Update app.py

app.py

@@ -14,8 +14,9 @@ tokenizer.pad_token = tokenizer.eos_token
 #Base model
 model = AutoModelForCausalLM.from_pretrained(
     BASE_MODEL,
-    torch_dtype=torch.float16 if DEVICE == "cuda" else torch.float32,
-    device_map="auto"
+    torch_dtype=torch.float32,
+    device_map={"": "cpu"},
+    low_cpu_mem_usage=True
 )