Fallback to CPU if loading model: vk::PhysicalDevice::createDevice: ErrorExtensionNotPresent.

[gaykawadpk]

I have built llama.cpp using -DGGML_VULKAN=ON
I am running example on android emulator.
./llama-cli -m /storage/emulated/0/Android/data/com.example.llama/files/tinyllama-1.1-f16.gguf

But if fails -

load_tensors: loading model tensors, this can take a while... (mmap = true)

llama_model_load: error loading model: vk::PhysicalDevice::createDevice: ErrorExtensionNotPresent

llama_model_load_from_file_impl: failed to load model

How do I make it fallback to CPU if extension is not supported.
Or how do I support this?