Using the same model and data, the accuracy can reach 0.998 in version 2.0, but after upgrading version 2.6, it can only reach about 0.288 in the same training time.Who can tell me why this happened
Thanks ！
Please provide standalone code to reproduce the issue? Thank you