该模型基于一个定制化的Transformer架构，可通过以下命令安装：

pip install git+https://github.com/lucadiliello/bleurt-pytorch.git

加载模型并进行预测的代码如下：

import torch
from bleurt_pytorch import BleurtConfig, BleurtForSequenceClassification, BleurtTokenizer

config = BleurtConfig.from_pretrained('lucadiliello/BLEURT-20')
model = BleurtForSequenceClassification.from_pretrained('lucadiliello/BLEURT-20')
tokenizer = BleurtTokenizer.from_pretrained('lucadiliello/BLEURT-20')

references = ["窗边有鸟儿在啁啾", "这是一个随机句子"]
candidates = ["窗边有鸟儿在啁啾", "这看起来像是个随机句子"]

model.eval()
with torch.no_grad():
    inputs = tokenizer(references, candidates, padding='longest', return_tensors='pt')
    res = model(**inputs).logits.flatten().tolist()
print(res)
# [0.9990496635437012, 0.7930182218551636]

有关PyTorch中BleurtConfig、BleurtForSequenceClassification和BleurtTokenizer的定义，请参阅该代码仓库。