Ecosyste.ms: Issues

An open API service for providing issue and pull request metadata for open source projects.

GitHub / HobbitLong/SupContrast issues and pull requests

#150 - Solving Loss Saturation

Issue - State: open - Opened by avernal1819 15 days ago

#149 - about loss_in and loss_out in paper

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

#148 - How to change backbone

Issue - State: open - Opened by lucasPV 7 months ago - 1 comment

#147 - solved

Issue - State: closed - Opened by limengting-evelyn 8 months ago

#146 - Code for using the ImageNet pretrained model

Issue - State: open - Opened by LightingMc 8 months ago - 4 comments

#145 - Moco trick?

Issue - State: open - Opened by LightingMc 8 months ago

#143 - Pre-trained model

Issue - State: open - Opened by tanushreebanerjee 8 months ago - 1 comment

#142 - fix: nan loss values

Pull Request - State: closed - Opened by srsawant34 10 months ago - 6 comments

#141 - NaN loss for no positive pair from SupCon

Issue - State: closed - Opened by srsawant34 10 months ago

#140 - Feature/pre commit

Pull Request - State: open - Opened by seanswyi 11 months ago

#139 - Mvp

Pull Request - State: closed - Opened by NinaLong2077 11 months ago - 1 comment

#138 - .

Pull Request - State: closed - Opened by NinaLong2077 11 months ago

#137 - Purpose of logits_mask?

Issue - State: closed - Opened by adv010 11 months ago - 3 comments

#136 - CE vs CE + SupCon

Issue - State: closed - Opened by qsunyuan about 1 year ago - 2 comments

#135 - loss computation: mean and not sum

Issue - State: open - Opened by CassNot about 1 year ago - 4 comments

#134 - Execution of code for ImageNet code (Moco trick)

Issue - State: open - Opened by Angelina1996 about 1 year ago - 1 comment

#132 - Is this using resnet50x4 width?

Issue - State: closed - Opened by hdeng26 over 1 year ago - 2 comments

#131 - Question about potentially incomplete usage of 'logits_mask'

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

#130 - Local pipeline

Pull Request - State: closed - Opened by LukasMahieu over 1 year ago - 1 comment

#128 - the mean and std of cifar dataset

Issue - State: open - Opened by shuaiNJU almost 2 years ago

#127 - Loss function is not convergent when batch-sizes smaller？

Issue - State: open - Opened by kyre-99 almost 2 years ago - 4 comments

#126 - 3673 adding image reducer

Pull Request - State: closed - Opened by chrishickey almost 2 years ago

#125 - fix comments!

Pull Request - State: closed - Opened by SlowMonk almost 2 years ago

#124 - Why use two anchors?

Issue - State: open - Opened by qiaoyu1002 almost 2 years ago - 1 comment

#123 - The problem about the number of positives in code

Issue - State: open - Opened by shuaiNJU almost 2 years ago

#122 - Supervised Contrastive Learning with n_views=1

Issue - State: open - Opened by piconti almost 2 years ago - 1 comment

#121 - Using SupContrast loss for non-image application

Issue - State: open - Opened by mossishahi almost 2 years ago - 1 comment

#120 - Hyperparameters for ImageNet training?

Issue - State: open - Opened by InugYoon almost 2 years ago - 4 comments

#119 - Run custom dataset

Issue - State: closed - Opened by geokarant almost 2 years ago

#118 - Can anyone reproduce this result or is it just me?

Issue - State: closed - Opened by isacyoo about 2 years ago - 1 comment

#117 - MoCo code

Issue - State: open - Opened by carloalbertobarbano about 2 years ago - 1 comment

#116 - About the loss. Sincerely, I would like to ask：

Issue - State: open - Opened by Struggle-Forever about 2 years ago - 5 comments

#115 - Provided pre-trained load issue

Issue - State: open - Opened by Rasoul77 about 2 years ago - 3 comments

#114 - compare

Pull Request - State: closed - Opened by shovelingpig over 2 years ago

#113 - About train resent-18 on the dataset CIFAR-100, the accuracy is just 55.

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

#112 - f1, f2 = torch.split(features, [bsz, bsz], dim=0)

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

#111 - fix a bug, which has the little probability of producing nan in the loss

Pull Request - State: open - Opened by yuanlonghui over 2 years ago - 2 comments

#110 - Question about logits_mask

Issue - State: open - Opened by Alva-2020 over 2 years ago - 5 comments

#109 - Another bug?

Issue - State: open - Opened by W2Q3Q1 over 2 years ago

#108 - May be a bug?

Issue - State: open - Opened by W2Q3Q1 over 2 years ago - 1 comment

#107 - Fixes the usage of SimCLR loss in README.md

Pull Request - State: open - Opened by vaydemir3 over 2 years ago

#106 - SimCLR Loss: False Output (Possibly with Correction)

Issue - State: open - Opened by vaydemir3 over 2 years ago - 2 comments

#105 - The implementation using the trick of the momentum encoder

Issue - State: open - Opened by yinguoweiOvO over 2 years ago - 2 comments

#104 - Why the func returns "nan"?

