Chongruo Wu
Chongruo Wu
> > I think @LiheYoung is correct -- w/ `DistributedDataParallel` you launch N copies of the program. If you don't set the seed, then `np.random` will sample the dataset differently,...
https://github.com/kekmodel/FixMatch-pytorch/blob/10db592088432a0d18f95d74a2e3f6a2dbc25518/dataset/cifar.py#L102-L106 For each GPU, the corresponding process will create a CIFAR dataset. Since we don't set the fixed seed, the idx is shuffled (line 104) in different ways on different...
One possible reason is that the input mean and std are not set correctly. I found the std is [1,1,1] in openseg.pytorch https://github.com/openseg-group/openseg.pytorch/blob/6787d215ccc28714b265c48bd9cfc41b41c81e28/configs/cityscapes/R_101_D_8.json#L70. The mean and std values in our...
Thanks. We did not train the RetinaNet.
We integrate it with resnet.py. Some new hyper-parameters(e.g., RADIX) are defined for ResNeSt.
We didn't modify the dataloder
Thanks for your advice! For the pretrained model, you can download the model with standard names from torch.hub ```shell # load pretrained models, using ResNeSt-50 as an example net =...
please refer to #25
should use the current codebase. We define several hyper-parameters, including radix. #14