tamakoji/pytorch-syncbn

Issues

• Error about BatchNorm2dSync for 1d input in eval model, but I can use this BatchNorm2dSync for 1d input in train model

  #9 opened 4 years ago by XiwuChen
  0

• I have not seen any differences between m1 and m2, Why?

  #11 opened 5 years ago by tjpulkl
  2

• subprocess.CalledProcessError: Command '['which', 'c++']' returned non-zero exit status 1

  #7 opened 6 years ago by koncle
  2

• support for fp16

  #12 opened 5 years ago by antorsae
  0

• Please make this into a pip package.

  #10 opened 5 years ago by ericlearning
  0

• Does this support 3dBN?

  #8 opened 6 years ago by zyl1336110861
  0

• I get stuck after several iterations

  #6 opened 6 years ago by gakkiri
  0

• The gpu memory increase when number of gpus > 4.

  #5 opened 6 years ago by justttry
  0

• Gpu memory problem. In training phase, it will gradually occupy all the gpu memory.

  #1 opened 6 years ago by weitaoatvison
  13

• RuntimeError: one of the variables needed for gradient computation has been modified by an inplace operation

  #4 opened 6 years ago by ElbertFang
  0

• Torch 1.0.1 compatibility

  #3 opened 6 years ago by sunggukcha
  1

• Can it use in 3d？

  #2 opened 7 years ago by wangxiaodong1021
  0