soumith/ganhacks

Issues

• Generator and discriminator schedule

  #77 opened a year ago by jkdev100
  0

• GAN training example

  #76 opened a year ago by wendlerc
  0

• Activation functions for training and testing phases

  #75 opened 2 years ago by SITUSITU
  1

• MAE increasing when training!

  #74 opened 2 years ago by DISAPPEARED13
  0

• Why Discriminator Loss 0 is a failure mode ?

  #36 opened 7 years ago by arijitx
  15

• Generator ADAM X Discriminator SGD

  #48 opened 6 years ago by nuneslu
  3

• CycleGAN

  #73 opened 2 years ago by Blakey-Gavin
  0

• Tricks to sharpen GAN's synthesis

  #72 opened 3 years ago by DISAPPEARED13
  0

• Dose it need to transfer pretrained discriminator when transfer learning for GAN?

  #69 opened 4 years ago by micklexqg
  1

• Need help in understanding #1 and #16

  #56 opened 5 years ago by miqbal23
  2

• Reference for trick#16

  #68 opened 4 years ago by shaurov2253
  0

• Unexplainable GAN losses, need help

  #57 opened 5 years ago by lennert130994
  3

• Why "Construct different mini-batches for real and fake, i.e. each mini-batch needs to contain only all real images or all generated images"?

  #52 opened 6 years ago by mrgloom
  3

• > @KrnTneja : Thanks for your tricks. Could you provide any code to do it? I also meet the problem of loss D goes to zero

  #67 opened 4 years ago by shaurov2253
  1

• Discriminator loss of WGAN-gp drop sharply to zero

  #66 opened 4 years ago by AboMad
  0

• Hacks relevancy in 2020

  #64 opened 5 years ago by phillipiv
  1

• balance loss maybe is good idea?

  #65 opened 5 years ago by Oktai15
  0

• PixelShuffle, ConvTranspose2d + stride will get bad checkerboard

  #59 opened 5 years ago by gmt710
  2

• Understand #10

  #63 opened 5 years ago by fantasy-fish
  0

• Generator try to maximize, Discriminator try to minimize

  #49 opened 5 years ago by Auth0rM0rgan
  1

• Paper / talk associated?

  #62 opened 5 years ago by cossio
  0

• Validate a GAN

  #22 opened 7 years ago by dantp-ai
  5

• for GAN ，why my D loss is increse，and G loss decrease to 0 at the begining

  #29 opened 7 years ago by hefeiwangyande
  10

• Discriminator with accuracy 1, generator fools it perfectly. What's going on?

  #45 opened 6 years ago by danielegrattarola
  5

• Evaluate the Discriminator module over different set of images

  #60 opened 5 years ago by anamika06jain
  0

• is the training normal? how is the right train loss curve?

  #58 opened 5 years ago by micklexqg
  0

• How to handle failure mode

  #40 opened 7 years ago by solomatov
  3

• Question: Any interest in putting together a premium version for sale on SugarKubes?

  #54 opened 6 years ago by wrannaman
  3

• How would one go about training the network with one channel images?

  #34 opened 7 years ago by patakk
  2

• label smoothing one-sided?

  #53 opened 6 years ago by makeyourownalgorithmicart
  0

• Standard G & D Loss Values & Trends?

  #46 opened 6 years ago by James-G-Hill
  5

• Regarding sampling image in batches

  #51 opened 6 years ago by miqbal23
  0

• not working in trick 13: adding gaussian noise to every layer of generator

  #26 opened 7 years ago by danielkaifeng
  1

• Using generator and discriminator loss to evaluate GANs.

  #43 opened 7 years ago by Sinha-Raunak
  0

• Regarding 13: Add noise to inputs...

  #39 opened 7 years ago by vmvargas
  1

• G is always really smaller than D

  #47 opened 6 years ago by kristosh
  0

• There is no source that has empirically demonstrated a positive effect of replay buffer on GAN training

  #31 opened 7 years ago by AranKomat
  1

• Discriminator accuracy: 0.5 on average, but collapsed into 0 for negative samples and 1 for positive samples

  #44 opened 6 years ago by danielegrattarola
  2

• GAN - Multiple iterations for generator training

  #42 opened 7 years ago by ahmed-fau
  0

• Label smoothing

  #41 opened 7 years ago by knazeri
  0

• Saturation of G_loss and D_loss

  #38 opened 7 years ago by rishabhkumar296
  0

• How hard it is to make a GAN experiment reproducible?

  #37 opened 7 years ago by sbmaruf
  0

• Alternating training vs. DCGAN?

  #35 opened 7 years ago by meder411
  0

• Feasible loss simplification?

  #33 opened 7 years ago by florian-boehm
  2

• Trick 11: How to schedule thresholds A and B?

  #23 opened 7 years ago by benstaf
  2

• about generator loss

  #32 opened 7 years ago by fsalmasri
  0

• Replace BatchNorm with WeightNorm

  #24 opened 7 years ago by andyli
  1

• Update tips for WGANs

  #27 opened 7 years ago by ndahlquist
  3

• measuring quality

  #30 opened 7 years ago by malie
  0

• Sampling from a sphere instead of a cube

  #28 opened 7 years ago by mgarbade
  3