How to fine-tune the llama-adapter-v2 for llama-2 7b models.

[Harsh-raj]

I am trying to use llama-adapter-v2 for llama2-7b model but there is got to be differences in the fine tuning process due to changes in llama model to llama 2 models. how to do the fine tuning with llama 2 instead of llama model

[gaopengpjlab]

Please try the upgraded llama-adapter-v2 model SPHINX in the following link:

https://github.com/Alpha-VLLM/LLaMA2-Accessory/tree/main/SPHINX

SPHINX using 1440/288 visual tokens and LLaMA2-13B while LLaMa-Adapter-V2 only use 10 visual tokens and LLaMa-7B.