How to use gpu to accelerate inference?

[跃跃欲试]

How to use gpu to accelerate inference? I deployed it exactly according to howtouse on GitHub, but it was still very slow and did not use gpu for inference

[跃跃欲试]

import torch
from transformers import AutoTokenizer, AutoModelForCausalLM

tokenizer = AutoTokenizer.from_pretrained("Salesforce/codegen25-7b-mono", trust_remote_code=True)
model = AutoModelForCausalLM.from_pretrained("Salesforce/codegen25-7b-mono")
inputs = tokenizer("# this function prints hello world", return_tensors="pt")
sample = model.generate(**inputs, max_length=128)
print(tokenizer.decode(sample[0]))