Update app.py

app.py

@@ -5,11 +5,13 @@ import torch
 
 import bitsandbytes as bnb
 
-# 設置模型 ID 和加載 Hugging Face API token
 model_id = "dingckc/FineLlama-3.1-8B"
-tokenizer = AutoTokenizer.from_pretrained(model_id, token=os.getenv('ACCESS_KEY'))
-model = AutoModelForCausalLM.from_pretrained(model_id, load_in_8bit=True, token=os.getenv('ACCESS_KEY'))
-model = model.to("cuda" if torch.cuda.is_available() else "cpu")
+tokenizer = AutoTokenizer.from_pretrained(model_id, token=os.getenv("ACCESS_KEY"))
+model = AutoModelForCausalLM.from_pretrained(
+    model_id,
+    device_map="auto",  # 自動映射到可用設備（CPU）
+    torch_dtype=torch.float16  # 減少內存占用
+)
 
 # 定義推理函數
 def evaluate_essay(title, essay):