How To FineTune Llama3

Follow Along: console.brev.d...
Join Our Discord: / discord
Request Access To the Model: huggingface.co...
In this guide, we fine tune the popular open sourced model, Llama3-8B using the powerful finetuning method DPO (Direct Preference Optimization). You can do this all from your laptop! Try it out and let us know what you think
Please leave any future guides you would like made below!

[@codetour]

This is goooood good stuff

[@ashokchhetri]

Why is your viewer pov lower? I think bit higher would be fine. Good going with your content. Keep them coming. ❤🎉

[@lilflameeeee]

keep going bro thank you for this

[@brev-dev]

you’re welcome - baxate

[@inusahseidu3620]

Here from TikTok 🫡

[@sachinperuri2249]

The chat template is not available for download anymore. Could you check please? Thanks.

[@brev-dev]

hey there, looks like the user who created that formatter changed the name! We will update the notebook. To get it to work simply replace the word “chat” with “instruct” in the url, so
raw.githubusercontent.com/chujiezheng/chat_templates/main/chat_templates/llama-3-instruct.jinja

[@sachinperuri2249]

@@brev-dev Thanks!

[@bamboom6172]

Lebanon

[@RAGEEcs]

Lebanon Gang

[@RabeeQasem]

thank you

[@McLovinit5338]

understood 0% but we gotta start somewhere ig

[@kennethnavarro661]

Does Brev also provide cloud storage similar to an aws s3 bucket ?

[@Baxate]

currently working on this feature! Be on the lookout for it soon

[@85bitsdeveloper]

Can you try it using ORPO instead DPO?

[@ahmedbader5721]

Keep up the good work man 🫡

[@RaulCorrea-xi6je]

you have to pay to deploy it?

[@surajsukumar2721]

171k

[@surajsukumar2721]

157k

[@RaulCorrea-xi6je]

you have to pay to deploy?

[@brev-dev]

yes, because this is on an A100 GPU, you will have to purchase brev credits. However, this particular instance is a few dollars an hour so in total this guide should be about 4$