Support for llama-2 70B

[qizzzh]

Would it work natively or we need to train new adapters?

[gaopengpjlab]

Our new repo support fullfinetune / PEFT / Quantized PEFT of LLaMa2-70B. Please check the following repo:

https://github.com/Alpha-VLLM/LLaMA2-Accessory

[qizzzh]

Nice. Do you have a fine-tuned llama-2 70B multi-modal model?

[gaopengpjlab]

Please check the model zoo of llama2-accessory. We released llama2-70B finetuned on sharegpt.

[qizzzh]

Sorry where is it? I didn’t find it in the repo.

[gaopengpjlab]

Please check this website:

https://github.com/Alpha-VLLM/LLaMA2-Accessory/blob/main/docs/finetune.md#multi-turn-instruction-tuning-of-llama2-7b-on-sharegpt

[LutaoChu]

Great work! Has the LLaMA Adapter V2 model based on LLaMA2 been released? Or release plan?

[gaopengpjlab]

@LutaoChu Please check the new repo. https://llama2-accessory.readthedocs.io/en/latest/finetune/sg_peft.html