facebookresearch/simsiam

Issues

• Loss increasing

  #50 opened 10 months ago by karthik1145
  0

• Could not find any class folder

  #49 opened a year ago by chandagrover
  0

• Why are training and testing so slow

  #48 opened a year ago by InstantWindy
  2

• Slow convergence with SGD linear evaluation

  #37 opened 2 years ago by gergopool
  1

• which vector is used in experiment (the vector atfer predictor or after projection)?

  #47 opened a year ago by aruba01
  0

• self.encoder.fc in builder.py

  #46 opened a year ago by MSutharsan
  0

• [Problem] problem occured when trained on custom dataset

  #33 opened 2 years ago by CaptainEven
  4

• l2 normalize

  #17 opened 3 years ago by UQ-DongYuan
  2

• About the shape of image[0] and image[1]

  #42 opened 2 years ago by Richard-Lu-badbird
  1

• The value of loss function nn.CosineSimilarity is negative

  #44 opened 2 years ago by zhang1hongliang
  1

• Pre-trained weights for SimCLR, MoCov2, BYOL, SwaV

  #43 opened 2 years ago by ninatu
  1

• Reproducing CIFAR-10 Results

  #9 opened 3 years ago by ardywibowo
  14

• Loss collapse during training

  #40 opened 2 years ago by aisagarw
  1

• A question about projection and prediction MLP

  #6 opened 3 years ago by CaptainPrice12
  3

• About the projection and prediction head dimension configs

  #39 opened 2 years ago by Dylan-H-Wang
  0

• loss question

  #29 opened 2 years ago by daeing
  4

• SyncBatchnorm usage in main_lincls.py

  #38 opened 2 years ago by myoh97
  2

• checkpoint for epoch 800

  #36 opened 2 years ago by elias-ramzi
  0

• larger batch size with linear scale does not work

  #35 opened 2 years ago by liming-ai
  0

• AssertionError: assert set(msg.missing_keys) == {"fc.weight", "fc.bias"}

  #8 opened 3 years ago by etetteh
  3

• Stuck in the DDP initiliazation step

  #16 opened 2 years ago by kxgong
  2

• difference between byol in table19 and simsiam?

  #34 opened 2 years ago by LMerCy
  1

• Single GPU Training

  #30 opened 2 years ago by nixczhou
  3

• Loss descent slowly

  #31 opened 2 years ago by vincezengqiang
  1

• Checkpoints with 50 epochs

  #32 opened 2 years ago by kevinbro96
  1

• Checkpoint with 200 epochs?

  #18 opened 3 years ago by lzyhha
  7

• zero_init_residual

  #28 opened 2 years ago by ifountalis
  1

• Problems of reproducing the transfering results with SimSiam/MoCo-v2 on COCO

  #26 opened 2 years ago by thss15fyt
  1

• General question regarding weight sharing and stop gradient

  #27 opened 3 years ago by ramchandracheke
  2

• Question about the calculated details of std in the paper

  #25 opened 3 years ago by qxm1995716
  2

• Questions about the optimal model for object detection

  #24 opened 3 years ago by bluenights
  2

• Failing on Reproducing ImageNet Linear Classification Results with SGD

  #23 opened 3 years ago by ferreirafabio
  6

• Checkpoints for COCO and VOC

  #19 opened 3 years ago by alexlioralexli
  1

• Learning rate scheduler setting for different epochs

  #22 opened 3 years ago by BaohaoLiao
  1

• Negative Loss

  #21 opened 3 years ago by say2sarwar
  1

• What's the point if we do not gather all outputs in different GPUs to compute contrastive loss

  #20 opened 3 years ago by BaohaoLiao
  3

• Open-source improved implementations of BYOL, SWAV, etc.?

  #5 opened 3 years ago by DrJimFan
  3

• SyncBatchNorm

  #12 opened 3 years ago by dengandong
  5

• issue about the cosine similarity loss function

  #14 opened 3 years ago by sonnefred
  1

• Question about batch size

  #15 opened 3 years ago by CoinCheung
  0

• Question about the learning rate used in linear evaluation

  #13 opened 3 years ago by Vickeyhw
  1

• Fixed LR on predictor with LR warmup

  #11 opened 3 years ago by patrickpoirson
  1

• kNN evaluation during training

  #10 opened 3 years ago by elias-ramzi
  2

• Why BYOL without the momentum encoder collapses but SimSiam does not?

  #7 opened 3 years ago by zhujiagang
  1

• teaser

  #2 opened 3 years ago by endernewton
  0

• Teaser

  #1 opened 3 years ago by endernewton
  0