Cross encoder training example (train_samples for dataloader but dev_samples for evaluator, why?

Question

I am trying to use a pre-trained model (same as in the example below) and add training data to it but when I look at the evaluator part, the dev_samples is parsed to the function.

Why is it not the train_samples?

    # We wrap train_samples (which is a List[InputExample]) into a pytorch DataLoader
    train_dataloader = DataLoader(train_samples, shuffle=True, batch_size=train_batch_size)
    
    
    # We add an evaluator, which evaluates the performance during training
    evaluator = CECorrelationEvaluator.from_input_examples(dev_samples, name="sts-dev")

Github Code

    train_samples = []
    dev_samples = []
    test_samples = []
    with gzip.open(sts_dataset_path, "rt", encoding="utf8") as fIn:
        reader = csv.DictReader(fIn, delimiter="\t", quoting=csv.QUOTE_NONE)
        for row in reader:
            score = float(row["score"]) / 5.0  # Normalize score to range 0 ... 1
    
            if row["split"] == "dev":
                dev_samples.append(InputExample(texts=[row["sentence1"], row["sentence2"]], label=score))
            elif row["split"] == "test":
                test_samples.append(InputExample(texts=[row["sentence1"], row["sentence2"]], label=score))
            else:
                # As we want to get symmetric scores, i.e. CrossEncoder(A,B) = CrossEncoder(B,A), we pass both combinations to the train set
                train_samples.append(InputExample(texts=[row["sentence1"], row["sentence2"]], label=score))
                train_samples.append(InputExample(texts=[row["sentence2"], row["sentence1"]], label=score))
    
    
    # We wrap train_samples (which is a List[InputExample]) into a pytorch DataLoader
    train_dataloader = DataLoader(train_samples, shuffle=True, batch_size=train_batch_size)
    
    
    # We add an evaluator, which evaluates the performance during training
    evaluator = CECorrelationEvaluator.from_input_examples(dev_samples, name="sts-dev")
    
    
    # Configure the training
    warmup_steps = math.ceil(len(train_dataloader) * num_epochs * 0.1)  # 10% of train data for warm-up
    logger.info("Warmup-steps: {}".format(warmup_steps))
    
    
    # Train the model
    model.fit(
        train_dataloader=train_dataloader,
        evaluator=evaluator,
        epochs=num_epochs,
        warmup_steps=warmup_steps,
        output_path=model_save_path,
    )
    
    
    ##### Load model and eval on test set
    model = CrossEncoder(model_save_path)
    
    evaluator = CECorrelationEvaluator.from_input_examples(test_samples, name="sts-test")
    evaluator(model)