怎么获取XLNet里某个单词的词向量呢？

import torch
from scipy.spatial.distance import cosine
from torch.utils.data import DataLoader, Dataset
from transformers import XLNetTokenizer, XLNetModel
from d2l import torch as d2l

tokenizer = XLNetTokenizer.from_pretrained("BaseXLNet")
XLNet = XLNetModel.from_pretrained("BaseXLNet")
sentens = "I will give you some hopes, do not sad and grieved"
tokens = tokenizer(sentens, return_tensors="pt")
with torch.no_grad():
    out = XLNet(**tokens)

怎么根据预训练模型，获取这句话里某个单词的词向量呢？