Why "vision_model.post_layernorm" is skipped in llama.cpp/examples /llava/convert_image_encoder_to_gguf.py

[DanialTaheri]

I am looking at the https://github.com/ggerganov/llama.cpp/blob/master/examples/llava/convert_image_encoder_to_gguf.py#L19 and I see that the "post_layernorm" weight and biass is skipped in the gguf conversion of the encoder, and subsequently in the inference graph for Llava. What is the reason for that?