DPOTrainer and sequence length

What happens when using the DPO Trainer and there’s a data point with sequence length greater than the model’s sequence length? Does it get cut off from the beginning until it fits? Does it throw an error?