How to perform real quantization model inference with SmoothQuant?

[wanto-rest]

I want to know how to perform inference. I have made several attempts, including using the Transformers framework, but there are always issues.
here is my code. i always meet problem like 'Int8OPTDecoder' object has no attribute 'dropout'...if i change the version of transformers ，missing attribute will also change.
import torch
from transformers import AutoTokenizer
from smoothquant.opt import Int8OPTForCausalLM
import os
os.environ["CUDA_VISIBLE_DEVICES"] = "0"

def load_pretrained_int8_model():
# 官方提供的INT8模型
model_name = "/data/zyh/models/mit-han-lab/opt-125m-smoothquant"
tokenizer = AutoTokenizer.from_pretrained("/data/zyh/models/opt-125m")

# 直接加载INT8模型
model = Int8OPTForCausalLM.from_pretrained(
    model_name,
    torch_dtype=torch.float16,
    device_map="auto"
)

return model, tokenizer

def int8_inference_demo():
model, tokenizer = load_pretrained_int8_model()

# 推理
prompt = "The future of artificial intelligence is"
inputs = tokenizer(prompt, return_tensors="pt")
input_ids = inputs.input_ids.cuda()

with torch.no_grad():
    outputs = model.generate(
        input_ids,
        max_new_tokens=50,
        temperature=0.7,
        do_sample=True,
        pad_token_id=tokenizer.eos_token_id
    )

generated_text = tokenizer.decode(outputs[0], skip_special_tokens=True)
print(f"Generated: {generated_text}")

return generated_text

if name == "main":
int8_inference_demo()