Issue - State: open - Opened by Zhudongsheng75 over 2 years ago - 8 comments

#103 - Meaning of joint training

Issue - State: open - Opened by htn274 over 2 years ago - 3 comments

#102 - about loss function

Issue - State: open - Opened by tommying over 2 years ago - 3 comments

#101 - Why use logits without exp() and log() - torch.log(exp_logits)

Issue - State: open - Opened by Zhen-Zohn-WANG over 2 years ago - 3 comments

#100 - output Nan loss

Issue - State: closed - Opened by ChuangLee over 2 years ago - 1 comment

#99 - Why main_supcon

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

#98 - Understanding the SupContrast loss

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

#97 - Two different Data augmentations?

Issue - State: open - Opened by priyarana almost 3 years ago - 7 comments

#96 - About Training

Issue - State: closed - Opened by xianxuan-z almost 3 years ago - 2 comments

#95 - About args n_views

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

#94 - Release of pre-trained model on C-10/100 and ImageNet

Issue - State: closed - Opened by sunyiyou almost 3 years ago - 2 comments

#93 - Confusion about training and validation when pretraining

Issue - State: closed - Opened by XiHuYan almost 3 years ago - 1 comment

#92 - negative distance

Issue - State: open - Opened by mese79 about 3 years ago - 1 comment

#91 - Update losses.py

Pull Request - State: closed - Opened by YJiangcm about 3 years ago

#90 - can supCon loss used in multi-label classification?

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

#89 - How do we add contrastive accuracy to the SupCon

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

#88 - Embedding dimension proportional to capacity or depth?

Issue - State: open - Opened by arjung128 about 3 years ago

#87 - SupConLoss on non-image data (no augmentation)

Issue - State: open - Opened by mese79 about 3 years ago - 1 comment

#86 - Update losses.py

Pull Request - State: open - Opened by zhangzhenyu13 about 3 years ago - 3 comments

#85 - where is Self-Supervised Contrastive Loss?

Issue - State: closed - Opened by seyeeet about 3 years ago

#84 - Importance of embedding dimensions

Issue - State: open - Opened by shaunakjoshi12 about 3 years ago - 3 comments

#83 - Non-decreasing loss for custom dataset

Issue - State: open - Opened by Bedrettin-Cetinkaya about 3 years ago - 10 comments

#82 - Is the entire dataset used for pre-training?

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

#81 - How necessary is apex? install errors

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

#80 - # of Batch

Issue - State: closed - Opened by wjun0830 over 3 years ago - 2 comments

#79 - Pass batch size and workers to val loader

Pull Request - State: open - Opened by FrankXinqi over 3 years ago

#78 - About Custom dataset

Issue - State: open - Opened by LDZ233-ldz over 3 years ago

#77 - Make network weights available

Issue - State: closed - Opened by jnboehm over 3 years ago

#76 - About --ckpt /path/to/model.pth

Issue - State: open - Opened by Lin-xuejuan-an over 3 years ago - 2 comments

#75 - About Custom dataset

Issue - State: closed - Opened by LDZ233 over 3 years ago - 1 comment

#74 - Input image resolution for ImageNet

Issue - State: closed - Opened by littleredxh over 3 years ago - 9 comments

#73 - losses related problem

Issue - State: open - Opened by yypurpose over 3 years ago - 1 comment

#72 - batch=32

Issue - State: open - Opened by LDZ233-ldz over 3 years ago - 2 comments

#71 - running main_ce.py does not work

Issue - State: closed - Opened by gr986 over 3 years ago - 2 comments

#70 - RuntimeError: cuDNN error: CUDNN_STATUS_NOT_INITIALIZED

Issue - State: closed - Opened by LDZ233-ldz over 3 years ago

#69 - Training on custom dataset

Issue - State: open - Opened by himansh1314 over 3 years ago - 3 comments

#68 - opened by mistake - please ignore

Issue - State: closed - Opened by prakharguptaz over 3 years ago

#67 - An error in main_supcon.py

Issue - State: closed - Opened by holdfire over 3 years ago - 2 comments

#66 - About multi-crop strategy.

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

#65 - Regarding the "for numerical stability" block

Issue - State: open - Opened by AlexMoreo over 3 years ago - 4 comments

#64 - Question about SupConLoss implementation

Issue - State: open - Opened by vkhoi over 3 years ago - 11 comments

#63 - Training on CIFAR-10 takes a lot of GPU memory space

Issue - State: open - Opened by jainpulkit54 over 3 years ago - 1 comment

#62 - What is "contrast_mode" in loss?

Issue - State: open - Opened by omg777 over 3 years ago - 2 comments

#61 - loss_out >= loss_in

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

#60 - The code of visualization

Issue - State: open - Opened by echobear313 over 3 years ago

#59 - Result by using other augmentations(e.g. RandAug)

Issue - State: open - Opened by sooonwoo over 3 years ago - 4 comments

#58 - losses.py

Issue - State: open - Opened by yangyt-2020 over 3 years ago - 1 comment

#57 - question about line:141 in main_supcon.py(custom dataset std)

Issue - State: open - Opened by cucutone over 3 years ago - 2 comments

#56 - projection head

Issue - State: closed - Opened by arjung128 over 3 years ago - 1 comment

#55 - T-SNE visualization

Issue - State: open - Opened by munirfarzeen over 3 years ago - 3 comments

#54 - ImageNet

Issue - State: open - Opened by arjung128 over 3 years ago

#53 - When do I stop training?

Issue - State: open - Opened by Taeuk-Jang over 3 years ago

#52 - Performance problem

Issue - State: open - Opened by zkcys001 almost 4 years ago - 2 comments