Commits

Branch selector

 main

User selector

All users

Datepicker

All time

Commit History

Commits on Dec 5, 2022

• need simple vit with patch dropout for another project

  lucidrains

  committed
  500e231

  Copy full SHA for 500e231

Commits on Dec 2, 2022

• add vit with patch dropout, fully embrace structured dropout as multiple papers are now corroborating each other

  lucidrains

  committed
  89e1996

  Copy full SHA for 89e1996

Commits on Dec 1, 2022

• offer 1d versions, in light of https://arxiv.org/abs/2211.14730

  lucidrains

  committed
  2f87c0c

  Copy full SHA for 2f87c0c

Commits on Oct 29, 2022

• try to fix tests

  lucidrains

  committed
  59c8948

  Copy full SHA for 59c8948

• add a 3d version of cct, addressing lucidrains#238 0.38.1

  lucidrains

  committed
  cb6d749

  Copy full SHA for cb6d749

Commits on Oct 25, 2022

• make sure global average pool can be used for vivit in place of cls token

  lucidrains

  committed
  6ec8fda

  Copy full SHA for 6ec8fda

Commits on Oct 24, 2022

• add vivit

  lucidrains

  committed
  13fabf9

  Copy full SHA for 13fabf9

Commits on Oct 17, 2022

• Improving Readability (lucidrains#220)

  Show description for c0eb4c0
  ryanrussell

  authored
  c0eb4c0

  Copy full SHA for c0eb4c0

• release updated mae where one can more easily visualize reconstructions, thanks to @Vishu26

  lucidrains

  committed
  5f1a6a0

  Copy full SHA for 5f1a6a0

• Update mae.py (lucidrains#242)

  Show description for 9a95e79
  Vishu26

  authored
  9a95e79

  Copy full SHA for 9a95e79

• add the 3d simple vit

  lucidrains

  committed
  b4853d3

  Copy full SHA for b4853d3

Commits on Oct 16, 2022

• begin extending some of the architectures over to 3d, starting with basic ViT

  lucidrains

  committed
  29fbf0a

  Copy full SHA for 29fbf0a

Commits on Jul 27, 2022

• add link to Flax translation by @conceptofmind

  lucidrains

  authored
  4b8f5bc

  Copy full SHA for 4b8f5bc

Commits on Jul 16, 2022

• offer way for extractor to return latents without detaching them

  lucidrains

  committed
  f86e052

  Copy full SHA for f86e052

Commits on Jun 30, 2022

• slightly more clear of einops rearrange for cls token, for lucidrains#224

  lucidrains

  committed
  2fa2b62

  Copy full SHA for 2fa2b62

Commits on Jun 29, 2022

• follow @arquolo feedback and advice for MaxViT

  lucidrains

  committed
  9f87d1c

  Copy full SHA for 9f87d1c

Commits on Jun 25, 2022

• fix hidden dimension in MaxViT thanks to @arquolo

  lucidrains

  committed
  2c6dd70

  Copy full SHA for 2c6dd70

Commits on Jun 19, 2022

• be able to accept a reference to a layer within the model for forward hooking and extracting the embedding output, for regionvit to work with extractor

  lucidrains

  committed
  6460119

  Copy full SHA for 6460119

• make extractor flexible for layers that output multiple tensors, show CrossViT example

  lucidrains

  committed
  4e62e5f

  Copy full SHA for 4e62e5f

Commits on May 4, 2022

• add simple vit, from https://arxiv.org/abs/2205.01580

  lucidrains

  committed
  b3e90a2

  Copy full SHA for b3e90a2

Commits on May 3, 2022

• add EsViT, by popular request, an alternative to Dino that is compatible with efficient ViTs with accounting for regional self-supervised loss

  lucidrains

  committed
  4ef72fc

  Copy full SHA for 4ef72fc

Commits on Apr 7, 2022

• fix bibtex typo (lucidrains#212)

  vztu

  authored
  c2aab05

  Copy full SHA for c2aab05

Commits on Apr 6, 2022

• fix mbconv residual block

  lucidrains

  committed
  81661e3

  Copy full SHA for 81661e3

• fix maxvit - need feedforwards after attention

  lucidrains

  committed
  13f8e12

  Copy full SHA for 13f8e12

• link to maxvit in readme

  lucidrains

  committed
  2d4089c

  Copy full SHA for 2d4089c

• maxvit intent to build (lucidrains#211)

  Show description for c7bb5fc
  lucidrains

  authored
  c7bb5fc

  Copy full SHA for c7bb5fc

• sponsor button

  lucidrains

  authored
  946b19b

  Copy full SHA for 946b19b

Commits on Mar 31, 2022

• let windowed tokens exchange information across heads a la talking heads prior to pointwise attention in sep-vit

  lucidrains

  committed
  d93cd84

  Copy full SHA for d93cd84

• cleanup sepvit

  lucidrains

  committed
  5d4c798

  Copy full SHA for 5d4c798

• intent to build (lucidrains#210)

  Show description for d65a742
  lucidrains

  authored
  d65a742

  Copy full SHA for d65a742

• do not layernorm on last transformer block for scalable vit, as there is already one in mlp head

  lucidrains

  committed
  8c54e01

  Copy full SHA for 8c54e01

• complete learnable memory ViT, for efficient fine-tuning and potentially plays into continual learning

  lucidrains

  committed
  df656fe

  Copy full SHA for df656fe

Commits on Mar 30, 2022

• correct need for post-attention dropout

  lucidrains

  committed
  4e6a42a

  Copy full SHA for 4e6a42a

Commits on Mar 28, 2022

• link to tensorflow2 translation by @taki0112

  lucidrains

  committed
  6d7298d

  Copy full SHA for 6d7298d

Commits on Mar 26, 2022

• CCT allow for rectangular images

  lucidrains

  committed
  9cd56ff

  Copy full SHA for 9cd56ff

Pagination

Previous

Next