datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
infinity3m-koboExperimental upload of a modified version of the BAAI/Infinity-Instruct 3M dataset.
Filtering tools were used to cut down on foreign languages, refusals and writing tasks.
While the removal of writing tasks is unusual for our community this allows tuners to use this dataset alongside superior writing data, preventing short story biases from being inserted.
KoboldAI_infinity3m-kobo-KTOSloPreferenceShareGPT