When trying to use IREE to convert to large language model, an error is reported

[DFXX34]

import torch
from transformers import AutoModelForCausalLM, AutoTokenizer
import torch._dynamo

import os
os.environ["TOKENIZERS_PARALLELISM"] = "false"
torch._dynamo.config.suppress_errors = True

model_name = "../Meta-Llama-3-8B"
device = torch.device("cpu")
model = AutoModelForCausalLM.from_pretrained(
model_name,
torch_dtype="auto",
device_map="auto",
offload_folder="offload_folder"
)
tokenizer = AutoTokenizer.from_pretrained(model_name)
model = model.half()
prompt = "请给我一个关于大型语言模型的简短介绍。"
inputs = tokenizer(prompt, return_tensors="pt").to(device)

opt_linear_module = torch.compile(model, backend="turbine_cpu")
print(opt_linear_module)
with torch.no_grad():
export = opt_linear_module(**inputs)

error：