitter, fix_double_space=True, ignore_errors=
 
 class Translator:
     def __init__(self):
-        self.model = AutoModelForSeq2SeqLM.from_pretrained(MODEL_URL, low_cpu_mem_usage=False, load_in_8bit=True)
+        self.model = AutoModelForSeq2SeqLM.from_pretrained(MODEL_URL, low_cpu_mem_usage=True)
         if torch.cuda.is_available():
             self.model.cuda()
         self.tokenizer = NllbTokenizer.from_pretrained(MODEL_URL)