St.cache throws unexpected mutation warning

I have a streamlit app where I want to first load a sklearn model from s3.

import streamlit as st
import pandas as pd 

def load_model():
    DATA_URL = "path/to/model"
    rf_mod = pd.read_pickle(DATA_URL)
    return rf_mod

load_state = st.text("loading ...")
rf_mod = load_model()

options = st.selectbox("select a model", ["Random Forest", "XGBoost"])

f"you select {options}"

However, when I toggle the select box the error was raised CachedObjectMutationWarning: Return value of load_model() was mutated between runs. I didn’t try to modify the loaded model. Anyone has idea how the mutation happened?