代码如下：
from transformers import *
import torch
import logging
logging.basicConfig(level=logging.INFO)
bert_model_path = "../pretrain_model/bert_base_cased"
tokenizer = BertTokenizer.from_pretrained(bert_model_path)
model = BertForSequenceClassification.from_pretrained(bert_model_path)
classes = ["not paraphrase", "is paraphrase"]
sequence_0 = "The company HuggingFace is based in New York City"
sequence_1 = "Apples are especially bad for your health"
sequence_2 = "HuggingFace's headquarters are situated in Manhattan"
paraphrase = tokenizer.encode_plus(sequence_0, sequence_2, return_tensors="pt")
not_paraphrase = tokenizer.encode_plus(sequence_0, sequence_1, return_tensors="pt")
paraphrase_classification_logits = model(**paraphrase)[0]
not_paraphrase_classification_logits = model(**not_paraphrase)[0]
paraphrase_results = torch.softmax(paraphrase_classification_logits, dim=1).tolist()[0]
not_paraphrase_results = torch.softmax(not_paraphrase_classification_logits, dim=1).tolist()[0]
print("Should be paraphrase")
for i in range(len(classes)):
print(f"{classes[i]}: {round(paraphrase_results[i] * 100)}%")
print("\nShould not be paraphrase")
for i in range(len(classes)):
print(f"{classes[i]}: {round(not_paraphrase_results[i] * 100)}%")
预期输出格式： Should be paraphrase not paraphrase: 10% is paraphrase: 90% Should not be paraphrase not paraphrase: 94% is paraphrase: 6%在服务器上输出：
weight.t() size: torch.Size([768, 3072]) input size: torch.Size([1, 21, 3072]) weight.t() size: torch.Size([3072, 768]) input size: torch.Size([1, 21, 768]) weight.t() size: torch.Size([768, 768]) input size: torch.Size([1, 21, 768]) weight.t() size: torch.Size([768, 768]) input size: torch.Size([1, 21, 768]) weight.t() size: torch.Size([768, 768]) input size: torch.Size([1, 21, 768]) weight.t() size: torch.Size([768, 768]) input size: torch.Size([1, 21, 768]) weight.t() size: torch.Size([768, 3072]) input size: torch.Size([1, 21, 3072]) weight.t() size: torch.Size([3072, 768]) #还有很多行像上面一样格式数据输出 Should be paraphrase not paraphrase: 10% is paraphrase: 90% Should not be paraphrase not paraphrase: 94% is paraphrase: 6%求问中间输出是怎么回事，我并没有主动打印上面的信息啊，我换了好几个模型都是这样子，请问有人碰到过这样的情况么。
我去看transformer的loging信息，但是并没有发现有weight.t()这样的信息打印。去github上看issue也没发现有人碰到这样的问题。
google直接搜不到
###救救孩子吧，开学不出成果会被延毕啊啊啊啊啊。
全部评论
(4) 回帖