SwinTransformer/Transformer-SSL

Issues

• Gradient overflow. Skipping step, loss scaler 0 reducing loss scale to xxxx

  #19 opened a year ago by AlphaNext
  0

• pretrain fails when image categories are similar

  #18 opened a year ago by 894665169
  0

• Strange output log

  #2 opened 4 years ago by launchauto
  12

• Will apex mixed precision training affect the accuracy of the model?

  #17 opened 2 years ago by Pang-b0
  0

• AttributeError: TRAINING_IMAGES

  #14 opened 2 years ago by Rocky1salady-killer
  1

• Download links of DeiT-S and Swin-T backbone models are interchanged

  #16 opened 2 years ago by kavin-du
  0

• I am training chuxian on a local server with only one GPU error

  #15 opened 2 years ago by L-xiaolu
  0

• Train MoBy-SwinT on local machine with one GPU

  #11 opened 3 years ago by TNA8
  2

• The interpolation method for BYOL augmentation is wrong

  #12 opened 3 years ago by songkangsg
  1

• TypeError: 'Compose' object is not iterable

  #5 opened 3 years ago by chenhaoxing
  2

• Have you tried any other initial patch size in the swin transformer apart from the patch size = 4?

  #13 opened 3 years ago by sfarkya04
  0

• Some questions about relative_position_index and attn_mask

  #10 opened 3 years ago by Haoqing-Wang
  0

• 300 epoch DeiT-S and Swin-T checkpoints dowload links are same

  #9 opened 3 years ago by Breeze-Zero
  0

• dataloader error

  #8 opened 3 years ago by niutransWZY
  0

• How to load a checkpoint when using the swin transformer as the backbone in a Mask-RCNN model

  #7 opened 3 years ago by amobiny
  0

• Question about the detection/segmentation results

  #6 opened 3 years ago by amobiny
  0

• Cannot import vit_deit_small

  #3 opened 3 years ago by michuanhaohao
  1

• config setting "NORM_BEFORE_MLP" takes no effect

  #4 opened 3 years ago by flishwang
  0

• Multi-machine training

  #1 opened 4 years ago by launchauto
  0