Ecosyste.ms: Issues

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

GitHub / SHI-Labs/Neighborhood-Attention-Transformer issues and pull requests

#99 - Some comparisons against Deformable Attention

Issue - State: open - Opened by haiphamcse 4 months ago

#98 - Corrected num_heads in DiNAT-s Large

Pull Request - State: closed - Opened by AdityaKane2001 5 months ago - 1 comment

#97 - instance segmentation mask2former + dinat

Issue - State: open - Opened by AM97Y 6 months ago

#96 - For 3D segmentation

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

#95 - How to calculate the number of params?

Issue - State: closed - Opened by HanzhouLiu 7 months ago - 1 comment

#94 - mmdetection on COCO2017 not converge

Issue - State: closed - Opened by jamesben6688 8 months ago - 1 comment

#93 - Cannot repeat the results of Mask2Former+DiNAT-Large on ADE20K

Issue - State: closed - Opened by Linwei-Chen 9 months ago - 12 comments
Labels: help wanted

#92 - training from scratch with different size for height and width

Issue - State: closed - Opened by mr17m 9 months ago - 3 comments
Labels: question

#91 - NAT Tiny performance on ImageNet 1k

Issue - State: closed - Opened by jamesben6688 10 months ago - 7 comments
Labels: question

#90 - About the receptive field of image pixel

Issue - State: closed - Opened by money6651626 11 months ago - 4 comments

#89 - freeze_at be set to 2 to freeze the pretrained weight downloaded from the official website?

Issue - State: closed - Opened by aihcyllop 12 months ago - 2 comments
Labels: question

#88 - Is dectect model available?

Issue - State: closed - Opened by chenzx2 about 1 year ago - 2 comments
Labels: question

#87 - Fix DiNAT_s issues in det/seg

Pull Request - State: closed - Opened by alihassanijr about 1 year ago

#86 - Is DiNAT code is runnable?

Issue - State: closed - Opened by chhkang about 1 year ago - 2 comments
Labels: bug

#85 - some problem during train

Issue - State: closed - Opened by guoguo1314 about 1 year ago - 9 comments

#83 - May I ask whether the code of coco instance segmentation mask2former is dinat or NAT?

Issue - State: closed - Opened by aihcyllop about 1 year ago - 1 comment
Labels: question

#82 - Where is natten.py

Issue - State: closed - Opened by SantaFlang about 1 year ago

#81 - Is it possible to do upsampling using NAT ?

Issue - State: closed - Opened by jimmysue about 1 year ago - 2 comments
Labels: question

#80 - Update NAT bibtex -> CVPR

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

#79 - Fix very minor spelling error

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

#78 - Welcome update to OpenMMLab 2.0

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

#77 - How to visualize the attention map?

Issue - State: open - Opened by Amo5 over 1 year ago - 2 comments
Labels: question

#76 - Update DiNAT chart

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

#75 - Updated DiNAT-L

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

#74 - ONNX

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

#73 - Update README.md

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

#72 - HF announcement + minor fixes.

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

#71 - News Update

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

#70 - New models and other fixes

Pull Request - State: closed - Opened by alihassanijr over 1 year ago
Labels: enhancement

#68 - Fix requirements

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

#67 - Motivation on choosing NAT depth

Issue - State: closed - Opened by oksanadanilova over 1 year ago - 2 comments
Labels: question

#66 - Update checkpoint links

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

#65 - Move NATTEN into a separate repository

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

#64 - Gradcheck fixes

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

#63 - Compile problem

Issue - State: closed - Opened by yuhaoliu7456 over 1 year ago - 2 comments
Labels: bug

#62 - Add `dilations` to DET and SEG

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

#61 - A question about the rpb in LegacyNeighborhoodAttention2D

Issue - State: closed - Opened by lartpang over 1 year ago - 5 comments
Labels: question

#60 - Relation to visual attention network (VAN).

