search

clevaway / J.A.R.V.I.S

Jarvis model: A Fine-tune of llama2, works 100% offline with Ollama
MIT License
68 stars 12 forks source link

Try a different LLM #10

Open SakthiMahendran opened 3 months ago

SakthiMahendran commented 3 months ago

As people mentioned before LLaMa is very slow due it's huge size... So why don't we try different models for example "phi3" from microsoft.... in the context of day to day uses llama and phi3 won't have big difference.... but phi3 is lightwieght and is much more quick than llama. It has 3.8B parameters.... If we wan't to make to useable then LLaMa won't be good choice...

try "ollama run phi3"... You will get an idea

FotieMConstant commented 3 months ago

Hey there @SakthiMahendran thanks for the input on this. I really appreciate you suggestion. any thoughts on fine-tuning phi3? perhaps we can start from there. I have never fine-tuned it before and I suppose the fine-tuning processing is different from llama.

let me know. Thanks!