Code

Quality-of-Life for Google Colab#3

Code

Draft

woctezuma wants to merge 13 commits intoNVlabs:mainNVlabs/stylegan2-ada-pytorch:mainfrom woctezuma:google-colabwoctezuma/stylegan2-ada-pytorch:google-colabCopy head branch name to clipboard

+136-12

Commits

Commits on Jan 10, 2024

• pytorch 1.11 support: don't use conv2d_gradfix on v1.11, port grid_sample_gradfix to the new API

  Show description for e74827f
  

  jannehellsten

  authored and

  woctezuma

  committed
  e74827f

  Copy full SHA for e74827f

• Fix custom ops bug for pytorch 1.12 and onwards

  Show description for 4397f15
  

  jannehellsten

  authored and

  woctezuma

  committed
  4397f15

  Copy full SHA for 4397f15

• Support newer versions of PyTorch (v1.1X and v2)

  woctezuma

  committed
  471c7cb

  Copy full SHA for 471c7cb

• Save output as JPG instead of PNG

  woctezuma

  committed
  edc9f16

  Copy full SHA for edc9f16

• Add utility functions

  woctezuma

  committed
  bd645cf

  Copy full SHA for bd645cf

• Resume from the latest pickle

  woctezuma

  committed
  61bc29c

  Copy full SHA for 61bc29c

• Automatically set the resume value of kimg

  woctezuma

  committed
  7eeedaf

  Copy full SHA for 7eeedaf

• Allow to manually set the resume value of the augmentation strength

  woctezuma

  committed
  0e4e93e

  Copy full SHA for 0e4e93e

• Automatically set the resume value of augment_p

  woctezuma

  committed
  7e29bb5

  Copy full SHA for 7e29bb5

• Add cfg (auto_norp): auto cfg without EMA rampup

  woctezuma

  committed
  2b82b6a

  Copy full SHA for 2b82b6a

• Allow to override mapping net depth with --cfg_map

  woctezuma

  committed
  a51de58

  Copy full SHA for a51de58

• Allow to enforce CIFAR-specific architecture tuning with --cifar_tune

  woctezuma

  committed
  f83452f

  Copy full SHA for f83452f

• README: add changelog

  woctezuma

  committed
  2c507a6

  Copy full SHA for 2c507a6