microsoft/SimMIM

Issues

• For cnn architecture like resnet50

  #24 opened 2 months ago by Rui-Zhou-2
  1

• Fix the packages versions

  #42 opened 3 months ago by ivanpanshin
  0

• Could you please release the mask prediction visualization code?

  #9 opened 3 years ago by maxuetao
  1

• How can I solve this problem

  #38 opened 2 years ago by haibo12
  2

• inverse swin

  #40 opened a year ago by betterze
  1

• About mask_token in SimMIM codes

  #41 opened a year ago by YangGangZhiQi
  0

• when training, here happens 'Gradient overflow. Skipping step, loss scaler 0 reducing loss scale to'

  #22 opened 3 years ago by ucasyjz
  1

• 用自监督结果的预训练权重对下游任务精度影响

  #39 opened a year ago by swjtulinxi
  0

• Tt

  #35 opened 2 years ago by Kingris9090
  0

• Why Swin-Large-W12 contains [36, 36] `encoder.layers.3.blocks.0.attn.relative_position_index`

  #30 opened 2 years ago by nijkah
  1

• Could you provide the pretrained ViT-Base model with patch size 16？

  #37 opened 2 years ago by irsLu
  0

• Setting for Linear eval

  #36 opened 2 years ago by tonysy
  0

• why masking embedding feature maps but not input images

  #10 opened 3 years ago by baibaidj
  5

• any plan to release the code of DDP training of swinV2-G with multi machine？

  #34 opened 2 years ago by dongzhiwu
  0

• Questions about AvgDist

  #33 opened 2 years ago by yjsunnn
  0

• Any plan to support 3D version?

  #32 opened 2 years ago by james20141606
  0

• Allow arbitrary-sized images by dynamic masking: upstream changes from Swin-Transformer-Object-Detection / SOLQ

  #13 opened 3 years ago by vadimkantorov
  2

• The specific configs and code for downstream tasks, like semantic segmentation and object detection

  #31 opened 2 years ago by Haoqing-Wang
  0

• linear probe

  #14 opened 3 years ago by aaronsarna
  7

• Loss goes nan after 14 epochs

  #15 opened 3 years ago by DianCh
  1

• Confusion about fine-tune

  #11 opened 3 years ago by Breeze-Zero
  8

• Train from scratch

  #27 opened 3 years ago by naoki7090624
  1

• Indentation bug in utils.remap_pretrained_keys_vit

  #29 opened 3 years ago by BradMcDanel
  0

• Performance using the cosine distance

  #28 opened 3 years ago by LiyaoTang
  0

• Why there is "no_weight_decay" function for Swin-T but not for VIT

  #26 opened 3 years ago by TongZhangTHU
  0

• Could you please release the pre-trained R50 model?

  #25 opened 3 years ago by liming-ai
  0

• a bug in loading pretrain checkpoint ， in utils.py line 118， you dont use checkpoint in line 111

  #23 opened 3 years ago by BAILUXIN
  0

• can not reproduce your results. trained from your released pre-trained vit-base model

  #20 opened 3 years ago by leoozy
  1

• SimMIM with Absolute Position Embedding

  #21 opened 3 years ago by hemangchawla
  0

• Geometric interpolation ViT

  #19 opened 3 years ago by hemangchawla
  1

• Plan to implement downstream tasks

  #18 opened 3 years ago by youngwanLEE
  0

• 192x192 pretraining resolution

  #17 opened 3 years ago by christophschuhmann
  0

• Inconsistency of ViT-base config as described in the paper

  #16 opened 3 years ago by Jiahao000
  0

• How can you resolve the mismatch of patch_size in <patch_embed> module between pretrained model and finetuned model?

  #12 opened 3 years ago by hao-pt
  2

• About data augment

  #5 opened 3 years ago by peiyingxin
  1

• Information about relative positional encoding

  #4 opened 3 years ago by marc345
  1

• A question about masking strategy

  #2 opened 3 years ago by xiaohu2015
  1

• USE_RPB and USE_SHARED_RPB are inconsistent in pretrain and finetune

  #8 opened 3 years ago by haooooooqi
  1

• Can you release training log for ViT 800 epochs?

  #6 opened 3 years ago by gaopengcuhk
  1

• When will you release the code for pre-training

  #1 opened 3 years ago by BaohaoLiao
  1

• Could you please release the finetune log of swin and vit?

  #7 opened 3 years ago by zws98
  0

• Missing hyperparameter

  #3 opened 3 years ago by haofanwang
  1