Can Flash Attention patched LoRA be converted to ggml?

[Alasdair0]

If I patch a LLaMa model with Flash Attention, then create a LoRA with it, can I still convert that to ggml and infer on it with Llama.cpp?