fix cache

app.py

@@ -76,7 +76,7 @@ def load_lora(lora_path, base_model="decapoda-research/llama-7b-hf"):
 
 
 base_model = "decapoda-research/llama-13b-hf"
-tokenizer = LlamaTokenizer.from_pretrained(base_model, cache_dir="data/hf")
+tokenizer = LlamaTokenizer.from_pretrained(base_model)
 # question = "如果今天是星期五, 那么后天是星期几?"
 model = load_lora(lora_path="facat/alpaca-lora-cn-13b", base_model=base_model)