This is a mix of three datasets LimaRP, PIPPA and openerotica/freedom-rp and it has been made into 8K token chunks (using mistralai/Mistral-Nemo-Base-2407's tokenizer).
I reduced the amount of data from PIPPA and freedom-rp for it to be more balanced with the LimaRP dataset. (PIPPA is normally around 20k examples and freedom-rp is 9k but LimaRP is about 1k, so for this I added 1669 examples from PIPPA and 2590 from freedom-rp respectively.)
It uses ShareGPT formatting.
Original datasets:… See the full description on the dataset page:
https://huggingface.co/datasets/mpasila/LimaRP-PIPPA-freedom-rp-Mix-8K.
