facebookresearch/deit

Issues

• ViT-B Training for DeiT

  #233 opened 10 months ago by ziqipang
  2

• Will you be releasing the accuracy of the official deit III framework trained tiny version on IN1k?

  #241 opened a year ago by chenziwenhaoshuai
  0

• DeiT depth 24 (CaiT - TABLE 1)

  #218 opened a year ago by GoJunHyeong
  2

• Gradient accumulation code

  #240 opened a year ago by King4819
  0

• Question about different seeds per gpu with DDP

  #239 opened a year ago by HIT-LiuChen
  0

• Training

  #238 opened a year ago by ali-88123
  0

• Inclusion of Transformers Need Registers

  #237 opened a year ago by mileseverett
  0

• Slow Training

  #234 opened a year ago by mueller-mp
  2

• random.seed(seed) in line 205 is commented

  #236 opened a year ago by Phuoc-Hoan-Le
  0

• Checkpoints of IN21K pretrained deit III

  #232 opened a year ago by Byakuya-zi
  0

• Hi，Why can't I find deit_tiny_distilled_patch16_224 in hubconf

  #231 opened a year ago by GerogeD
  0

• TracerWarning

  #230 opened a year ago by maingoc1605
  0

• batch_size flag

  #220 opened 2 years ago by tsengalb99
  2

• how to implement document layout analysis use Deit-B

  #210 opened 2 years ago by sherryhsy
  2

• How to launch a training of CAIT models ?

  #226 opened 2 years ago by elias-ramzi
  0

• Code for cosub

  #224 opened 2 years ago by ppalantir
  0

• The ablation experiment of DeiT

  #215 opened 2 years ago by Berry-Wu
  2

• ImageNet21K data preparation for pre-training

  #219 opened 2 years ago by mxjecho
  5

• Meaning of the model name ( ResMLP)

  #207 opened 2 years ago by YHYeooooong
  1

• Can I use timm==0.4.12 instead of timm==0.3.2 ?

  #206 opened 2 years ago by irhallac
  1

• unexpected keyword argument 'pretrained_cfg'

  #212 opened 2 years ago by entron
  2

• Are the hyperparameters for DeiT-T and for DeiT-S any different than DeiT-B?

  #201 opened 2 years ago by Phuoc-Hoan-Le
  1

• ImageNet21k pretrained model without finetuning on 1k

  #181 opened 2 years ago by bhheo
  2

• How long is it supposed to take to train on ImageNet21k for 90 epochs with 8 V100 GPUs

  #198 opened 2 years ago by Phuoc-Hoan-Le
  1

• number of classes

  #197 opened 2 years ago by Ye-Na-Kim
  1

• What's the accuracy of deit-S without pre-trained on CIFAR10

  #202 opened 2 years ago by hanwenran1
  1

• how to implement cosub training use deit-III

  #217 opened 2 years ago by xiaoguang-1
  2

• What are the hyperparameters for DeiT-III (epoch 400 or 600)?

  #214 opened 2 years ago by GoJunHyeong
  0

• how to implement cosub training use deit-III

  #216 opened 2 years ago by xiaoguang-1
  0

• Single machine multi-GPU training

  #213 opened 2 years ago by AlexNmSED
  0

• Multi-node support

  #208 opened 2 years ago by Phuoc-Hoan-Le
  0

• Multinode Slurm Training

  #204 opened 2 years ago by yazdanimehdi
  0

• What batch size number other than 1024 have been tried when training a DeiT model?

  #205 opened 2 years ago by Phuoc-Hoan-Le
  0

• Does the EMA is used in DeiT-III?

  #203 opened 2 years ago by mzr1996
  3

• Question about Throughput

  #189 opened 2 years ago by techmonsterwang
  1

• cifar100 pretrain model?

  #191 opened 2 years ago by Wang-Y-S
  1

• The concatenation of 'cls_tokens' and 'patch_embedding' is not necceassay.

  #188 opened 2 years ago by thb1314
  1

• What is the difference between class attention in the paper CaiT and traditional multi-headed self-attention？

  #196 opened 2 years ago by hutingz
  1

• What is the ImageNet-1K Top-1 accuracy of Training from 0 to 400 epochs (Fig. 5 of Deit III paper)

  #199 opened 2 years ago by sanyalsunny111
  0

• Config file of ViT-B/16

  #195 opened 2 years ago by shashankvkt
  2

• Object detection from the DeiT III pretrained model

  #187 opened 2 years ago by YuanLiuuuuuu
  2

• Uneven memory usage among GPUs with DistributedDataParallel

  #194 opened 2 years ago by Phuoc-Hoan-Le
  0

• Is it possible if I can see how the validation accuracy changes over the number of epochs for DeiT?

  #193 opened 2 years ago by Phuoc-Hoan-Le
  0

• Is "unscale-lr" used in DeiT training on ImageNet1k

  #192 opened 2 years ago by Phuoc-Hoan-Le
  0

• Reproduce PatchConvnet

  #185 opened 2 years ago by billpsomas
  2

• LAMB and amp

  #182 opened 2 years ago by sgunasekar
  1

• Question about training DeiT-small distilled

  #186 opened 2 years ago by mingqiJ
  2

• Is uniform drop-path rates beneficial ?

  #184 opened 2 years ago by bhheo
  2

• DeiT-tiny pth size is not 5M, it is 22M

  #183 opened 2 years ago by witding
  1

• Confusion about fine-tuning

  #180 opened 2 years ago by WangWenhao0716
  4