Model Information

[phanphong13]

I see your models are built on the llama architecture, so which version of llama do you use and do you fine-tune on pre-trained models?

[yukiarimo]

V1-V3 = LLaMA 2 7B Instruct
V4 = LLaMA 3.1 8B Base
Himitsu V1 = LLaMA 3.2 1B Base (upcoming)

It's all in the documentation. Please read it again. https://github.com/yukiarimo/yuna-ai?tab=readme-ov-file#dataset

[phanphong13]

Oh, I am sorry . UmI want to ask a little about the infrastructure to fine-tune and deploy your product. Can you tell me more? Vào Th 5, 7 thg 11, 2024 vào lúc 15:15 Yuki Arimo < ***@***.***> đã viết:

…

V1-V3 = LLaMA 2 7B Instruct V4 = LLaMA 3.1 8B Base Himitsu V1 = LLaMA 3.2 1B Base (upcoming) It's all in the documentation. Please read it again. https://github.com/yukiarimo/yuna-ai?tab=readme-ov-file#dataset — Reply to this email directly, view it on GitHub <#117 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ATOW3QEZW6KYXG4INBNIV2DZ7MORXAVCNFSM6AAAAABRKQOGK6VHI2DSMVQWIX3LMV43URDJONRXK43TNFXW4Q3PNVWWK3TUHMYTCMJXGQ3TONY> . You are receiving this because you authored the thread.Message ID: ***@***.***>

[yukiarimo]

1xA100 was used for fine-tuning. For inference, almost any consumer-grade NVIDIA GPU should be fine. Personally, I’m using a MacBook Pro M1 with 16GB of RAM.

If you want me to train a custom model for you or are planning to deploy it, please refer to the Yuna Ai Marketplace: https://www.patreon.com/c/yukiarimo/membership

Note: to use/deploy Yuna commercially, you must be a member of Yuna Premium and have a commercial license purchased from the marketplace