Issue - State: closed - Opened by MenghaoGuo over 1 year ago - 4 comments
Labels: question

#59 - Legacy Torch implementation for Dilated Neighborhood Attention (DiNAT)

Issue - State: closed - Opened by iliaschalkidis over 1 year ago - 5 comments
Labels: enhancement, question

#58 - abbreviation for rpb

Issue - State: closed - Opened by qsh-zh over 1 year ago - 2 comments
Labels: question

#57 - Doc adjustment

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

#56 - DiNAT Checkpoints

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

#55 - DiNAT release

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

#54 - No position encoding? Could you explain some your thoughts?

Issue - State: closed - Opened by laisimiao almost 2 years ago - 2 comments
Labels: question

#53 - Is it necessary to write dedicated fp16 kernel ?

Issue - State: closed - Opened by rayleizhu almost 2 years ago - 8 comments
Labels: question

#52 - Does forward always use fp16?

Issue - State: closed - Opened by rayleizhu almost 2 years ago - 5 comments
Labels: question

#51 - failure occured in building at pytorch 1.11.0 / CUDA 11.3 / Win10 / VS2019 error

Issue - State: closed - Opened by helonin almost 2 years ago - 10 comments
Labels: help wanted

#49 - Where is the Neighborhood Attention 2D?

Issue - State: closed - Opened by ady-yu almost 2 years ago - 2 comments
Labels: documentation

#48 - Running NAT require Removing Visibility for Old (sm_52 or earlier) GPUs

Issue - State: closed - Opened by gauenk almost 2 years ago - 8 comments
Labels: Compilation error

#47 - Minor bug fix and cleanup

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

#46 - Tiny Bug in nattencuda.py

Issue - State: closed - Opened by z-jiaming almost 2 years ago - 1 comment
Labels: bug

#45 - Cleanup

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

#44 - NA CUDA Extension v0.12!

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

#43 - pytorch 1.12.0 CUDA 11.6 Win10 VS2019 build error

Issue - State: closed - Opened by Ken1256 almost 2 years ago - 19 comments
Labels: help wanted

#42 - how to debug cuda kernel?

Issue - State: closed - Opened by m1710545173 about 2 years ago - 3 comments
Labels: question

#40 - how to do Visualization

Issue - State: closed - Opened by skyfallsss about 2 years ago - 6 comments
Labels: help wanted

#39 - Do you try larger resolution when evaluation or pretrained on ImageNet-22k to get higher performance?

Issue - State: closed - Opened by laisimiao about 2 years ago - 4 comments
Labels: question

#38 - throughput of nat_tiny vs resnet50

Issue - State: closed - Opened by jiaojile about 2 years ago - 4 comments
Labels: question

#37 - PE added on query and key

Issue - State: closed - Opened by XiaoyuShi97 about 2 years ago - 2 comments
Labels: question

#36 - cpp version of nattenav layer

Issue - State: closed - Opened by AntonyDrovalov about 2 years ago - 3 comments
Labels: enhancement

#35 - about the kernel size

Issue - State: closed - Opened by XiaoyuShi97 about 2 years ago - 3 comments
Labels: question

#34 - FLOP Counter

Pull Request - State: closed - Opened by alihassanijr about 2 years ago

#33 - CUDA extension error

Issue - State: closed - Opened by wytcsuch about 2 years ago - 4 comments
Labels: help wanted

#32 - NeighborhoodAttention1d

Pull Request - State: closed - Opened by alihassanijr about 2 years ago
Labels: enhancement

#31 - The testing results of the whole dataset is empty

Issue - State: closed - Opened by blowhen about 2 years ago - 14 comments
Labels: help wanted

#30 - Is there an efficient way to apply NAtten to 1D data?

