是否支持lora微调visual model

[HiXxxSss]

您好，当我希望使用lora微调visual model时，发生报错。具体来说，我在Qwen2_5_VLForConditionalGeneration定义下面的方法：

def wrap_visual_lora(self, r=128, lora_alpha=256, lora_dropout=0.05):
    lora_config = LoraConfig(
        r=r,
        target_modules=['attn.qkv', 'attn.proj', 'mlp.gate_proj', 'mlp.up_proj', 'mlp.down_proj'],
        lora_alpha=lora_alpha,
        lora_dropout=lora_dropout,
    )
    self.visual = get_peft_model(self.visual, lora_config)
    self.visual.enable_input_require_grads()
    self.visual.print_trainable_parameters()

但是报错如下。似乎是因为找不到visual模型的get_input_embeddings方法。有没有其他lora微调visual部分的办法呢？

Traceback (most recent call last):
  File "/DATA/workshop/personal/Qwen2_5_VL/qwen2_5_vl/train/train.py", line 379, in <module>
    main()
  File "/DATA/workshop/personal/Qwen2_5_VL/qwen2_5_vl/train/train.py", line 324, in main
    model.wrap_visual_lora(r=model_args.use_visual_lora, lora_alpha=2 * model_args.use_visual_lora)
  File "/DATA/workshop/personal/Qwen2_5_VL/qwen2_5_vl/model/modeling_qwen2_5_vlForClassification.py", line 241, in wrap_visual_lora
    self.visual = get_peft_model(self.visual, lora_config)
  File "/usr/local/lib/python3.10/dist-packages/peft/mapping.py", line 95, in get_peft_model
    return PeftModel(model, peft_config, adapter_name=adapter_name)
  File "/usr/local/lib/python3.10/dist-packages/peft/peft_model.py", line 120, in __init__
    model = self._prepare_model_for_gradient_checkpointing(model)
  File "/usr/local/lib/python3.10/dist-packages/peft/peft_model.py", line 320, in _prepare_model_for_gradient_checkpointing
    model.enable_input_require_grads()
  File "/usr/local/lib/python3.10/dist-packages/transformers/modeling_utils.py", line 1883, in enable_input_require_grads
    self._require_grads_hook = self.get_input_embeddings().register_forward_hook(make_inputs_require_grads)
  File "/usr/local/lib/python3.10/dist-packages/transformers/modeling_utils.py", line 1902, in get_input_embeddings
    raise NotImplementedError
NotImplementedError