Lora参数融合后保存模型导入报错

pipe = DiffusionPipeline.from_pretrained(base_model_id, torch_dtype=torch.float16).to("cuda")
pipe.load_lora_weights("lora.safetensors")

pipe.save_pretrained("model")

这样会带来错误：

ValueError: Cannot load <class 'diffusers.models.unets.unet_2d_condition.UNet2DConditionModel'> from hypersd/unet because the following keys are missing: 
 down_blocks.2.attentions.1.transformer_blocks.0.attn2.to_q.weight, down_blocks.3.resnets.1.conv1.weight, down_blocks.3.resnets.0.conv2.weight, up_blocks.3.attentions.0.transformer_blocks.0.attn1.to_k.weight, up_blocks.1.resnets.2.conv1.weight, up_blocks.2.attentions.1.transformer_blocks.0.ff.net.0.proj.bias, up_blocks.0.resnets.1.conv_shortcut.bias, up_blocks.3.attentions.2.transformer_blocks.

因为lora参数还没真正融合，要先

pipe.unload_lora_weights()

再保存，就保存lora参数和pipe融合的版本了。