xuemo188

Results 6 issues of xuemo188

大佬请问下,在10分类中vgg,resnet,densenet所对应的稀疏化系数0.0001 0.00001,0.00001,那么在100分类稀疏化系数是怎样的?

论文采用的是对BN中缩放因子进行L1稀疏化,我看了代码发现好像和其他的过程的L1稀疏化不一样,torch中设置weight_decay默认是所有参数L2正则化训练 ![image](https://user-images.githubusercontent.com/55520460/69701603-4445a380-1128-11ea-8375-7ba93f2e9e94.png) 请问这个是怎么体现L1正则化的

File "vggprune.py", line 73, in mask = weight_copy.gt(thre).float().cuda() RuntimeError: Expected object of type torch.cuda.FloatTensor but found type torch.FloatTensor for argument #2 'other'

Traceback (most recent call last): File "vggprune.py", line 73, in mask = weight_copy.gt(thre).float().cuda() RuntimeError: Expected object of type torch.cuda.FloatTensor but found type torch.FloatTensor for argument #2 'other' 我用的环境是python3.6 torch0.4.1

VOC数据集稀疏化训练map很低,使用的yolov3.weights,训练100步map只有10几,loss也在震荡,是不是稀疏化训练只适合单一目标,例如:手的数据集 在不用稀疏化训练时,就能保证很高的map,大佬请教下