QNN support for other models (e.g., Gemma2, phi-2)

[Seoyoung-Ko]

I am unable to experiment with MLLM-NPU w/ Qwen because my current device has 12GB memory.
In the MLLM-NPU paper, I saw that the paper evaluated with Gemma2, phi-2, etc. models. So I was wondering if you could share the code for these models. Also, I want to know why evaluating with Qwen-1.5 1.8B needs 16GB device because the model parameter size is not that large. I hope to try MLLM on my own device to see how it performs.