microsoft/SimMIM issues and pull requests

#42 - Fix the packages versions

Issue - State: open - Opened by ivanpanshin 5 months ago

#41 - About mask_token in SimMIM codes

Issue - State: open - Opened by YangGangZhiQi about 1 year ago

#40 - inverse swin

Issue - State: open - Opened by betterze over 1 year ago - 1 comment

#39 - 用自监督结果的预训练权重对下游任务精度影响

Issue - State: open - Opened by swjtulinxi over 1 year ago

#38 - How can I solve this problem

Issue - State: open - Opened by haibo12 about 2 years ago - 2 comments

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

Issue - State: open - Opened by irsLu about 2 years ago

#36 - Setting for Linear eval

Issue - State: open - Opened by tonysy over 2 years ago

#35 - Tt

Issue - State: closed - Opened by Kingris9090 over 2 years ago

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

Issue - State: open - Opened by dongzhiwu over 2 years ago

#33 - Questions about AvgDist

Issue - State: open - Opened by yjsunnn over 2 years ago

#32 - Any plan to support 3D version?

Issue - State: open - Opened by james20141606 over 2 years ago

#31 - The specific configs and code for downstream tasks, like semantic segmentation and object detection

Issue - State: open - Opened by Haoqing-Wang over 2 years ago

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

Issue - State: open - Opened by nijkah over 2 years ago - 1 comment

#29 - Indentation bug in utils.remap_pretrained_keys_vit

Issue - State: open - Opened by BradMcDanel over 2 years ago

#28 - Performance using the cosine distance

Issue - State: open - Opened by LiyaoTang over 2 years ago

#27 - Train from scratch

Issue - State: open - Opened by naoki7090624 almost 3 years ago - 1 comment

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

Issue - State: open - Opened by TongZhangTHU almost 3 years ago

#25 - Could you please release the pre-trained R50 model?

Issue - State: open - Opened by liming-ai almost 3 years ago

#24 - For cnn architecture like resnet50

Issue - State: closed - Opened by Rui-Zhou-2 almost 3 years ago - 1 comment

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

Issue - State: closed - Opened by BAILUXIN almost 3 years ago

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

Issue - State: open - Opened by ucasyjz almost 3 years ago - 1 comment

#21 - SimMIM with Absolute Position Embedding

Issue - State: open - Opened by hemangchawla almost 3 years ago

#21 - SimMIM with Absolute Position Embedding

Issue - State: open - Opened by hemangchawla almost 3 years ago

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

Issue - State: open - Opened by leoozy almost 3 years ago - 1 comment

#19 - Geometric interpolation ViT

Issue - State: open - Opened by hemangchawla almost 3 years ago - 1 comment

#18 - Plan to implement downstream tasks

Issue - State: open - Opened by youngwanLEE almost 3 years ago

#17 - 192x192 pretraining resolution

Issue - State: open - Opened by christophschuhmann almost 3 years ago

#16 - Inconsistency of ViT-base config as described in the paper

Issue - State: open - Opened by Jiahao000 about 3 years ago

#15 - Loss goes nan after 14 epochs

Issue - State: open - Opened by DianCh about 3 years ago - 1 comment

#14 - linear probe

Issue - State: open - Opened by aaronsarna about 3 years ago - 7 comments

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

Issue - State: open - Opened by vadimkantorov about 3 years ago - 2 comments

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

Issue - State: closed - Opened by hao-pt about 3 years ago - 2 comments

#11 - Confusion about fine-tune

Issue - State: closed - Opened by Breeze-Zero about 3 years ago - 8 comments

#10 - why masking embedding feature maps but not input images

Issue - State: closed - Opened by baibaidj about 3 years ago - 5 comments

#9 - Could you please release the mask prediction visualization code?

Issue - State: closed - Opened by maxuetao about 3 years ago - 1 comment

#8 - USE_RPB and USE_SHARED_RPB are inconsistent in pretrain and finetune

Issue - State: closed - Opened by haooooooqi about 3 years ago - 1 comment

#7 - Could you please release the finetune log of swin and vit?

Issue - State: open - Opened by zws98 about 3 years ago

#6 - Can you release training log for ViT 800 epochs?

Issue - State: open - Opened by gaopengcuhk about 3 years ago - 1 comment

#5 - About data augment

Issue - State: closed - Opened by peiyingxin about 3 years ago - 1 comment

#4 - Information about relative positional encoding

Issue - State: closed - Opened by marc345 about 3 years ago - 1 comment

#3 - Missing hyperparameter

Issue - State: closed - Opened by haofanwang about 3 years ago - 1 comment

#2 - A question about masking strategy

Issue - State: closed - Opened by xiaohu2015 about 3 years ago - 1 comment

#1 - When will you release the code for pre-training

Issue - State: closed - Opened by BaohaoLiao over 3 years ago - 1 comment

GitHub / microsoft/SimMIM issues and pull requests