GeorgeCazenavette/mtt-distillation

Issues

• Why grand_loss will gradually increase and become NaN

  #42 opened 4 months ago by ManZhao123
  0

• Checkpoints of models

  #38 opened a year ago by DeepOceanDeep
  1

• Quick Question regarding class index mapping for the custom imagenet subsets

  #41 opened a year ago by meghbhalerao
  4

• values for max_start_epoch

  #4 opened 2 years ago by ankanbhunia
  8

• how to do distillation for a model other than VGG?

  #33 opened 2 years ago by coloedrainbow
  2

• details about the full real subsets of ImageNet

  #35 opened a year ago by zeyuanyin
  2

• have trouble at distillating with VGG networks

  #20 opened a year ago by ArmandXiao
  4

• Question about imagenette

  #22 opened a year ago by yaolu-zjut
  1

• Grand loss curve

  #27 opened 2 years ago by XuyangZhong-29
  1

• N << M or N >> M

  #28 opened a year ago by HaowenGuan
  7

• Model training with the released synthetic CIFAR-10 (ipc=50)

  #29 opened 2 years ago by LeavesLei
  4

• ReparamModule Usage

  #30 opened a year ago by Sinp17
  2

• CIFAR 10 ipc 10 Hyperparameters

  #31 opened 2 years ago by WeizhiGao
  2

• About Hyper-paramters

  #32 opened 2 years ago by maple-zhou
  1

• Normalization of dataset

  #37 opened a year ago by f-amerehi
  4

• Running this project with PCAM instead of CFAR

  #36 opened a year ago by wojo501
  0

• Could you please share the variance of the performance in the KIP to NN column?

  #34 opened 2 years ago by NiaLiu
  0

• question about learning rate

  #24 opened 2 years ago by harrylee999
  2

• What is ZCA

  #26 opened 2 years ago by youdutaidi
  2

• How did you get x̄ ± s in table 1

  #23 opened 2 years ago by NiaLiu
  2

• question on Tiny imagenet ipc=50

  #25 opened 2 years ago by DongyaoZhu
  2

• GPU requirement

  #21 opened 2 years ago by rave78
  2

• Unrolled optimization

  #17 opened 2 years ago by vadimkantorov
  1

• Reproduce cross-architecture performance

  #18 opened 2 years ago by NiaLiu
  1

• Negative LR

  #16 opened 2 years ago by cliangyu
  3

• Expert trajactory performance

  #8 opened 3 years ago by 1215481871
  3

• A question about backbone networks

  #15 opened 2 years ago by imesu2378
  3

• distill.py loss = nan

  #6 opened 2 years ago by yangyangtiaoguo
  1

• Experience on hyper-parameters

  #14 opened 2 years ago by Huage001
  1

• how to use images？

  #11 opened 2 years ago by Fduxiaozhige
  7

• A question for the paper

  #12 opened 2 years ago by alittleCVer
  2

• The clip value

  #13 opened 2 years ago by tao-bai
  2

• Where did you get the acc 36.1% from the paper Dataset Distillation with Infinitely Wide Convolutional Networks

  #10 opened 2 years ago by NiaLiu
  3

• args.mom in buffer.py

  #9 opened 3 years ago by liuyugeng
  2

• Buffer.py has no args.texture

  #5 opened 3 years ago by liuyugeng
  1

• about hyperparameter: learning rate about updating condenses samples

  #3 opened 3 years ago by Alan-Qin
  5

• how does it work on a new model arch?

  #2 opened 3 years ago by lucasjinreal
  1

• Great work! What is the training time for the distillation stage?

  #1 opened 3 years ago by ankanbhunia
  2