mrq/DL-Art-School
1
3
Fork 5
Code Issues Pull Requests Packages Projects Releases Wiki Activity
1,690 Commits 1 Branch 0 Tags 5.6 MiB
71b73db044
Commit Graph

278 Commits

Author SHA1 Message Date
James Betker
54bff35171 Fix issue where eval was not being used by all ddp processes 2021-06-14 09:50:04 -06:00
James Betker
3e3ad7825f Add support for training an EMA network alongside the main networks 2021-06-12 21:01:41 -06:00
James Betker
696f320820 Get rid of feature networks 2021-06-11 20:50:07 -06:00
James Betker
65c474eecf Various changes to fix testing 2021-06-11 15:31:10 -06:00
James Betker
2ad2b56438 Don't do wandb except on rank 0 2021-06-06 16:52:07 -06:00
James Betker
692e9c417b Support diffusion unet 2021-06-06 13:57:22 -06:00
James Betker
fa908a6a15 Fix wandb import issue 2021-06-04 23:27:15 -06:00
James Betker
103a88506e Log eval to wandb 2021-06-04 23:23:20 -06:00
James Betker
6084915af8 Support gaussian diffusion models 
Adds support for GD models, courtesy of some maths from openai.

Also:
- Fixes requirement for eval{} even when it isn't being used
- Adds support for denormalizing an imagenet norm
 2021-06-02 21:47:32 -06:00
James Betker
45bc76ba92 Fixes and mods to support training classifiers on imagenet 2021-06-01 17:25:24 -06:00
James Betker
f129eaa39e Clean up byol a bit 
- Remove option to aug in dataset (there's really no reason for this now that kornia works on GPU on windows)
- Other stufff
 2021-05-24 21:35:46 -06:00
James Betker
119f17c808 Add testing capabilities for segformer & contrastive feature 2021-04-27 09:59:50 -06:00
James Betker
9bbe6fc81e Get segformer to a trainable state 2021-04-25 11:45:20 -06:00
James Betker
17555e7d07 misc adjustments for stylegan 2021-04-21 18:14:17 -06:00
James Betker
f89ea5f1c6 Mods to support lightweight_gan model 2021-03-02 20:51:48 -07:00
James Betker
784b96c059 Misc options to add support for training stylegan2-rosinality models: 
- Allow image_folder_dataset to normalize inbound images
- ExtensibleTrainer can denormalize images on the output path
- Support .webp - an output from LSUN
- Support logistic GAN divergence loss
- Support stylegan2 TF weight extraction for discriminator
- New injector that produces latent noise (with separated paths)
- Modify FID evaluator to be operable with rosinality-style GANs
 2021-02-08 08:09:21 -07:00
James Betker
7070142805 Make vqvae3_hard more configurable 2021-02-04 09:03:22 -07:00
James Betker
51b63b2aa6 Add switched_conv with hard routing and make vqvae use it. 2021-01-25 08:25:29 -07:00
James Betker
557cdec116 misc 2021-01-23 13:45:17 -07:00
James Betker
587a4f4050 resnet_unet_3 
I'm being really lazy here - these nets are not really different from each other
except at which layer they terminate. This one terminates at 2x downsampling,
which is simply indicative of a direction I want to go for testing these pixpro networks.
 2021-01-15 14:51:03 -07:00
James Betker
d1007ccfe7 Adjustments to pixpro to allow training against networks with arbitrarily large structural latents 
- The pixpro latent now rescales the latent space instead of using a "coordinate vector", which
   **might** have performance implications.
- The latent against which the pixel loss is computed can now be a small, randomly sampled patch
   out of the entire latent, allowing further memory/computational discounts. Since the loss
   computation does not have a receptive field, this should not alter the loss.
- The instance projection size can now be separate from the pixel projection size.
- PixContrast removed entirely.
- ResUnet with full resolution added.
 2021-01-12 09:17:45 -07:00
James Betker
34f8c8641f Support training imagenet classifier 2021-01-11 20:09:16 -07:00
James Betker
07168ecfb4 Enable vqvae to use a switched_conv variant 2021-01-09 20:53:14 -07:00
James Betker
61a86a3c1e VQVAE 2021-01-07 10:20:15 -07:00
James Betker
