Zijie Guo

> > > > > > 就像我说的.先检查一下你的记忆.正如我所看到的，您正在使用窗口，因此任务管理器将帮助您查看内存状态。 > > > > > > > > > > > > > > > 我想这是关于注意力图的问题 > > > > > > >...

> 大佬换成你的代码后 这个错误是为什么 You don't have sufficient GPU mem. Try with smaller batch_size. Or you should give up on this.

For this I would recommand this paper **Rethinking ImageNet Pre-training** from Kaiming He. @ICCV2019. In short. No, pre-train is not necessary. You can achieve the same level with longer training.

They are different for different scales. For small you devide by 8 then middle you divide them by 16 and for large you divide them by 32. see https://github.com/jingtianyilong/YOLOv4-pytorch/blob/master/get_anchor.py

> Thanks for your rapid reply! > > Sorry for my confusing writing. What I want to point out is the mismatch between `MODEL['ANCHORS']` values and the commented anchor values....

network and pretrain weight unmatch

So obviouslly yolov4 works but the mobilenet version don't. For the current error, label number don't match. Check your dataset setting. And try to debug yourself first. Tell us what...

Pls do something on your own and tell us what you have done to debug. Simply saying "I came across a bug" doesn't provide enough information to even understand your...

This is really straight forward. not enough values to unpack (expected 3, got 2) `_, p_d, beta` are 3 value you expected. But obviously enough, `self.model(img)` only return 2 value,...

> > @jingtianyilong请问 > > 该怎么解决 > > 可以将您的代码分享一下吗？由于我运行的博主代码有bug，目前无法解决。邮箱：[[email protected]](mailto:[email protected]),非常感谢 I use my own code which is heavily modified version of this repo. You can check my repo： https://github.com/jingtianyilong/YOLOv4-pytorch