microsoft / SimMIM

This is an official implementation for "SimMIM: A Simple Framework for Masked Image Modeling".

https://arxiv.org/abs/2111.09886

microsoft/SimMIM Issues

inverse swin
Updated 2 months ago1
About mask_token in SimMIM codes
Updated 4 months ago
when training, here happens 'Gradient overflow. Skipping step, loss scaler 0 reducing loss scale to'
Updated 8 months ago1
用自监督结果的预训练权重对下游任务精度影响
Updated 9 months ago
How can I solve this problem
Updated 10 months ago2
Tt
Closed a year ago
Why Swin-Large-W12 contains [36, 36] `encoder.layers.3.blocks.0.attn.relative_position_index`
Updated a year ago1
Could you provide the pretrained ViT-Base model with patch size 16？
Updated a year ago
Setting for Linear eval
Updated 2 years ago
why masking embedding feature maps but not input images
Closed 2 years ago5
any plan to release the code of DDP training of swinV2-G with multi machine？
Updated 2 years ago
Questions about AvgDist
Updated 2 years ago
Any plan to support 3D version?
Updated 2 years ago
Allow arbitrary-sized images by dynamic masking: upstream changes from Swin-Transformer-Object-Detection / SOLQ
Updated 2 years ago2
The specific configs and code for downstream tasks, like semantic segmentation and object detection
Updated 2 years ago
linear probe
Updated 2 years ago7
Loss goes nan after 14 epochs
Updated 2 years ago1
Confusion about fine-tune
Closed 2 years ago8
Train from scratch
Updated 2 years ago1
Indentation bug in utils.remap_pretrained_keys_vit
Updated 2 years ago
Performance using the cosine distance
Updated 2 years ago
Why there is "no_weight_decay" function for Swin-T but not for VIT
Updated 2 years ago
Could you please release the pre-trained R50 model?
Updated 2 years ago
For cnn architecture like resnet50
Updated 2 years ago
a bug in loading pretrain checkpoint ， in utils.py line 118， you dont use checkpoint in line 111
Closed 2 years ago
can not reproduce your results. trained from your released pre-trained vit-base model
Updated 2 years ago1
SimMIM with Absolute Position Embedding
Updated 2 years ago
Geometric interpolation ViT
Updated 2 years ago1
Plan to implement downstream tasks
Updated 2 years ago
192x192 pretraining resolution
Updated 2 years ago
Inconsistency of ViT-base config as described in the paper
Updated 2 years ago
How can you resolve the mismatch of patch_size in <patch_embed> module between pretrained model and finetuned model?
Closed 2 years ago2
Could you please release the mask prediction visualization code?
Closed 2 years ago1
About data augment
Closed 2 years ago1
Information about relative positional encoding
Closed 2 years ago1
A question about masking strategy
Closed 2 years ago1
USE_RPB and USE_SHARED_RPB are inconsistent in pretrain and finetune
Closed 2 years ago1
Can you release training log for ViT 800 epochs?
Updated 2 years ago1
When will you release the code for pre-training
Closed 2 years ago1
Could you please release the finetune log of swin and vit?
Updated 2 years ago
Missing hyperparameter
Closed 2 years ago1