ZF/vit-pytorch
1
0
Fork 0
mirror of https://github.com/lucidrains/vit-pytorch.git synced 2025-12-30 08:02:29 +00:00
Code Issues Packages Projects Releases Wiki Activity
222 Commits 3 Branches 254 Tags
6db20debb4c882cfde8a1759aa56e4fc295f0713
Commit Graph

110 Commits

Author SHA1 Message Date
Phil Wang
6db20debb4 add patch merger 2022-03-01 16:50:17 -08:00
Phil Wang
1bae5d3cc5 allow for rectangular images for efficient adapter 2022-01-31 08:55:31 -08:00
Phil Wang
25b384297d return None from extractor if no attention layers 2022-01-28 17:49:58 -08:00
Phil Wang
64a07f50e6 epsilon should be inside square root 2022-01-24 17:24:41 -08:00
Phil Wang
c1528acd46 fix feature maps in Nest, thanks to @MarkYangjiayi 2022-01-22 13:17:30 -08:00
Phil Wang
1cc0f182a6 decoder positional embedding needs to be reapplied https://twitter.com/giffmana/status/1479195631587631104 2022-01-06 13:14:41 -08:00
Phil Wang
28eaba6115 0.26.2 2022-01-03 12:56:34 -08:00
Phil Wang
91ed738731 0.26.1 2021-12-30 19:31:26 -08:00
Phil Wang
70ba532599 add ViT for small datasets https://arxiv.org/abs/2112.13492 2021-12-28 10:58:21 -08:00
Phil Wang
e52ac41955 allow extractor to only return embeddings, to ready for vision transformers to be used in x-clip 2021-12-25 12:31:21 -08:00
Phil Wang
0891885485 include tests in package for conda 2021-12-22 12:44:29 -08:00
Phil Wang
976f489230 add some tests 2021-12-22 09:13:31 -08:00
Phil Wang
2c368d1d4e add extractor wrapper 2021-12-21 11:11:39 -08:00
Phil Wang
b983bbee39 release MobileViT, from @murufeng 2021-12-21 10:22:59 -08:00
Phil Wang
365b4d931e add adaptive token sampling paper 2021-12-03 19:52:40 -08:00
Phil Wang
ff44d97cb0 make initial channels customizable for PiT 2021-11-22 18:08:49 -08:00
Phil Wang
b69b5af34f dynamic positional bias for crossformer the more efficient way as described in appendix of paper 2021-11-22 17:39:36 -08:00
Phil Wang
36e32b70fb complete and release crossformer 2021-11-22 17:10:53 -08:00
Phil Wang
5ae555750f add SimMIM 2021-11-21 15:50:19 -08:00
Phil Wang
e8f6d72033 release masked autoencoder 2021-11-12 20:08:48 -08:00
Phil Wang
cb1729af28 more efficient feedforward for regionvit 2021-11-07 17:18:59 -08:00
Phil Wang
06d375351e add RegionViT paper 2021-11-07 09:47:28 -08:00
Phil Wang
f196d1ec5b move freqs in RvT to linspace 2021-10-05 09:23:44 -07:00
Phil Wang
d2d6de01d3 0.20.7 2021-08-30 08:14:43 -07:00
Phil Wang
6e1be11517 0.20.6 2021-08-21 09:03:54 -07:00
Phil Wang
aeed3381c1 use hardswish for levit 2021-08-19 08:22:55 -07:00
Phil Wang
3f754956fb remove last transformer layer in t2t 2021-08-14 08:06:23 -07:00
Phil Wang
918869571c fix hard distillation, thanks to @CiaoHe 2021-08-12 08:40:57 -07:00
Phil Wang
e5324242be fix wrong norm in nest 2021-08-05 12:55:48 -07:00
Phil Wang
22da26fa4b fix recorder in data parallel situation 2021-07-08 10:15:07 -07:00
Phil Wang
a6c085a2df 0.20.0 for cct 2021-07-02 15:48:48 -07:00
Phil Wang
64a2ef6462 fix mpp 2021-06-16 16:46:32 -07:00
Phil Wang
53884f583f 0.19.5 2021-06-16 14:24:46 -07:00
Phil Wang
60ad4e266e layernorm on channel dimension == instancenorm2d with affine set to true 2021-06-03 16:41:45 -07:00
Phil Wang
a254a0258a fix typo 2021-06-01 07:33:00 -07:00
Phil Wang
26df10c0b7 fix max pool in nest 2021-05-28 11:06:02 -07:00
Phil Wang
17cb8976df make nest resilient to dimension that are not divisible by number of heads 2021-05-27 22:41:07 -07:00
Phil Wang
daf3abbeb5 add NesT 2021-05-27 22:02:17 -07:00
Phil Wang
b483b16833 0.18.4 2021-05-18 14:40:33 -07:00
Phil Wang
679e5be3e7 apply scale to 2d rel pos bias in levit 2021-05-10 11:37:23 -07:00
Phil Wang
04f86dee3c implement SOTA new self-supervised learning technique from facebook for vision transformers, Dino 2021-05-02 14:00:36 -07:00
Phil Wang
6549522629 be able to accept non-square patches, thanks to @FilipAndersson245 2021-05-01 20:04:41 -07:00
Phil Wang
65bb350e85 0.17.2 2021-04-30 06:44:54 -07:00
Phil Wang
7807f24509 fix small bug 2021-04-29 15:39:41 -07:00
Phil Wang
30a1335d31 release twins svt 2021-04-29 14:55:25 -07:00
Phil Wang
4f3dbd003f for PiT, project to increased dimensions on first grouped conv for depthwise-conv 2021-04-29 12:41:00 -07:00
Phil Wang
0df1505662 add zeroing of weight parameters of batchnorm in levit just before residual connection, noticed by @EelcoHoogendoorn 2021-04-27 08:41:16 -07:00
Phil Wang
3df6c31c61 fix norm issues in cvt 2021-04-27 08:36:17 -07:00
Phil Wang
54af220930 fix cvt 2021-04-26 20:37:51 -07:00
Phil Wang
bad4b94e7b fix all issues with rotary vision transformer 2021-04-25 12:09:32 -07:00