JunrQ/NAS

Issues

• training FBNet on CIFAR-10

  #15 opened 6 years ago by sl-zeng
  8

• imagenet training code bug.

  #23 opened 6 years ago by weitaoatvison
  2

• Some questions in measure_speed.py

  #22 opened 6 years ago by Jihao-Li
  0

• Forward and backward for FBNet

  #13 opened 6 years ago by GuntherZhong
  5

• got nan when calculate gumbel_softmax

  #21 opened 6 years ago by ChenMinQi
  1

• Model structure for the CIFAR-10, FLOPS and params

  #20 opened 6 years ago by mrastegari
  0

• Params, BN, Learned Distribution and FC width layer about FBnet and your implementation

  #19 opened 6 years ago by zhaohui-yang
  5

• Could you please upload the code for retraining SNAS?

  #17 opened 6 years ago by xdhhh
  1

• How to reproduce your result?

  #18 opened 6 years ago by LeoWang
  0

• Do you have any good choice for the three levels of resource constraint?

  #12 opened 6 years ago by MarkAlive
  2

• have you reproduced the results in original paper?

  #2 opened 6 years ago by iamhankai
  12

• I try to train a sample net

  #16 opened 6 years ago by ChengTsang
  2

• I mesure the speed in my device

  #14 opened 6 years ago by ChengTsang
  4

• kernel size in class FBnetblock

  #8 opened 6 years ago by 666zz666
  9

• what meaning MAC means in MixedOp?

  #9 opened 6 years ago by MarkAlive
  5

• Why the calculation of FLOPs of dilconv didnt need to consider the parameter 'dilation'?

  #10 opened 6 years ago by MarkAlive
  1

• FLOP should be changed to FLOP1

  #11 opened 6 years ago by MarkAlive
  0

• Which branch is latest

  #7 opened 6 years ago by MichaelGou1105
  1

• bug in model_parameters of snas

  #6 opened 6 years ago by MarkAlive
  3

• Loss of snas doesn't decrease in `train_cifar10.py`

  #5 opened 6 years ago by JunrQ
  0

• The value of lat is different from total_loss - ce

  #4 opened 6 years ago by JunrQ
  0

• Test, initialization in multi-dev training is different for parameters

  #1 opened 6 years ago by JunrQ
  0