Commit Graph

Select branches

Hide Pull Requests

main

patch-1

0.0.1

0.0.2

0.1.0

0.1.0a

0.1.1

0.1.2

0.1.3

0.1.4

0.1.5

0.10.0

0.10.1

0.11.0

0.11.1

0.11.2

0.12.0

0.12.1

0.14.0

0.14.1

0.14.2

0.14.3

0.15.0

0.15.1

0.15.2

0.15.3

0.15.4

0.15.6

0.15.7

0.2.0

0.2.1

0.2.2

0.2.3

0.2.4

0.26.3

0.26.4

0.26.5

0.27.0

0.27.1

0.27.10

0.27.11

0.27.12

0.27.2

0.27.3

0.27.4

0.27.5

0.27.6

0.27.8

0.27.9

0.28.0

0.29.0

0.29.1

0.3.0

0.3.1

0.3.2

0.30.0

0.31.0

0.31.1

0.32.0

0.4.0

0.5.0

0.5.1

0.5.2

0.6.0

0.6.1

0.6.2

0.6.3

0.6.4

0.6.5

0.6.6

0.6.7

0.6.8

0.6.9

0.7.0

0.7.1

0.8.0

0.8.1

0.9.0

0.9.1

0.9.2

1.16.6

v0.16.0

v0.16.1

v0.16.2

v0.16.3

v0.16.4

v0.16.5

v0.16.6

v0.16.7

v0.17.0

v0.17.1

v0.17.2

v0.17.3

v0.17.4

v0.17.5

v0.17.6

v0.17.7

v0.18.0

v0.18.1

v0.18.2

v0.18.3

v0.18.4

v0.19.0

v0.19.1

v0.19.2

v0.20.0

v0.20.1

v0.20.2

v0.20.2a

v0.21.0

v0.21.1

v0.21.2

v0.22.0

v0.23.0

v0.23.1

v0.23.2

v0.23.3

v0.23.4

v0.24.0

v0.24.1

v0.24.2

v0.24.3

v0.24.4

v0.25.0

v0.25.1

v0.25.2

v0.25.3

v0.26.0

v0.26.1

Mono Color

• 4284c8840d clipping for continuous time diffusion not working v0.17.4 Phil Wang 2022-06-08 16:26:13 -07:00
• d4ffa3fced link to annotated ddpm Phil Wang 2022-06-08 12:55:15 -07:00
• c44d3ea01d learned noise schedule seems to be working, allow for one to make the monotonic net learn a bit more slowly than the unet v0.17.3 Phil Wang 2022-06-08 12:34:07 -07:00
• c4991f576f allow for configuring the hidden dimension of the monotonic mlp parameterizing the noise schedule v0.17.2 Phil Wang 2022-06-08 11:18:54 -07:00
• a19331aa59 fix learned noise schedule v0.17.1 Phil Wang 2022-06-08 10:19:02 -07:00
• 94eabaca1a complete learned noise schedule for variational ddpm paper, still need to finish cosine alpha schedule in log(snr) form v0.17.0 Phil Wang 2022-06-08 09:47:09 -07:00
• eaf9d9fdc4 unet needs to be conditioned on log(snr) in p_mean_variance for continuous time gaussian diffusion v0.16.7 Phil Wang 2022-06-08 00:41:41 -07:00
• 33eee3bb4a unet needs to be conditioned on log(snr) in p_mean_variance for continuous time gaussian diffusion v0.16.6 Phil Wang 2022-06-08 00:19:14 -07:00
• 3bf5e768c2 use a non-sinusoidal embedded condition for continuous time gaussian diffusion conditioned on log(snr) v0.16.5 Phil Wang 2022-06-07 21:15:27 -07:00
• 532178a6a3 assume when sampling all batch samples are at the same time, and do not noise for the last time step v0.16.4 Phil Wang 2022-06-07 16:12:44 -07:00
• 3bbb6ebf16 get working version of gaussian diffusion with continuous time (only beta linear schedule for now, but will eventually contain alpha cosine schedule as well as parameterized, learned monotonic MLP) v0.16.3 Phil Wang 2022-06-07 15:59:29 -07:00
• b50c9f6f13 get working version of gaussian diffusion with continuous time (only beta linear schedule for now, but will eventually contain alpha cosine schedule as well as parameterized, learned monotonic MLP) v0.16.2 Phil Wang 2022-06-07 15:49:39 -07:00
• 6b93fa48f6 fix comment Phil Wang 2022-06-06 17:30:46 -07:00
• a291da5098 bring back linear noise schedule, but default to cosine v0.16.1 Phil Wang 2022-05-27 19:13:05 -07:00
• e5a18bb25c switch over to film like conditioning, used by both openai and google at this point v0.16.0 Phil Wang 2022-05-24 23:47:34 -07:00
• fc8e4547aa higher default learning rate 0.15.7 Phil Wang 2022-05-16 13:39:55 -07:00
• cae9f4a71f whoops 0.15.6 Phil Wang 2022-05-14 13:59:21 -07:00
• 91f03fb88b optimize for simplicity and clarity - researcher does not need to worry about normalizing and unnormalizing now 0.15.4 Phil Wang 2022-05-14 11:38:43 -07:00
• 60128257c5 use tqdm pbar during training 0.15.3 Phil Wang 2022-05-13 20:25:49 -07:00
