SHI-Labs/Neighborhood-Attention-Transformer

Issues

• Question: What are some recommended better alternatives for RBP ?

  #105 opened 7 months ago by fzimmermann89
  6

• Alternatives to Relative Positional Biases?

  #104 opened 5 months ago by nikhilmishra000
  2

• Rule of thumb for dilations value?

  #103 opened 6 months ago by rafiibnsultan
  1

• Some comparisons against Deformable Attention

  #99 opened 7 months ago by haiphamcse
  4

• Can you release your training log of NAT? I mean, the summary.csv in output folder.

  #100 opened 9 months ago by Yonghao-Yu
  3

• instance segmentation mask2former + dinat

  #97 opened 7 months ago by AM97Y
  1

• How to visualize the attention map?

  #77 opened 7 months ago by Amo5
  3

• For 3D segmentation

  #96 opened 7 months ago by LambdaLi
  2

• Issue with Neighborhood Attention Model (NAT) Pretrained Weights

  #101 opened 7 months ago by Miiitiii
  1

• NAT Tiny performance on ImageNet 1k

  #91 opened a year ago by jamesben6688
  7

• training from scratch with different size for height and width

  #92 opened a year ago by mr17m
  3

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

  #93 opened a year ago by Linwei-Chen
  12

• How to calculate the number of params?

  #95 opened a year ago by HanzhouLiu
  1

• mmdetection on COCO2017 not converge

  #94 opened a year ago by jamesben6688
  1

• Is it possible to do upsampling using NAT ?

  #81 opened a year ago by jimmysue
  2

• Is dectect model available?

  #88 opened a year ago by chenzx2
  2

• freeze_at be set to 2 to freeze the pretrained weight downloaded from the official website?

  #89 opened a year ago by aihcyllop
  2

• About the receptive field of image pixel

  #90 opened a year ago by money6651626
  4

• Is DiNAT code is runnable?

  #86 opened 2 years ago by chhkang
  2

• Running NAT require Removing Visibility for Old (sm_52 or earlier) GPUs

  #48 opened 2 years ago by gauenk
  8

• Welcome update to OpenMMLab 2.0

  #78 opened 2 years ago by vansin
  1

• some problem during train

  #85 opened 2 years ago by guoguo1314
  9

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

  #83 opened 2 years ago by aihcyllop
  1

• Where is natten.py

  #82 opened 2 years ago by SantaFlang
  0

• ONNX

  #74 opened 2 years ago by idanpd
  2

• Does forward always use fp16?

  #52 opened 2 years ago by rayleizhu
  5

• While ruuning the code, I got this types of problem. Could you please tell me the solution

  #69 opened 2 years ago by Mehulk43
  11

• Legacy Torch implementation for Dilated Neighborhood Attention (DiNAT)

  #59 opened 2 years ago by iliaschalkidis
  5

• Motivation on choosing NAT depth

  #67 opened 2 years ago by oksanadanilova
  2

• cpp version of nattenav layer

  #36 opened 2 years ago by AntonyDrovalov
  3

• Relation to visual attention network (VAN).

  #60 opened 2 years ago by MenghaoGuo
  4

• A question about the rpb in LegacyNeighborhoodAttention2D

  #61 opened 2 years ago by lartpang
  5

• Compile problem

  #63 opened 2 years ago by yuhaoliu7456
  2

• abbreviation for rpb

  #58 opened 2 years ago by qsh-zh
  2

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

  #39 opened 2 years ago by laisimiao
  4

• Is it necessary to write dedicated fp16 kernel ?

  #53 opened 2 years ago by rayleizhu
  8

• pytorch 1.12.0 CUDA 11.6 Win10 VS2019 build error

  #43 opened 2 years ago by Ken1256
  19

• RuntimeError: CUDA error: CUBLAS_STATUS_INVALID_VALUE when calling `cublasSgemm( handle, opa, opb, m, n, k, &alpha, a, lda, b, ldb, &beta, c, ldc)`

  #50 opened 2 years ago by yxchng
  3

• No position encoding? Could you explain some your thoughts?

  #54 opened 2 years ago by laisimiao
  2

• Where is the Neighborhood Attention 2D?

  #49 opened 2 years ago by ady-yu
  2

• failure occured in building at pytorch 1.11.0 / CUDA 11.3 / Win10 / VS2019 error

  #51 opened 2 years ago by helonin
  10

• Tiny Bug in nattencuda.py

  #46 opened 2 years ago by z-jiaming
  1

• how to debug cuda kernel?

  #42 opened 2 years ago by m1710545173
  3

• throughput of nat_tiny vs resnet50

  #38 opened 3 years ago by jiaojile
  4

• PE added on query and key

  #37 opened 3 years ago by XiaoyuShi97
  2

• how to do Visualization

  #40 opened 3 years ago by skyfallsss
  6

• Assertion `idx_dim >= 0 && idx_dim < index_size && "index out of bounds"` failed.

  #41 opened 3 years ago by YangParky
  0

• about the kernel size

  #35 opened 3 years ago by XiaoyuShi97
  3

• The testing results of the whole dataset is empty

  #31 opened 3 years ago by blowhen
  14

• CUDA extension error

  #33 opened 3 years ago by wytcsuch
  4