Issue - State: closed - Opened by Guaishou74851 about 2 years ago - 4 comments
Labels: enhancement

#29 - CUDA out of memory

Issue - State: closed - Opened by zhangzheng2242 about 2 years ago - 8 comments
Labels: question

#28 - How to download the pre training of image1k

Issue - State: closed - Opened by zhouyi-git about 2 years ago - 2 comments
Labels: question

#27 - About the size of neighborhood

Issue - State: closed - Opened by wangning7149 about 2 years ago - 4 comments
Labels: question

#26 - Salient map

Issue - State: closed - Opened by Yuuuumie about 2 years ago - 2 comments
Labels: enhancement

#25 - Fix for import errors

Pull Request - State: closed - Opened by alihassanijr about 2 years ago - 2 comments

#24 - Minor error fixed

Pull Request - State: closed - Opened by alihassanijr about 2 years ago - 1 comment

#23 - April 30th update

Pull Request - State: closed - Opened by alihassanijr about 2 years ago

#22 - nattenav_cuda.so: cannot open shared object file: No such file or directory

Issue - State: closed - Opened by julliffbistu about 2 years ago - 12 comments
Labels: help wanted

#21 - Hello and thank you for your interest.

Issue - State: closed - Opened by Wisdom2wisdom about 2 years ago

#20 - hello , I have already installed CUDA on requirement ，why have CUDA extension error ?

Issue - State: closed - Opened by Wisdom2wisdom about 2 years ago - 5 comments
Labels: help wanted

#19 - how to get model object?

Issue - State: closed - Opened by fire717 about 2 years ago - 2 comments
Labels: help wanted

#18 - unable to get the pretrain model for classification

Issue - State: closed - Opened by zhou-rui1 about 2 years ago - 3 comments

#17 - About license

Issue - State: closed - Opened by WangWenhao0716 about 2 years ago - 2 comments

#16 - Any plans to release the Neighborhood Attention code using only Pytorch?

Issue - State: closed - Opened by qwopqwop200 about 2 years ago - 2 comments
Labels: question

#15 - how to build two NAT cuda

Issue - State: closed - Opened by XiaoyuShi97 about 2 years ago - 2 comments
Labels: question

#14 - Valid Padded Behavior

Issue - State: closed - Opened by mgriffin1994 about 2 years ago - 3 comments
Labels: enhancement

#13 - Questions about the algorithm speed.

Issue - State: closed - Opened by jikerWRN about 2 years ago - 2 comments
Labels: documentation, question

#12 - problem about the dim of Q of transformers QKV

Issue - State: closed - Opened by xiaochang129 about 2 years ago - 2 comments
Labels: question

#11 - Question about the dimension of head

Issue - State: closed - Opened by PeiqinZhuang about 2 years ago - 7 comments
Labels: question

#10 - How to train without the cuda extension?

Issue - State: closed - Opened by Geek-lixiang about 2 years ago - 6 comments
Labels: enhancement

#9 - How can I implement a function like joint image filtering?

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

#8 - Details of Training

Issue - State: closed - Opened by achen46 about 2 years ago - 2 comments

#7 - About the detail of NATTENQKRPBFunction and NATTENAVFunction

Issue - State: closed - Opened by myt889 about 2 years ago - 2 comments

#6 - add web demo/model to Huggingface

Issue - State: closed - Opened by AK391 about 2 years ago
Labels: enhancement

#5 - Question on the compilation with different version of cuda or pytorch

Issue - State: closed - Opened by PeiqinZhuang about 2 years ago - 9 comments
Labels: question

#4 - Downsampling using Overlapping vs non-overlapping Convolutions

Issue - State: closed - Opened by mmaaz60 about 2 years ago - 2 comments
Labels: question

#3 - Comparison with zero-padding version.

Issue - State: closed - Opened by weigq about 2 years ago - 3 comments

#2 - When will the code be released?

Issue - State: closed - Opened by linjing7 about 2 years ago - 3 comments

#1 - This seems to be an idea that has been demonstrated by the existing method.

Issue - State: closed - Opened by lartpang about 2 years ago - 6 comments