storage

app.py

@@ -20,7 +20,7 @@ MAX_INPUT_TOKEN_LENGTH = 4096
 
 if torch.cuda.is_available():
     model_id = "codys12/MergeLlama-7b"
-    model = AutoModelForCausalLM.from_pretrained(model_id, torch_dtype=torch.float16, device_map=0,)
+    model = AutoModelForCausalLM.from_pretrained(model_id, torch_dtype=torch.float16, device_map=0, cache_dir="/Data")
     model.cuda()
     tokenizer = AutoTokenizer.from_pretrained(model_id)