keyu-tian/SparK issues and pull requests

#88 - Docker + smaller ConvNeXt-V2 models

Pull Request - State: open - Opened by MarkoHaralovic about 2 months ago

#87 - _cur_active

Issue - State: closed - Opened by LeeHaoRanRan 2 months ago

#86 - dist

Issue - State: closed - Opened by LeeHaoRanRan 3 months ago

#85 - Training Efficiency

Issue - State: open - Opened by LM986 3 months ago

#84 - The versions of mmdet and mmcv?

Issue - State: open - Opened by shuuchen 6 months ago

#83 - Necessity of Mask Tokens

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

#82 - Increasing batch size

Issue - State: open - Opened by knightron0 10 months ago - 1 comment

#81 - ConvNext implementation performance

Issue - State: closed - Opened by ds2268 10 months ago - 4 comments

#80 - SparK ResNet and global feature interaction

Issue - State: open - Opened by csvance 10 months ago - 8 comments

#79 - A Code Issue About “pretrain/main.py”

Issue - State: closed - Opened by lynnz7 10 months ago - 2 comments

#78 - SparseBatchNorm2d can not mask correctly ？

Issue - State: closed - Opened by FengheTan9 10 months ago - 3 comments

#77 - Can SparK be used for few-shot learning?

Issue - State: closed - Opened by Y-T-G 10 months ago - 2 comments

#76 - recommend a great library designed for sparse tensors

Issue - State: closed - Opened by xubin04 10 months ago - 1 comment

#75 - ConvNext B for reconstruct images

Issue - State: closed - Opened by M-SunRise 10 months ago - 3 comments

#74 - How to transfer this method to 3D situation.

Issue - State: closed - Opened by xubin04 10 months ago - 1 comment

#73 - About sparse convolution

Issue - State: closed - Opened by Itsanewday 11 months ago - 4 comments

#72 - Resuming ImageNet fine-tuning

Issue - State: closed - Opened by ds2268 11 months ago - 2 comments

#71 - SparK for semantic segmentation

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

#70 - there is no requirements.txt file.

Issue - State: closed - Opened by mandal4 12 months ago - 1 comment

#69 - ImageNet finetuning exploding

Issue - State: open - Opened by ds2268 12 months ago - 10 comments

#68 - Are there any plans to make a port to tensorflow and Keras?

Issue - State: closed - Opened by noahcjones 12 months ago - 1 comment

#67 - Tutorial for finetune on my own dataset

Issue - State: closed - Opened by WeiCL7777 almost 1 year ago - 1 comment

#66 - reducing pre-training to 200 epochs

Issue - State: closed - Opened by bollossom about 1 year ago - 9 comments

#65 - 对比convnextv2

Issue - State: closed - Opened by leoxxxxxD about 1 year ago - 1 comment

#64 - Target dataset and augmentation

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

#63 - There is no activation after the 2nd Conv in each decoder block

Issue - State: closed - Opened by JiamingJiao about 1 year ago - 1 comment

#62 - infer问题

Issue - State: closed - Opened by Shlre about 1 year ago - 3 comments

#61 - Spark.forward

Issue - State: closed - Opened by ricklisz about 1 year ago - 1 comment

#60 - Can not load the pretrained convnext_small model

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

#59 - 小模型自监督效果

Issue - State: closed - Opened by leoxxxxxD about 1 year ago - 11 comments

#58 - Unexpected key(s) in state_dict: "arch", "module", "is_pretrain".

Issue - State: closed - Opened by Chole0529 about 1 year ago - 2 comments

#57 - 迁移到医学图像分割任务上性能 gap 好大

Issue - State: closed - Opened by David-19940718 about 1 year ago - 2 comments

#56 - does this code work with CNN with groupnorm?

Issue - State: closed - Opened by yxchng about 1 year ago - 1 comment

#55 - Using 3D convolution

Issue - State: closed - Opened by chinmay5 over 1 year ago - 2 comments

#54 - How to deal with the conv in SE module？

Issue - State: closed - Opened by RainFrost1 over 1 year ago - 4 comments

#53 - 想问一下这个loss训练到什么数量级才可以拿去做预训练权重

Issue - State: closed - Opened by heng3366 over 1 year ago - 1 comment

#52 - Finetuning epochs

Issue - State: closed - Opened by ds2268 over 1 year ago - 2 comments

#50 - Self-supervised training time is too long compared to MAE

Issue - State: closed - Opened by HIT-SIRS over 1 year ago - 4 comments

#49 - 稀疏卷积问题

Issue - State: closed - Opened by MENG1996 over 1 year ago - 1 comment

#48 - 训练问题

Issue - State: closed - Opened by Wuqiman over 1 year ago - 1 comment

#47 - Contrastive learning methods performance in paper

Issue - State: closed - Opened by Vickeyhw over 1 year ago - 1 comment

#46 - Modified to 3D network

Issue - State: closed - Opened by XIANYUNYEHE-DEL over 1 year ago - 1 comment

#45 - fatal: not a git repository (or any parent up to mount point /)

Issue - State: closed - Opened by Moreland-cas over 1 year ago - 1 comment

#44 - Pretraining with my own dataset

Issue - State: closed - Opened by alskdjfasdfsadf over 1 year ago - 3 comments

#43 - How do you adjust lr and wd?

