HobbitLong RepDistiller issues

HobbitLong / RepDistiller

[ICLR 2020] Contrastive Representation Distillation (CRD), and benchmark of recent knowledge distillation methods

BSD 2-Clause "Simplified" License

2.11k stars 388 forks source link

issues

Newest

Newest Most commented Recently updated Oldest Least commented Least recently updated

With tensorboard try

#60 geetHonve closed 9 months ago
1
No dev set split

#59 guzy0324 closed 9 months ago
0
How to use myself datasets?

#58 1997Jessie opened 1 year ago
1
ERROR :run ./fetch_pretrained_teachers.sh

#57 DPingWu opened 1 year ago
1
the result is different in resnet56

#56 SuWideSun opened 1 year ago
1
Compatibility for `torch==1.12.1`

#55 sieu-n closed 8 months ago
2
Is Ensemble distillation also included?

#54 YanjingLiLi closed 1 year ago
0
Hyperparameter Settings for KD on Imagenet

#53 Calmepro777 opened 1 year ago
0
Why using log_softmax instead of softmax?

#52 nguyenvulong opened 2 years ago
1
Question about normalization constant Z_v1 and Z_v2 in the ContrastMemory

#51 YujieZheng99 opened 2 years ago
0
Ensemble Task Implementation

#50 sdsawtelle opened 2 years ago
2
crd used in image enhancement task like Denoise\SR\Deblur.

#49 YangGangZhiQi opened 2 years ago
0
about using the resnet models for cifar10

#48 EmnaGuermazi97 opened 2 years ago
1
Failed to download the teacher models

#47 Prisoneryc opened 2 years ago
2
resnet structure seems to be a bit wrong

#46 surprisedong opened 2 years ago
3
Problem of the order of the normalization in Similarity-Preserving loss.

#45 seacj opened 2 years ago
0
Training scheme for linear probe on STL10 and TinyImagenet

#44 4m4n5 opened 2 years ago
0
Error while running the code

#43 frestuc closed 2 years ago
0
Has anyone implemented Wasserstein Contrastive Representation Distillation

#42 Xinxinatg opened 2 years ago
1
Why "opt.nce_k" in dataset cifar100 is 16384? How can I get this ?

#41 MuHeDing opened 3 years ago
2
Question on memory consumption for CRD loss when the dataset is very large

#40 TMaysGGS opened 3 years ago
3
Cross modal KD implementation release?

#39 liu115 opened 3 years ago
1
AttributeError: 'CIFAR100InstanceSample' object has no attribute 'train_data'

#38 Jiawen-huang opened 3 years ago
3
what is the difference between the position of putting "with torch.no_grad()"

#37 ChriswooTalent opened 3 years ago
1
Hyper-parameters for reproducing the results on ImageNet

#36 kumamonatseu opened 3 years ago
9
how to train my model?

#35 972461099 opened 3 years ago
0
About deep mutual learning setting

#34 swlzq opened 3 years ago
0
About the CE loss

#33 XiXiRuPan opened 3 years ago
0
ImageNet results

#32 senya-ashukha opened 3 years ago
0
How to train teacher model

#31 tiancity-NJU opened 3 years ago
0
teacher model is too big to run with batch_size 64

#30 tiancity-NJU opened 3 years ago
0
Question about pretrained teacher model

#29 MaorunZhang closed 3 years ago
0
Multiple GPU training

#28 deropty closed 3 years ago
0
the introduction of ContrastMemory

#27 sanshanxiashi opened 3 years ago
1
KD method in both configurations seems to be doing better than all other methods except the one from your paper

#26 ksachdeva closed 3 years ago
1
hyperparameters for other methods

#25 wukailu opened 3 years ago
1
questions about ContrastMemory

#24 jianxiangm opened 4 years ago
3
AttributeError: 'CIFAR100Instance' object has no attribute 'train_data'

#23 Yejing-Lai closed 4 years ago
0
How can I use CRD_loss to face landmark detetct for model compression? There is no "opt.nce_k: number of negatives paired with each positive".

#22 gjd2017 opened 4 years ago
0
The calculation of correlation matrix

#21 winycg opened 4 years ago
0
How do you choose the optimal hyper-parameters?

#20 JinYang88 opened 4 years ago
2
Does the crd can be applied to cross domain distillation

#19 Doraemonzm closed 4 years ago
2
The sampler is not consistent with the original implementation of CCKD

#18 winycg closed 4 years ago
1
code for ensemble distillation

#17 tonmoy-saikia closed 4 years ago
1
Reported results based on early stopping?

#16 VladimirLi closed 4 years ago
3
the implementation of cckd is not consistent with the paper

#15 xiaojieli0903 closed 4 years ago
2
Add backward compat properties to CIFAR100 dataset classes

#14 ml-illustrated opened 4 years ago
2
Form of the h function for infinite dataset

#13 brotherofken closed 4 years ago
2
Memory issue about the NST LOSS

#12 leoozy closed 4 years ago
2
Questions about ContrastMemory

#11 HelloTobe closed 4 years ago
5