RuntimeError: The size of tensor a (812) must match the size of tensor b (6300) at non-singleton dimension 0

[zapqqqwe]

generated_ids = model.generate(**inputs, max_new_tokens=32,)为什么推理的时候报错，但是添加use_cache=False就可以不报错呢，，下面的报错

        outputs_cd = self(
            **model_inputs_cd,
            return_dict=True,
            output_attentions=output_attentions_wo_img,
            output_hidden_states=output_hidden_states_wo_img,
        )