simple-llm-finetuner
simple-llm-finetuner copied to clipboard
Examples to get started with
I see there's a few examples in the repo - it would be great for a tutorial to accompany them, so newbies dipping their feet in this for the first time can get an idea of what's happening / score some early wins.
I have 3 example datasets under example-datasets directory to play with.
Could you add a tutorial for using them as well, that most importantly shows the result of a successful training?
https://docs.google.com/spreadsheets/d/1QSwJFiyzUQ6H1CloDmJWcHJfYiT7SVxfwBDOOcbvFEo/edit#gid=1356873557
I have some training samples using here - ive had very good luck using reflection methods to get imo much better responsiveness.
What are reflection methods - and would you mind writing up a tutorial using simple-llama-finetuner that given a dataset X, parameters Y, will reliably produce results Z? This way beginners will be able to get confidence in what they're doing.