Wukong-0.1-Mistral-7B-v0.2 is a dealigned chat finetune of the original fantastic Mistral-7B-v0.2 model by the Mistral team.
This model was trained on the teknium OpenHeremes-2.5 dataset, code datasets from Multimodal Art Projection
https://m-a-p.ai, and the Dolphin dataset from Cognitive Computations
https://erichartford.com/dolphin 🐬
This model was trained for 3 epochs over 4 4090's.
