3 ms·
OpenHermesPreferences: Dataset of ~1M AI preferences from teknium/OpenHermes-2.5
- kashifr 3y agoThe dataset can be used for training preference models or aligning language models through techniques like Direct Preference Optimization (DPO).
loading thread…