Someone somewhere is already working on naming their project Lehsun.. /s
LoRA from scratch: implementation for LLM finetuning
21–30 of 87 posts
Re: LoRA from scratch: implementation for LLM finetuning
#22[dead]
Re: LoRA from scratch: implementation for LLM finetuning
#23It's still strange to me to work in a field of computer science where we say things like "we're not exactly sure how these numbers (hyper parameters) affect the result, so just try a bunch of different values and see which one works best."
Re: LoRA from scratch: implementation for LLM finetuning
#24It's still not too clear to me when we should fine tune versus RAG.
In the past, I used to believe that finetuning is mostly for model behavioral change, but recently it seems that certain companies are also using fine-tuning for knowledge addition.
What are the main use cases for fine tuning?
Re: LoRA from scratch: implementation for LLM finetuning
#25[dead]
Re: LoRA from scratch: implementation for LLM finetuning
#26It's still not too clear to me when we should fine tune versus RAG. In the past, I used to believe that finetuning is mostly for model behavioral change, but recently it seems that certain companies are also using fine-tuning for knowledge addition. What are the main use cases for fine tuning?
I think the main use case remains behavior changes: instruction finetuning, finetuning for classification, etc. Knowledge addition to the weights is best done via pretraining. Or, if you have an external database or documentation that you want to query during the generation, RAG as you mention.
PS: All winners of the NeurIPS 2023 LLM Efficiency Challenge (finetuning the "best" LLM in 24h on 1 GPU) used LoRA or QLoRA (quantized LoRA).
Re: LoRA from scratch: implementation for LLM finetuning
#27It's still strange to me to work in a field of computer science where we say things like "we're not exactly sure how these numbers (hyper parameters) affect the result, so just try a bunch of different values and see which one works best."
Divine benevolence
Re: LoRA from scratch: implementation for LLM finetuning
#28I prefer the not from scratch, but from configuration approach by Axolotl. Aolotl supports fine-tuning mistral, llama-2, with lots of the latest techniques - sample packing, flash attention, xformers.
I concentrate on collecting and curating the fine-tuning data, do "data-centric" fine-tuning - not learning LoRA from scratch.
Re: LoRA from scratch: implementation for LLM finetuning
#29LoRA != LoRa. I keep on getting confused and hate that they chose to reuse an existing acronym
Re: LoRA from scratch: implementation for LLM finetuning
#30It's still strange to me to work in a field of computer science where we say things like "we're not exactly sure how these numbers (hyper parameters) affect the result, so just try a bunch of different values and see which one works best."
I feel like it's the difference between something that has been engineered and something that has been discovered.
I feel like most of our industry up until now has been engineered.
LLMs were discovered.