Practical Fine-Tuning of LLMs

[@AI-Makerspace]

Google colab notebook: colab.research.google.com/drive/1Jw9jthx_S62MPwKH9lqb6xPRwec4OiI6?usp=sharing Event Slides: www.canva.com/design/DAF-v_5WxcU/s2SCPuVA7ikGR0VSJOG6Rw/view?DAF-v_5WxcU&

[@AI_by_AI_007]

Another Great One! --- Downloaded a set of the sessions for a road trip this past weekend -- was like I entered a time warp, and I almost wished the trip was longer so I could finish the talk I was jamming on! Thanks for sharing your gifts that are beyond price... may success find and reward you.

[@AI-Makerspace]

Haha road-tripping with AIM is legend status AI_by-AI! We love to hear it 🙏!!

[@sivi3883]

I watched an hr. long video on KZbin without taking a break! Very exciting to learn this! Awesome work guys! Side note - I am amazed on how Greg speaks. Slow and clear with the perfect pace! For the entire video, I think I caught two "hmms" lol Can you share more details on how to join discord to your space specifically? I would love to share and learn with the community!

[@AI-Makerspace]

Definitely! You can join with the link: discord.gg/RzhvYvAwzA

[@adrianemikko]

Can I fine tune in an unsupervised manner? Say I have a lot of legal docs or case proceedings. I just want the model to have more understanding of the jargons of the local law that the base LLM model might have not seen before.

[@AI-Makerspace]

Yes, technically you can!

[@adrianemikko]

Cool! Then I won’t need a labeled data set to fine tune. Thanks!

[@sushilkrsoni2796]

Thanks Great video Can you please make a video on how to fine tune dolphin-2.7-mixtral-8x7b. Or At least please tell what changes we need to make for the same

[@Clementdijoux]

Hi ! I need to know, I want use RAG and Fine-Tuning for a specific topic. This is my use case I need a chat bot base on complexe PDF's so basically the chat bot will talk like a expert so I need to give him some pdf's and fine tune this model so he can talk like a expert of this specific topic ! Please help me

[@AI-Makerspace]

This is THE question that everyone is asking! "Should I Use RAG or Fine-Tuning?" We actually just gave a talk this week on the topic through Analytics Vidhya! Check it out here: community.analyticsvidhya.com/c/datahour/should-i-use-rag-or-fine-tuning