• cf6db71985 add gaussian diffusion where model predicts both noise and x_start, with a learned weighting between the two (experimental) 0.15.2 Phil Wang 2022-05-13 13:56:32 -07:00
• 27853452a2 add gaussian diffusion where model predicts both noise and x_start, with a learned weighting between the two (experimental) 0.15.1 Phil Wang 2022-05-13 12:52:11 -07:00
• 84ebb9ad13 offer predict_x0 objective 0.15.0 Phil Wang 2022-05-13 10:15:54 -07:00
• caa5af170d final cleanup 0.14.3 Phil Wang 2022-05-12 13:58:16 -07:00
• 55c658b967 cleanup unused Phil Wang 2022-05-12 11:52:06 -07:00
• e0f26677d6 make sure predicted mean is actually detached for all of the kl loss calculations 0.14.2 Phil Wang 2022-05-12 11:12:34 -07:00
• e147839d74 make sure to clip when sampling from gaussian diffusion with learned variance 0.14.1 Phil Wang 2022-05-12 10:08:53 -07:00
• 62e8490385 complete the gaussian diffusion with hybrid loss (learned variance) as in the improved ddpm paper 0.14.0 Phil Wang 2022-05-12 08:54:47 -07:00
• d412d8816b first pass at ddpm with learned variance Phil Wang 2022-05-11 17:38:29 -07:00
• 402b7c26df calculate noise schedule with float64 for numerical accuracy 0.12.1 Phil Wang 2022-05-10 15:23:34 -07:00
• 09613a40f3 cleanup Phil Wang 2022-05-07 05:47:21 -07:00
• c6966ae95a Merge pull request #24 from kashif/patch-1 Phil Wang 2022-05-07 05:32:51 -07:00
• 73591cf1ad updated citation in README Kashif Rasul 2022-05-07 11:23:45 +02:00
• 989f0fcb8e remove convnext blocks, they do not work well, validated in video diffusion repository 0.12.0 Phil Wang 2022-05-05 07:03:55 -07:00
• 84731bb03d groupnorm groups should be actually configurable 0.11.2 Phil Wang 2022-05-04 10:38:29 -07:00
• c6ecca555b allow for configuring expansion factor in convnext 0.11.1 Phil Wang 2022-05-04 10:33:23 -07:00
• 1f5c233072 bring back resnet blocks, make convnext blocks an experimental option 0.11.0 Phil Wang 2022-05-04 10:29:59 -07:00
• de378158e5 readme Phil Wang 2022-05-01 13:16:06 -07:00
• e274fb305a give an initial conv 0.10.1 Phil Wang 2022-05-01 08:49:38 -07:00
• f39b3b1d3f make sure time embedding dimension is kept at 4 x dimension (thanks @borisdayma) 0.10.0 Phil Wang 2022-04-29 14:55:12 -07:00
• 782c904d3b fix cosine beta schedule, thanks to @Zhengxinyang 0.9.2 Phil Wang 2022-04-19 20:51:50 -07:00
• 71953ebd22 fix bug, thanks to @jihoonerd 0.9.1 Phil Wang 2022-04-15 06:37:31 -07:00
• 0b8cdb4c8b remove outdated apex in favor of native pytorch AMP 0.9.0 Phil Wang 2022-04-13 08:59:18 -07:00
• e504e0e554 cleanup Phil Wang 2022-04-12 13:02:18 -07:00
• bd1e3b676e get rid of numpy 0.8.1 Phil Wang 2022-04-12 11:58:46 -07:00
• f4615599bc use full attention at the center of the unet 0.8.0 Phil Wang 2022-04-04 09:03:41 -07:00
• eb6e1b508e greater kernel size in convnext blocks 0.7.1 Phil Wang 2022-01-31 17:13:27 -08:00
• 91cff45939 replace resnets with convnext blocks 0.7.0 Phil Wang 2022-01-25 09:02:45 -08:00
• 7b51e30da7 fix layernorm 0.6.9 Phil Wang 2021-08-24 14:28:15 -07:00
• dadbf20154 remove stray print 0.6.8 Phil Wang 2021-07-16 15:18:20 -07:00
• 7706bdfc6f use pre-layernorm with linear attention, and also allow for turning off time embedding 0.6.7 Phil Wang 2021-06-25 10:48:11 -07:00
• 183e5f3cc5 move all constants into configurable class init parameters 0.6.6 Phil Wang 2021-06-25 10:37:36 -07:00
• 16c9ae7bb3 fix data not being normalized to range of -1 to 1 0.6.5 Phil Wang 2021-06-21 18:51:36 -07:00
• ab4c51c72c fix data not being normalized to range of -1 to 1 0.6.4 Phil Wang 2021-06-21 18:30:38 -07:00
• f5916111f8 0.6.3 0.6.3 Phil Wang 2021-06-21 17:41:16 -07:00
• ff02164ad7 image is in range of 0 to 1 0.6.2 Phil Wang 2021-06-21 13:04:35 -07:00
• ad9e303ff3 fix channels 0.6.1 Phil Wang 2021-06-11 15:28:36 -07:00
