ShenNong TCM Dataset is a dataset designed specifically for large-scale language model training and evaluation in the field of TCM. It contains more than 110,000 instruction data, which are generated through entity-centric self-instruction. It focuses on the core entities and different intent scenarios in the field of TCM, which can not only improve the model's ability to answer questions related to TCM, but also assist TCM diagnosis and provide personalized medical advice.