DingXiaoH/RepVGG

Issues

• question about the newest work?

  #122 opened 3 months ago by CoinCheung
  0

• 请问RepVGGPlus中的辅助分类器应该如何使用？

  #121 opened 3 months ago by BUG423
  0

• 预训练模型是在IN-1k上的吗，有22k上的预训练模型吗？

  #120 opened 5 months ago by McCarthyjfq
  1

• a mistake in code

  #110 opened 6 months ago by Halance
  1

• Why would you use 3x3+1x1+identity instead of 3x3+3x3+identity?

  #119 opened a year ago by hopef
  0

• Running example_pspnet.py generates an error: raise NotImplementedError

  #118 opened a year ago by fhf526
  1

• The diff is large when BN use other initialization method in verify script

  #117 opened a year ago by 2h4dl
  0

• Question about training and inference models

  #116 opened a year ago by hosytuyen
  0

• 为什么SEBlock中使用Conv2D而不是Linear层？

  #115 opened a year ago by Z-Xiong
  1

• 多分支存在非线性的合并

  #113 opened a year ago by piaohe20221128
  1

• 关于量化的问题，一般是对重参数化后的模型，再进一步进行量化的吗

  #114 opened a year ago by LKAMING97
  0

• Question about RepVGG usage for downstream task

  #112 opened a year ago by imdsafi09
  0

• 为什么RepVgg的单个block拿出来测试，重参数化前后输出的结果是不一样的

  #111 opened 2 years ago by lokiwizard
  0

• jit.trace error

  #109 opened 2 years ago by nistarlwc
  0

• 关于多卡训练卡死的问题

  #108 opened 2 years ago by SEUvictor
  1

• anyone use repvgg backbone in insightface face recognition task?

  #76 opened 3 years ago by xiakj
  1

• Paper Question - Why less favored than MobileNets for low-powered devices?

  #101 opened 2 years ago by ksachdeva
  1

• The bug of QAT

  #107 opened 2 years ago by qhy991
  0

• Training script inquiry

  #106 opened 2 years ago by zw615
  0

• why get different result after reparam

  #100 opened 2 years ago by fenghuoxiguozu
  3

• Question about multi-branch and single-branch network

  #105 opened 2 years ago by MiaoJieF
  1

• SSD with RepVGG backbone

  #83 opened 3 years ago by thuangb
  2

• subprocess.CalledProcessError: Command '['/usr/bin/python3', '-u', 'main.py', '--local_rank=0']' returned non-zero exit status 1.

  #104 opened 2 years ago by starsky68
  0

• 请问如果我用RepVGG作为backbone，并且在neck中有其他的操作，比如5*5卷积，这样训练出来的模型还能做结构重参数化吗？

  #103 opened 2 years ago by AssassinJay
  1

• how to prune repvgg

  #102 opened 2 years ago by sinoaidi
  0

• eq_kernel in get_custom_L2 missing contribution of rbr_identity.

  #99 opened 2 years ago by vchiley
  0

• MegEngine Implemention of RepVGG

  #84 opened 2 years ago by Asthestarsfalll
  0

• question about insert BN before QAT

  #98 opened 2 years ago by zhaoxin111
  0

• How is 1x1 conv with identity matrix as kernel equal to Identity layer

  #97 opened 2 years ago by prakashjayy
  1

• Where is the re-parameter module?

  #74 opened 3 years ago by typhoonlee
  2

• mmsegmentation

  #96 opened 2 years ago by endaoguansanlu
  0

• convert.py

  #95 opened 2 years ago by HonestyBrave
  0

• CIFAR10训练的参数？

  #94 opened 2 years ago by wooozihui
  0

• Re-params quantized RepVGG model

  #91 opened 3 years ago by thancaocuong
  2

• 关于repconv switch-to-deploy的问题

  #93 opened 2 years ago by pierrelzw
  0

• what matter does "get_custom_L2" do?

  #92 opened 3 years ago by shoutOutYangJie
  1

• Training with another dataset?

  #81 opened 3 years ago by Enescigdem
  1

• 3x3 convolution

  #87 opened 3 years ago by SpiritAshes
  1

• Paper Question: Use compression idea and generate same result as multi-branch model?

  #89 opened 3 years ago by oconnor127
  1

• Any reference about get_custom_L2?

  #90 opened 3 years ago by JoJoistheBestOne
  1

• Some questions about quantization

  #85 opened 3 years ago by Desmond-97
  9

• Update Arxiv Paper with RepVGGplus

  #75 opened 3 years ago by Zacharias030
  2

• why not use python3 -m torch.distributed.launch

  #88 opened 3 years ago by xiangyun-hz
  0

• Hyperparameters for RepVGGplus-L2pse

  #86 opened 3 years ago by zhreshold
  0

• How to understand the Structural re-parameterization of a RepVGG block

  #78 opened 3 years ago by jyang68sh
  2

• PTQ精度问题

  #79 opened 3 years ago by huoshuai-dot
  1

• Given input size: (64x240x320). Calculated output size: (64x0x1). Output size is too small

  #77 opened 3 years ago by DonMuv
  1

• for style transfer task training

  #82 opened 3 years ago by MaltoseMai
  0

• questions about training on custom datasets？

  #80 opened 3 years ago by Hezhexi2002
  0

• Weight decay will be applied to the weight of BN in identity branch and the weight of FC when using --custwd?

  #73 opened 3 years ago by wangjinwei94
  2