Python 变形金刚：获取密钥错误：'；句子嵌入'；关于model.fit

我得到了这个错误

KeyError:“句子嵌入”

在线：

rep = self.model(sentence_features[0])['sentence_embedding']

设置变压器的模型拟合功能

代码的其他部分：

train_data = ParallelSentencesDataset(student_model=student_model, teacher_model=teacher_model)
train_data.load_data('parallel5500TabTrain.tsv')
train_dataloader = DataLoader(train_data, shuffle=True, batch_size=train_batch_size)
train_loss = losses.MSELoss(model=student_model)

###### Load test sets ######
f = open("parallel5500TabTest.txt","r")
pairs = [line.strip().split("\t") for line in  f]
f.close()

src = []
trg = []
for pair in pairs:
    src.append(pair[0])
    trg.append(pair[1])

evaluators = []
test_reader = ParallelSentencesDataset(student_model=model, teacher_model=teacher_model)
test_reader.load_data('parallel5500TabTest.tsv')
test_dataloader = DataLoader(test_reader, shuffle=False, batch_size=train_batch_size)
test_mse = evaluation.TranslationEvaluator(src, trg)
evaluators.append(test_mse)

###### Train model ######

output_path = "model-" + _datetime.date.today().strftime("%Y-%m-%d")
model.fit(train_objectives=[(train_dataloader, train_loss)],
         evaluator=evaluators,
         epochs=20,
         evaluation_steps=1000,
         warmup_steps=1000,
         output_path=output_path 
          )

---

有人知道这可能是什么吗？

但是单词“句子嵌入”在来自句子变压器包的model.fit函数中。我看到其他一些代码也使用了它，它们在调用model.fit（）之前没有创建任何“句子嵌入”变量或类似的东西错误在丢失文件中。无论我选择了什么样的损失，他们都给出了相同的错误