Issue - State: closed - Opened by pengcheng93 over 1 year ago - 1 comment

#42 - Pretrained weights on ResNet18?

Issue - State: open - Opened by shuuchen over 1 year ago - 1 comment
Labels: todo/gpu-needed

#41 - About reproducing MAE with ViT-S

Issue - State: closed - Opened by LinB203 over 1 year ago - 6 comments

#40 - How do I know it is pretraining works instead of longer finetuning epochs?

Issue - State: closed - Opened by rayleizhu over 1 year ago - 5 comments

#39 - How to install Tap?

Issue - State: closed - Opened by LinB203 over 1 year ago - 1 comment

#38 - A question about data leakage.

Issue - State: closed - Opened by r1cheu over 1 year ago - 5 comments

#37 - will spark work well on depthwise conv?

Issue - State: closed - Opened by LinB203 over 1 year ago - 1 comment

#36 - why pretraining on 384x384 needs larger learning rate?

Issue - State: closed - Opened by junwuzhang19 over 1 year ago - 1 comment

#35 - how to finetune resnet in my own dataset?

Issue - State: closed - Opened by merlinxu-zz over 1 year ago - 6 comments

#34 - error in downstream_imagenet

Issue - State: closed - Opened by alskdjfasdfsadf over 1 year ago - 1 comment

#33 - downstream_imagenet fine-tuning question

Issue - State: closed - Opened by alskdjfasdfsadf over 1 year ago - 1 comment

#32 - What should i do when input shape is a rectangle ?

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

#31 - how to convert sparse model to dense model?

Issue - State: closed - Opened by Mike575 over 1 year ago - 2 comments

#30 - Optimized sparse conv from open3d

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

#29 - Can you provide more model pth with decoder?

Issue - State: closed - Opened by WwwwYz666 over 1 year ago - 2 comments

#28 - The finetuning weights on imagenet

Issue - State: open - Opened by YuetongFang over 1 year ago - 1 comment
Labels: todo/gpu-needed

#27 - The finetuning hyperparameters of resnet50

Issue - State: open - Opened by Vickeyhw over 1 year ago - 7 comments
Labels: todo/gpu-needed

#26 - cosine learning rate annealing

Issue - State: closed - Opened by shuuchen over 1 year ago - 1 comment

#25 - a small model such as Mobilenet v2 for pre-training

Issue - State: open - Opened by mmmz28 over 1 year ago - 3 comments
Labels: todo/gpu-needed

#24 - Will pretrained models converge faster on downstream tasks than non pretrained models?

Issue - State: closed - Opened by xylcbd over 1 year ago - 1 comment

#23 - The time of training ConvNext-B

Issue - State: closed - Opened by kalelpark over 1 year ago - 1 comment

#22 - Single GPU

Issue - State: closed - Opened by hasayake007 over 1 year ago - 1 comment

#20 - About Datasets

Issue - State: closed - Opened by songbingyue over 1 year ago - 1 comment

#19 - torch.distributed.elastic.multiprocessing.api.SignalException

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

#18 - Weights of Resnet with decoder

Issue - State: closed - Opened by Vickeyhw over 1 year ago - 4 comments

#17 - How to add support for more custom models in pre-training

Issue - State: closed - Opened by zero0kiriyu over 1 year ago - 3 comments

#16 - Is it possible to support the unsquare input

Issue - State: closed - Opened by zero0kiriyu over 1 year ago - 1 comment

#15 - Lr layer decay

Issue - State: closed - Opened by Vickeyhw over 1 year ago - 1 comment

#14 - vis_reconstruction.py question

Issue - State: closed - Opened by buxuewushu1314 over 1 year ago - 6 comments

#13 - 请问下ppt中的动图是怎么画的

Issue - State: closed - Opened by whuhit over 1 year ago - 2 comments
Labels: slides/video animations

#12 - should sp_conv also multipies 1/ratio like the dropout in training mode?

Issue - State: closed - Opened by qianyizhang over 1 year ago - 1 comment

#11 - How to try other mask patch size like 16*16?

Issue - State: closed - Opened by rangek over 1 year ago - 5 comments
Labels: mask patch size

#10 - Performance gain of sparse convolution compared to zero-outing?

Issue - State: closed - Opened by rayleizhu over 1 year ago - 6 comments

#9 - The plan to release the visualization code of image reconstruction and “mask pattern vanishing” issue

Issue - State: closed - Opened by songkq over 1 year ago - 5 comments

#8 - The time of training resnet50

Issue - State: closed - Opened by ModulatedConvolutionalNetworks over 1 year ago - 2 comments
Labels: training speed and memory

GitHub / keyu-tian/SparK issues and pull requests