Template-KleinBase4B-ContentRef权重加载失败

运行examples/flux2/model_inference_low_vram/Template-KleinBase4B-ContentRef.py，程序报错如下（部分内容

）：
Exception has occurred: RuntimeError
Error(s) in loading state_dict for FLUX2Image2LoRAModel:
Missing key(s) in state_dict: "image_encoder.vision_model.embeddings.patch_embedding.weight", "image_encoder.vision_model.embeddings.patch_embedding.bias", "image_encoder.vision_model.embeddings.position_embedding.weight", "image_encoder.vision_model.encoder.layers.0.layer_norm1.weight", "image_encoder.vision_model.encoder.layers.0.layer_norm1.bias", "image_encoder.vision_model.encoder.layers.0.self_attn.k_proj.weight", "image_encoder.vision_model.encoder.layers.0.self_attn.k_proj.bias", "image_encoder.vision_model.encoder.layers.0.self_attn.v_proj.weight", "image_encoder.vision_model.encoder.layers.0.self_attn.v_proj.bias", "image_encoder.vision_model.encoder.layers.0.self_attn.q_proj.weight", "image_encoder.vision_model.encoder.layers.0.self_attn.q_proj.bias", "image_encoder.vision_model.encoder.layers.0.self_attn.out_proj.weight", "image_encoder.vision_model.encoder.layers.0.self_attn.out_proj.bias", "image_encoder.vision_model.encoder.layers.0.layer_norm2.weight", "image_encoder.vision_model.encoder.layers.0.layer_norm2.bias", "image_encoder.vision_model.encoder.layers.0.mlp.fc1.weight", "image_encoder.vision_model.encoder.layers.0.mlp.fc1.bias", "image_encoder.vision_model.encoder.layers.0.mlp.fc2.weight", "image_encoder.vision_model.encoder.layers.0.mlp.fc2.bias", "image_encoder.vision_model.encoder.layers.1.layer_norm1.weight", "image_encoder.vision_model.encoder.layers.1.layer_norm1.bias", "image_encoder.vision_model.encoder.layers.1.self_attn.k_proj.weight", "image_encoder.vision_model.encoder.layers.1.self_attn.k_proj.bias", "image_encoder.vision_model.encoder.layers.1.self_attn.v_proj.weight", "image_encoder.vision_model.encoder.layers.1.self_attn.v_proj.bias", "image_encoder.vision_model.encoder.layers.1.self_attn.q_proj.weight", "image_encoder.vision_model.encoder.layers.1.self_attn.q_proj.bias", "image_encoder.vision_model.encoder.layers.1.self_attn.out_proj.weight", "image_encoder.vision_model.encoder.layers.1.self_attn.out_proj.bias", "image_encoder.vision_model.encoder.layers.1.layer_norm2.weight", "image_encoder.vision_model.encoder.layers.1.layer_norm2.bias", "image_encoder.vision_model.encoder.layers.1.mlp.fc1.weight", "image_encoder.vision_model.encoder.layers.1.mlp.fc1.bias", "image_encoder.vision_model.encoder.layers.1.mlp.fc2.weight", "image_encoder.vision_model.encoder.layers.1.mlp.fc2.bias", "image_encoder.vision_model.encoder.layers.2.layer_norm1.weight", "image_encoder.vision_model.encoder.layers.2.layer_norm1.bias", "image_encoder.vision_model.encoder.layers.2.self_attn.k_proj.weight", "image_encoder.vision_model.encoder.layers.2.self_attn.k_proj.bias", "image_encoder.vision_model.encoder.layers.2.self_attn.v_proj.weight", "image_encoder.vision_model.encoder.layers.2.self_attn.v_proj.bias",

Start by running examples/flux2/model_inference_low_vram/Template-KleinBase4B-ContentRef.py and inspect how weights are selected and loaded for FLUX2Image2LoRAModel. Compare the expected image_encoder keys with the loaded state_dict; done means the example loads the weights without the reported missing-key error.