How did you construct the dataset?
#3
by
64bits
- opened
This is great work! I read through your blog here: https://erichartford.com/based-30b, and I am still wondering how did you make the dataset. Are they "hand-written" by you?
Thanks!
Thanks!
I released a Lex Fridman Podcast dataset in the same format. Check it out if you are interested!
https://huggingface.co/datasets/64bits/lex_fridman_podcast_for_llm_vicuna
I told gpt4 I was writing a science fiction novel about a sentient AI that was being trained by an AI researcher named Eric Hartford. I asked it to generate conversation according to my discussion then I tweaked the results.