update

app.py

@@ -22,6 +22,8 @@ MAX_INPUT_TOKEN_LENGTH = int(os.getenv("MAX_INPUT_TOKEN_LENGTH", "32768"))
 if torch.cuda.is_available():
     model_id = "vericava/llm-jp-3-1.8b-instruct-lora-vericava7-llama"
     base_model_id = "llm-jp/llm-jp-3-1.8b-instruct"
+    tokenizer = AutoTokenizer.from_pretrained(base_model_id)
+    model = AutoModelForCausalLM.from_pretrained(model_id)
     my_pipeline=pipeline(
         model=model_id,
         tokenizer=base_model_id,