yolov8 training experience is continuously updated

Table of contents

optimizer

learning rate decay

600 batches attenuate 0.7, and the effect is found to be better

Self-encapsulated training code:

---

I use Yolov8 to train my own data set with pre-trained model

It is found that the original optimizer is SGD, the initial learning rate is 0.1, the accuracy of the first epoch after training is good, and the subsequent