Hi Guys,
So to improve the efficiency of LLM, you start with Prompt, then RAG, then finetune if needed. Can some give me a brief of different finetuning methods.
Also what is instruction fine tuning, PEFT, LORA? I am getting so many different answers everywhere, its confusing me. can some give me a clear picture on finetuning and different types of it and any thing else i need to know related to fine tuning
Check out the Generative AI with Large Language Models specialization!
hi @madhu95
Instruction fine-tuning, PEFT, LORA, QLORA are all fine tuning methods of LLMs based on its specific features.
RAG again is used to improve the performance of an LLM based on using a curated database, then allowing to retrieve relevant information and create dynamic generated responses.
One can approach to create an LLM with these features at beginning or try to incorporate any of the fine tuning methods to improve the output results.
Use of LoRA or QLoRA again depends on the data you’re using as the tuning speed depends on the programmers GPU configuration and other features involved with an LLM creation.
As mentioned by @gent.spah, ideally to go with that specialisation will improve your basic understanding of LLM and then there are many shorts courses in LLM from building your own vector database to using RAG application, which will further increase your understanding towards LLMs.
No matter what LLMs is an evolving topic and something you learn today, might be old tomorrow and goes on but having an understanding of this all, will make your approach towards LLM understanding open and experimenting for creative ideas.
Hope this helps, keep learning!!!
Regards
DP
I’m planning to, Just wanted a brief on it for now
Thanks @Deepti_Prasad