• ae42f48f6a prepare so that unet can work with a channel of one, and also make it so image size is hard coded in diffusion class. preparing for training on protein distograms 0.6.0 Phil Wang 2021-06-11 14:06:39 -07:00
• 5989f4c77e recommit sample Phil Wang 2020-10-13 09:32:56 -07:00
• 2082046888 set higher num train steps, so non-practitioners do not think it is completed Phil Wang 2020-10-11 13:49:49 -07:00
• 3c5b7e2d56 update readme Phil Wang 2020-10-10 10:37:44 -07:00
• d4ce9f6c38 save samples and models to ./results path 0.5.2 Phil Wang 2020-10-09 21:50:23 -07:00
• 5ad56dda25 save samples and models to ./results path 0.5.1 Phil Wang 2020-10-09 21:39:46 -07:00
• ff451f697e update with new and improved cosine noise scheduler Phil Wang 2020-10-09 21:21:02 -07:00
• f2765c4614 update with new and improved cosine noise scheduler 0.5.0 Phil Wang 2020-10-09 17:32:18 -07:00
• 3d96532c60 update citations in preparation to add improvements from a iclr 2021 paper Phil Wang 2020-10-05 17:15:28 -07:00
• ef2ca0b625 new paper suggests image linear attention is more effective without query normalization 0.4.0 Phil Wang 2020-10-04 21:53:53 -07:00
• 9f95a03c07 fix bug with rezero and linear attention 0.3.2 Phil Wang 2020-09-21 20:11:34 -07:00
• a4c68d3569 fix bug 0.3.1 Phil Wang 2020-09-15 15:57:39 -07:00
• b33a48e342 make sure when sampling, batch does not exceed training batch size 0.3.0 Phil Wang 2020-09-15 15:15:10 -07:00
• 8e5fb17063 add badge Phil Wang 2020-09-14 13:38:06 -07:00
• 4bf28914bc allow for mixed precision training with fp16 flag 0.2.4 Phil Wang 2020-09-08 17:26:23 -07:00
• b72235e71d allow for mixed precision training with fp16 flag 0.2.3 Phil Wang 2020-09-08 17:25:26 -07:00
• 88f83d0ff2 fix loading checkpoint 0.2.2 Phil Wang 2020-09-08 15:15:59 -07:00
• 26b5cab6c8 update ema more frequently 0.2.1 Phil Wang 2020-09-08 15:03:38 -07:00
• 1307b3115d add exponential moving average of model, also allow passing in custom noise schedule 0.2.0 Phil Wang 2020-09-08 14:58:09 -07:00
• 81fb2a0386 add sample Phil Wang 2020-09-08 10:17:48 -07:00
• 698227ae13 fix small bug with gradient accumulation 0.1.4 Phil Wang 2020-09-07 23:14:38 -07:00
• c479adf960 add interpolation 0.1.3 Phil Wang 2020-09-07 22:38:59 -07:00
• d70fb08f8a small helper fn to make sampling more clear Phil Wang 2020-09-07 16:40:36 -07:00
• e700a7c6de add image back Phil Wang 2020-09-07 10:37:58 -07:00
• 9c758662a3 fix another stray bug 0.1.2 Phil Wang 2020-09-06 23:25:30 -07:00
• 1f1e42e9f9 update readme 0.1.1 Phil Wang 2020-09-06 22:35:37 -07:00
• e1800c1a8d remove wip, seems to be working Phil Wang 2020-09-06 14:26:12 -07:00
• 11f27032ba offer training class to easily train model off an image directory 0.1.0 Phil Wang 2020-09-06 14:22:44 -07:00
• d8472a6220 update readme Phil Wang 2020-09-06 03:44:25 -07:00
• d59d8b05f6 fix a small bug 0.0.2 Phil Wang 2020-09-06 01:50:59 -07:00
• 365ce0c335 update readme Phil Wang 2020-09-06 01:44:05 -07:00
• 0cc4aecfd5 first alpha release 0.0.1 Phil Wang 2020-09-06 01:40:57 -07:00
• 79e0765675 make sampling work Phil Wang 2020-09-06 01:24:44 -07:00
• ee76497b83 complete training, sans sampling Phil Wang 2020-09-05 23:41:02 -07:00
• e496bba350 more cleanup Phil Wang 2020-09-05 14:11:32 -07:00
• 1b05546b47 cleanup Phil Wang 2020-09-05 13:57:44 -07:00
• c9855c8c74 complete model portion Phil Wang 2020-09-05 13:55:09 -07:00
• 73dc91e9f0 Create python-publish.yml Phil Wang 2020-08-31 12:47:17 -07:00
• dda425dc61 add scaffold Phil Wang 2020-08-31 12:45:27 -07:00
• df94747c0c Update README.md Phil Wang 2020-08-25 19:24:13 -07:00
• f94876c6fb Update README.md Phil Wang 2020-08-25 19:23:06 -07:00
• 602421c714 Initial commit Phil Wang 2020-08-25 19:22:11 -07:00