Andrej Karpathy (@karpathy) “Nice new paper improving image generation and (generative) unsupervised represen” — TopicDigg
Andrej Karpathy
@karpathy
I like training large deep neural nets.
加入 April 2009
1.1K 正在关注 3M 粉丝
Nice new paper improving image generation and (generative) unsupervised representation learning uses ViT instead of CNN to improve VQGAN into a new "ViT-VQGAN" image patch tokenizer. Tokens are then fed into a GPT for image generation, or linear probing.
显示更多
