Why does phi3 implementation in vLLM inherit from llama?

[power-puff-gg]

This might be a naive question but does anyone know why does phi3 implementation in vLLM inherit from llama?

Inference-only Phi3 model code inherit from Llama.py

vllm/model_executor/models/phi3.py

Why does vLLM not have an implementation of Phi3Model?
I wanted to extend the phi3 class with custom code but unable to get it working.

[robertshaw]

@power-puff-gg

LlamaForCausalLM

The models are very similar architecture. If you need to extend Phi, I would suggest extending the Llama definition