This is data from nbeerbower/gutenberg-moderne-dpo, nbeerbower/gutenberg2-dpo and jondurbin/gutenberg-dpo-v0.1 combined and converted into a simple SFT format.
Update: Added some extra samples from PJMixers-Dev/Lit-axo-Shuffled-RL-v0.1.
The user turn contains the prompt field and the assistant turn contains the chosen (original) text from each set.
I did some text pre-processing to make the written text more compatible with other SFT training sets (e.g. changing line spacing, changing to… See the full description on the dataset page:
https://huggingface.co/datasets/ToastyPigeon/gutenberg-sft.
