Add support for Llama 3

[lajlev]

Could be cool if it could run on iOS

[yayoimizuha]

I’m using Llama-3 model at TestFlight app, Something is wrong?

[lajlev]

How do I add it?

From: Tomokazu Katayama @.> Date: Tuesday, 23 April 2024 at 08.04 To: guinmoon/LLMFarm @.> Cc: Michael Lajlev @.>, Author @.> Subject: Re: [guinmoon/LLMFarm] Add support for Llama 3 (Issue #57)

IMG_0044.png (view on web)https://github.com/guinmoon/LLMFarm/assets/32425706/16d4ce6e-61d0-4928-a00f-9115c7550f49 I’m using Llama-3 model at TestFlight app, Something is wrong?

— Reply to this email directly, view it on GitHubhttps://github.com/guinmoon/LLMFarm/issues/57#issuecomment-2071474268, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AABAVTESU3NTTT2QCC4XEZLY6X2WLAVCNFSM6AAAAABGSOERCOVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDANZRGQ3TIMRWHA. You are receiving this because you authored the thread.Message ID: @.***>

[yayoimizuha]

Config:

Used prompt:

<|begin_of_text|><|start_header_id|>system<|end_header_id|>
A chat between a curious user and an artificial intelligence assistant. The assistant gives helpful, detailed, and polite answers to the user's questions. <|eot_id|><|start_header_id|>user<|end_header_id|>
{prompt}<|eot_id|><|start_header_id|>assistant<|end_header_id|>

Model file from: https://huggingface.co/QuantFactory/Meta-Llama-3-8B-Instruct-GGUF/blob/main/Meta-Llama-3-8B-Instruct.Q5_K_M.gguf

Using: iPad Pro ‘11 Gen 4(2022) (512GB model)

Sorry for poorly English because I’m Japanese.

[yayoimizuha]

P.S. Used App ver.:

[amostamu]

maybe llama 3 is not good at Japanese , test llama 3 70b with your prompt. Output is English with a lot of !!!!!

[lajlev]

Thanks for your responses. I'll try it out 🚀