ZhaoYu@TGU
ZhaoYu@TGU
另外，请问在将图像打成16*16的Patch后，共196个Patch，那每个Patch维度应该是256*3=768，这里是否是像VIT一样有一个线性层来把其特征维度进行了一个转换呢？
您好，还有个问题，就是在VIG中MLP层全通过1x1的卷积层实现，这样有什么优势吗？我看到引用的那篇deepgcns也是这样操作的
Hello. I think I've solved the problem. The function _get_target_encoder has a decorator on it, which makes it run only once, not every time. You can see this process with...
For example, if the number of parameters for random cropping and horizontal flipping is not the same, how can aware-augment?
Or can AugSelf only compare the same enhancement category?
您好，另外我看算法1和2里面完全没有提更新Encoder的参数啊？
Should there be an operation like this? `data.x = data.x[idx_nondrop]`