Corpus Idea
#1
by Enderchef - opened
This comment has been hidden (marked as Resolved)
Actually, the best pretrain dataset for this is finephrase. Read into it; its much higher quality then Cosmopedia.
Read the blog here: https://huggingface.co/spaces/HuggingFaceFW/finephrase
Enderchef changed discussion status to closed
Cosmopedia is NOT the best lmao
Looking at finephrase, I agree
Discussion closed