JiahuiYu/slimmable_networks

Issues

• how to run this code with cpu

  #58 opened 2 years ago by Chenhait
  0

• model profiling

  #37 opened 5 years ago by lyn0102
  3

• How can I evaluate standalone smaller width model?

  #57 opened 2 years ago by abhishekaich27
  1

• the issues of fuction SlimableConv2D

  #56 opened 3 years ago by baiyu12345
  1

• Some questions with the code.

  #51 opened 3 years ago by tfwang08
  0

• No module named 'mmdet.ops.nms.gpu_nms'

  #55 opened 4 years ago by likunnb
  0

• Mobile Implementation

  #54 opened 4 years ago by jsiloto
  0

• Slimmable Network using Mixed Precision

  #53 opened 4 years ago by pprp
  0

• Slimming to lower width results in poor accuracy on ARM architectures

  #52 opened 4 years ago by octavianmm
  0

• BN initialization for tracking performance during training

  #50 opened 4 years ago by MingxiLi
  0

• reproducing CIFAR10 results for AutoSlim

  #40 opened 5 years ago by RudyChin
  8

• AutoSlim search

  #46 opened 4 years ago by HaloTrouvaille
  1

• US-WDSR

  #49 opened 4 years ago by Booooooooooo
  1

• Size mismatch when loading us_mobilenet_v1_calibrated.pt

  #24 opened 5 years ago by bermanmaxim
  2

• top1 error = 0.99 when testing

  #48 opened 4 years ago by YoungDav
  1

• Size mismatch for bn running_mean and running_var, us_mobilenet_v1

  #47 opened 4 years ago by octavianmm
  0

• why use slimmable_training instead of universally_slimmable_training: True in Autoslim

  #38 opened 5 years ago by GYxiaOH
  7

• Would you share Image Super Resolution example model for Universally Slimmable Network training?

  #45 opened 4 years ago by conson0214
  1

• A Question

  #44 opened 4 years ago by Youskrpig
  1

• memory increasing during testing

  #43 opened 4 years ago by octavianmm
  1

• Question about slimming

  #41 opened 4 years ago by dada-thu
  1

• Run the experiment with distributed mode

  #42 opened 4 years ago by hwang595
  1

• width_mult = None when training us-Net

  #28 opened 5 years ago by lucaskyle
  3

• Using Cifar-100 to train slimmable_networks

  #39 opened 5 years ago by BcomedianC
  1

• Question about arbitrary width in a universally slimmable network.

  #36 opened 5 years ago by sseung0703
  8

• When will you release Greedy Slimming?

  #32 opened 5 years ago by apxlwl
  1

• channel configuration of auto-slim mobilenet-v2

  #16 opened 5 years ago by twangnh
  1

• The accuracy decrease sharply after calibration, even on the given Model zoo

  #35 opened 5 years ago by ZhanHuaa
  1

• Size mismatch when loading us_mobilenet_v2_calibrated.pt

  #34 opened 5 years ago by pawopawo
  3

• us.mobilenet_v2 cifar

  #29 opened 5 years ago by lucaskyle
  5

• how can i load us_model at a certain ratio

  #30 opened 5 years ago by lucaskyle
  5

• Calibration with non-uniform widths at test time

  #27 opened 5 years ago by MohammadSamragh
  6

• Why bn layers do not have weight_decay?

  #33 opened 5 years ago by leoozy
  1

• Is num_sample_training=2 enough for training ?

  #31 opened 5 years ago by bhheo
  3

• How to retrain slim-MobilenetV2

  #11 opened 6 years ago by Zeitzmz
  5

• Good job!

  #7 opened 5 years ago by zh583007354
  10

• US-Net BatchNorm

  #22 opened 5 years ago by semin-park
  6

• How to collect post statistics of BN

  #10 opened 6 years ago by VectorYoung
  8

• USNet for Obeject Detection

  #21 opened 5 years ago by VectorYoung
  1

• Memory Leak when Training US-Net

  #20 opened 5 years ago by VectorYoung
  1

• Sliming Object detection models

  #19 opened 5 years ago by abhigoku10
  1

• Training Speed, S-Net vs. US-Net

  #18 opened 6 years ago by VectorYoung
  1

• Training Speed

  #17 opened 6 years ago by VectorYoung
  4

• some questions about pertained model

  #8 opened 6 years ago by bestfleer
  3

• Exciting work! Some questions about universally slimmable networks

  #9 opened 6 years ago by TaojiannanYang
  3

• How did you combine the hard label loss and the Distillation loss

  #12 opened 6 years ago by VectorYoung
  7

• Reproduced Result

  #14 opened 6 years ago by VectorYoung
  1

• weight decay = 0 for BN parameters

  #15 opened 6 years ago by wesleywesleyteng
  5

• is the shufflenet implemented v1 or v2? Thanks.

  #13 opened 6 years ago by liwei46
  1

• Add code

  #2 opened 6 years ago by axzfeeomrzrqeue
  1