jianlong-yuan/syncbn-tensorflow

Issues

• SyncBN加了之后模型不收敛

  #9 opened 3 years ago by zihao-lu
  1

• Do you need to construct the network twice for training and inference accordingly?

  #2 opened 4 years ago by panfengli
  1

• error when using recurrent bidirectional batch normalization (multi-gpu)

  #3 opened 4 years ago by AzizCode92
  3

• why not change from “tf.nn. fused_batch_norm” to "tf.nn.batch_normalization"?

  #6 opened 4 years ago by sudongxiang
  1

• Any training examples?Error happens when opt.compute_gradients(lossCTC)

  #8 opened 6 years ago by wangershi
  5

• 一个关于gen_nccl_ops.nccl_all_reduce的问题

  #4 opened 6 years ago by stillwaterman
  6

• Error of 'outputs.device[-1]'

  #7 opened 6 years ago by wangq95
  6

• Question: tf.nn.batch_normalization vs tf.nn.fused_batch_norm

  #5 opened 6 years ago by sryap
  2

• Undefined role of the 'center' and 'scale' argument

  #1 opened 6 years ago by camhao168
  1