Update main.py

main.py

@@ -15,7 +15,7 @@ app = FastAPI()
 
 # Load model once on startup
 MODEL_NAME = "facebook/nllb-200-1.3B"  # 3B version
-tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME)
+tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME,use_fast=False)
 model = AutoModelForSeq2SeqLM.from_pretrained(MODEL_NAME, torch_dtype=torch.float16).to("cuda" if torch.cuda.is_available() else "cpu")
 
 class TranslationRequest(BaseModel):