liuzechun/ReActNet

Issues

• Depth-wise convolutions

  #36 opened a month ago by shekhovt
  1

• Error occurred in loading resnet model.

  #6 opened 4 years ago by BobZhang321
  2

• KD loss training

  #35 opened 2 years ago by snownus
  0

• Does KD_Loss need temperature?

  #34 opened 2 years ago by XA23i
  0

• How many gpus was used when running the experiments? This can affect results especially with batchnorm + torch.DataParallel

  #33 opened 2 years ago by Phuoc-Hoan-Le
  1

• I see scheduler.step() is being invoked right before the for loop in the train function or before training for one epoch

  #32 opened 2 years ago by Phuoc-Hoan-Le
  1

• Is 69.49% top1 ok with reactnet with mobilenet backbone with full-precision weights and binary activations?

  #31 opened 2 years ago by Phuoc-Hoan-Le
  1

• Is there code for reactnet in complete full-precision?

  #30 opened 2 years ago by Phuoc-Hoan-Le
  0

• About the GPU memory consumption.

  #26 opened 2 years ago by AlbertHuyb
  2

• blocks where input_channels are bigger than output_channels

  #28 opened 2 years ago by rainyBJ
  2

• Question about the result of ReAct-ResNet

  #29 opened 2 years ago by ThisisBillhe
  1

• Binary weights

  #27 opened 2 years ago by techmonsterwang
  2

• Has anyone successfully trained on cifar10/100 dataset?

  #25 opened 2 years ago by ThisisBillhe
  3

• Some questions in the code

  #24 opened 2 years ago by THKIM-CAS
  1

• Intermediate training results

  #21 opened 2 years ago by sfalkena
  2

• Could you share ResNet_34 model with ReAct binary you have been trained ?

  #18 opened 2 years ago by puhan123
  1

• Question on Reported FLOPs

  #17 opened 2 years ago by ooodragon94
  5

• Reported OPs

  #23 opened 2 years ago by HatemSelim94
  1

• Regarding a line of codes

  #10 opened 2 years ago by BNAadministrator3
  7

• About the learnerable bias parameter

  #16 opened 2 years ago by zz-cloud
  2

• Question about flops

  #19 opened 3 years ago by TIEHua
  1

• About code "-out3.detach() + out3 (line 35 in the birealnet.py)"

  #22 opened 3 years ago by hadhe145
  1

• Training setup is different

  #15 opened 3 years ago by resurgo97
  5

• Question about Reduction Block structure in reactnet

  #20 opened 3 years ago by xvyaward
  1

• param shape error of Model-ReAct-ResNet

  #14 opened 3 years ago by ustczwq
  2

• binary convolution's weight are not -1or 1.

  #12 opened 3 years ago by yuxianzhi
  1

• Why use torch.sign()?

  #11 opened 4 years ago by theqihao
  4

• Weights

  #8 opened 4 years ago by Skyve1996
  6

• Test accuracy reduced to 0%, but the train accuracy remains high.

  #9 opened 4 years ago by dezengzang
  1

• Mismatch on the number of iterations in training MobileNet

  #7 opened 4 years ago by licj15
  0

• Are the parameters of trained model actually float32 type?

  #5 opened 4 years ago by youngboy52
  1

• The link is invalid

  #4 opened 4 years ago by BobZhang321
  1

• Are all the intermediate activations during inference 1 bit?

  #3 opened 4 years ago by jiangwx
  0

• Excuse me , why is your conv3x3 not DW conv?

  #2 opened 4 years ago by loganyan
  2

• About the computation complexity

  #1 opened 4 years ago by ycjing
  7