it cant run continuously on MaixGo

[JasonStraka]

I‘ve trained my yolo as your demo,but the kmodel cant run continuously,as it says RAM not enough.但是我把yolo变小（把卷积核调小）后，会卡死在kpu.run_yolo2这一步，五六分钟都没有输出结果。请问是我训练出错了吗

[dextercai]

I remember the K210 chips suggests 2 types of convolution kernel (1x1、3x3). Other types of convolution kernel will cause low effectiveness.