You want to check out this fine tuning dataset

#16
by KnutJaegersberg - opened

We've got a 7B llama with this dark magic created instruct tuning dataset which is as good as chatgpt.
you want to use it.

https://huggingface.co/TheBloke/wizardLM-7B-HF

https://github.com/nlpxucan/WizardLM

perhaps this dataset is all you need

Added to v11 :)
Although it's tiny comparing with other data so I wonder how much it helps lol.

BlinkDL changed discussion status to closed

That's my point. Perhaps you can fine tune only using this dataset, ignoring all others.
Quality>Quantity

Sign up or log in to comment