kekmodel/FixMatch-pytorch

Issues

• 训练的准确率一直是10%，显然是没有作用的因为一共就只有10个类别

  #54 opened 4 months ago by hanyiwei799
  0

• train failed when EMA mode is off

  #42 opened 4 years ago by frozenzo
  5

• 每次加载训练好的模型，结果都不一样

  #53 opened 10 months ago by onlinehuazai
  0

• RuntimeError: "nll_loss_forward_reduce_cuda_kernel_2d_index" not implemented for 'Int'

  #52 opened 10 months ago by wxqyyds
  0

• Using fixmatch to train on my dataset, LR = 0.03, batchsize = 64. When the unlabeled loss increases gradually, the model gets worse and worse, and the loss of training set will decrease, but the loss of verification set will increase, and the accuracy will decrease?

  #23 opened 3 years ago by a571-op
  1

• why add a interleave before the input?

  #50 opened a year ago by williamium3000
  3

• Why the result with this implementation is better than reported one in the original paper?

  #25 opened 4 years ago by sooonwoo
  2

• Poor Results on CIFAR-100 with Smaller Models (e.g., ResNet34, ResNet18)

  #51 opened 2 years ago by tsunghan-wu
  0

• total numer of iterations

  #35 opened 4 years ago by holzbock
  2

• num_labeled in DistributedDataParallel

  #5 opened 5 years ago by LiheYoung
  9

• Can't reproduce the result with single GPU

  #49 opened 3 years ago by moucheng2017
  0

• When fix patch is applied to my dataset, the dataset is unbalanced. Why does the training set perform well, but the loss of verification set first decreases and then keeps rising?

  #24 opened 4 years ago by a571-op
  3

• Training time without distributed training?

  #45 opened 3 years ago by 07Agarg
  2

• All labeled examples are as part of unlabeled data without using their labels

  #48 opened 3 years ago by chengerr
  1

• Why the batch_size of CIFAR-100 set to 64 (total number over 4 GPU) ?

  #46 opened 3 years ago by kevinbro96
  0

• The RandAugment Implementation Is Wrong

  #39 opened 3 years ago by psellcam
  3

• Performance not reproduced

  #38 opened 3 years ago by Hyun1A
  0

• FixMatch for Multi-class Classification

  #47 opened 3 years ago by GuoJingtao-1997
  0

• Invalid syntax in train.py

  #37 opened 3 years ago by mdziezyc
  1

• RandAugment Implementation

  #33 opened 3 years ago by psellcam
  3

• What is the purpose of the parameter 'activate_before_residual' in WRN?

  #32 opened 3 years ago by KittyStore
  0

• Have you tried SVHN dataset?

  #30 opened 3 years ago by fiona-lxd
  3

• No Validation Set

  #43 opened 3 years ago by 07Agarg
  1

• v divided by 2 problem in Cutout

  #44 opened 3 years ago by Fordacre
  1

• Purpose of Interleave

  #36 opened 4 years ago by feras-oughali
  3

• Lack of seed during random labeled image selection may be leading to better performance on training resumption

  #41 opened 3 years ago by dsouzinator
  5

• Can not get repo reported accuracy for cifar10@40.5

  #29 opened 4 years ago by Kyfafyd
  5

• Wrong use of AverageMeter

  #40 opened 4 years ago by huberl
  1

• set_epoch with DistributedSampler

  #34 opened 4 years ago by TonyLianLong
  1

• How can I train model using multi GPUS in a machine?

  #31 opened 4 years ago by Kyfafyd
  7

• Accuracy lower than reported

  #27 opened 4 years ago by fil-mp
  2

• Have you fetch labelled samples for each class in balance?

  #28 opened 4 years ago by Kyfafyd
  2

• RuntimeError after 1 epoch

  #26 opened 4 years ago by valeriydutt
  2

• How to delete this question, I've solved it

  #22 opened 4 years ago
  0

• BatchNorm has a biased estimation

  #7 opened 4 years ago by YUE-FAN
  4

• Any plans to add CTAugment?

  #12 opened 4 years ago by monkeysforever
  2

• Datasets

  #20 opened 4 years ago by fil-mp
  2

• fail to reproduce the results of your experiment

  #21 opened 4 years ago by Jiawen-huang
  1

• What does the function `interleave` do?

  #19 opened 4 years ago by charlesCXK
  4

• CIFAR100 has different architecture from official implementation

  #13 opened 4 years ago by zijian-hu
  1

• Accuracy curves

  #14 opened 4 years ago by bkj
  1

• Unsupervised loss part for one single class

  #11 opened 4 years ago by dendrobiumz
  1

• Cannot reproduce 94.72% Top-1 accuracy on CIFAR10 (4000 labeled data)

  #8 opened 4 years ago by tqtrunghnvn
  15

• PyTorch Distributed Training

  #16 opened 4 years ago by ZhuoranYu
  1

• unlabeled data contain the labeled data?

  #18 opened 4 years ago by yaojunr
  1

• All labeled examples are as part of unlabeled data without using their labels?

  #10 opened 4 years ago by Chen-Song
  6

• custom dataset

  #17 opened 4 years ago by KirillHiddleston
  1

• 1

  #9 opened 4 years ago by Chen-Song
  0

• Weight decay in EMA?

  #15 opened 4 years ago by bkj
  2

• cos learing rate

  #6 opened 5 years ago by YUE-FAN
  2