HyperGAN
  • About
  • Getting started
  • CLI guide
  • Configurations
    • Configurable Parameters
  • Showcase
    • AI Explorer for Android
    • Youtube, Twitter, Discord +
  • Examples
    • 2D
    • Text
    • Classification
    • Colorizer
    • Next Frame (video)
  • Tutorials
    • Training a GAN
    • Pygame inference
    • Creating an image dataset
    • Searching for hyperparameters
  • Components
    • GAN
      • Aligned GAN
      • Aligned Interpolated GAN
      • Standard GAN
    • Generator
      • Configurable Generator
      • DCGAN Generator
      • Resizable Generator
    • Discriminator
      • DCGAN Discriminator
      • Configurable Discriminator
    • Layers
      • add
      • cat
      • channel_attention
      • ez_norm
      • layer
      • mul
      • multi_head_attention
      • operation
      • pixel_shuffle
      • residual
      • resizable_stack
      • segment_softmax
      • upsample
    • Loss
      • ALI Loss
      • F Divergence Loss
      • Least Squares Loss
      • Logistic Loss
      • QP Loss
      • RAGAN Loss
      • Realness Loss
      • Softmax Loss
      • Standard Loss
      • Wasserstein Loss
    • Latent
      • Uniform Distribution
    • Trainer
      • Alternating Trainer
      • Simultaneous Trainer
      • Balanced Trainer
      • Accumulate Gradient Trainer
    • Optimizer
    • Train Hook
      • Adversarial Norm
      • Weight Constraint
      • Stabilizing Training
      • JARE
      • Learning Rate Dropout
      • Gradient Penalty
      • Rolling Memory
    • Other GAN implementations
Powered by GitBook
On this page
  • Access
  • Events
  • before_step(feed_dict)
  • after_step(feed_dict)
  • after_create()
  • gradients(d_grads, g_grads)

Was this helpful?

  1. Components

Train Hook

Train hooks provide training events and loss modification to trainers.

PreviousOptimizerNextAdversarial Norm

Last updated 4 years ago

Was this helpful?

Access

gan.trainer.train_hooks # => [...]

Train hooks are setup and invoked by the trainer.

Events

Override these methods to change the train loop

before_step(step, feed_dict)
after_step(step, feed_dict)
after_create()
gradients(d_grads, g_grads)

before_step(feed_dict)

after_step(feed_dict)

Executed before/after the step takes place. feed_dict is what is being sent to the graph during the training step.

after_create()

Ran after the trainer is created.

gradients(d_grads, g_grads)

Refines the gradients before they are applied to the optimizer.

https://github.com/HyperGAN/HyperGAN/tree/master/hypergan/train_hooks