9fed90393f Add lucidrains pixpro trainer 2021-01-05 20:14:22 -07:00
James Betker
2c65b6b28e More mods to support styledsr 2021-01-04 11:32:28 -07:00
James Betker
4d8064c32c Modifications to allow partially trained stylegan discriminators to be used 2021-01-03 16:37:18 -07:00
James Betker
ce6524184c Do the last commit but in a better way 2021-01-02 22:24:12 -07:00
James Betker
edf9c38198 Make ExtensibleTrainer set the starting step for the LR scheduler 2021-01-02 22:22:34 -07:00
James Betker
bdbab65082 Allow optimizers to train separate param groups, add higher dimensional VGG discriminator 
Did this to support training 512x512px networks off of a pretrained 256x256 network.
 2021-01-02 15:10:06 -07:00
James Betker
193cdc6636 Move discriminators to the create_model paradigm 
Also cleans up a lot of old discriminator models that I have no intention
of using again.
 2021-01-01 15:56:09 -07:00
James Betker
9864fe4c04 Fix for train.py 2021-01-01 11:59:00 -07:00
James Betker
0eb1f4dd67 Revert "Get rid of CUDA_VISIBLE_DEVICES" 
It is actually necessary for training in distributed mode. Only
do it then.
 2020-12-31 10:31:40 -07:00
James Betker
1de1fa30ac Disable refs and centers altogether in single_image_dataset 
I suspect that this might be a cause of failures on parallel datasets.
Plus it is unnecessary computation.
 2020-12-31 10:13:24 -07:00
James Betker
8f0984cacf Add sr_fid evaluator 2020-12-30 20:18:58 -07:00
James Betker
a777c1e4f9 Misc script fixes 2020-12-29 20:25:09 -07:00
James Betker
3fd627fc62 Mods to support image classification & filtering 2020-12-26 13:49:27 -07:00
James Betker
29db7c7a02 Further mods to BYOL 2020-12-24 09:28:41 -07:00
James Betker
1bbcb96ee8 Implement a few changes to support training BYOL networks 2020-12-23 10:50:23 -07:00
James Betker
2437b33e74 Fix srflow_latent_space_playground bug 2020-12-22 15:42:38 -07:00
James Betker
e82f4552db Update other docs with dumb config options 2020-12-18 16:21:28 -07:00
James Betker
5640e4efe4 More refactoring 2020-12-18 09:18:34 -07:00
James Betker
a8179ff53c Image label work 2020-12-18 08:53:18 -07:00
James Betker
fb2cfc795b Update requirements, add image_patch_classifier tool 2020-12-16 09:42:50 -07:00
James Betker
e5a3e6b9b5 srflow latent space misc 2020-12-14 23:59:49 -07:00
James Betker
ec0ee25f4b Structural latents checkpoint 2020-12-11 12:01:09 -07:00
James Betker
a5630d282f Get rid of 2nd trainer 2020-12-10 09:57:38 -07:00
James Betker
11155aead4 Directly use dataset keys 
This has been a long time coming. Cleans up messy "GT" nomenclature and simplifies ExtensibleTraner.feed_data
 2020-12-04 20:14:53 -07:00
James Betker
8a83b1c716 Go back to apex DDP, fix distributed bugs 2020-12-04 16:39:21 -07:00
James Betker
8a00f15746 Implement FlowGaussianNll evaluator 2020-12-02 14:09:54 -07:00
James Betker
2e0bbda640 Remove unused archs 2020-12-01 11:10:48 -07:00
James Betker
da604752e6 Misc RRDB changes 2020-11-29 12:21:31 -07:00
James Betker
a1d4c9f83c multires rrdb work 2020-11-28 14:35:46 -07:00
James Betker
ef8d5f88c1 Bring split gaussian nll out of split so it can be computed accurately with the rest of the nll component 2020-11-27 13:30:21 -07:00
James Betker
fd356580c0 Play with lambdas 2020-11-26 20:30:55 -07:00
James Betker
45a489110f Fix datasets 2020-11-26 11:50:38 -07:00
James Betker
f6098155cd Mods to tecogan to allow use of embeddings as input 2020-11-24 09:24:02 -07:00
James Betker
b10bcf6436 Rework stylegan_for_sr to incorporate structure as an adain block 2020-11-23 11:31:11 -07:00
James Betker
5ccdbcefe3 srflow_orig integration 2020-11-19 23:47:24 -07:00
James Betker
d7877d0a36 Fixes to teco losses and translational losses 2020-11-19 11:35:05 -07:00
James Betker
6b679e2b51 Make grad_penalty available to classical discs 2020-11-17 18:31:40 -07:00
James Betker
8a19c9ae15 Add additive mode to rrdb 2020-11-16 20:45:09 -07:00
James Betker
125cb16dce Add a FID evaluator for stylegan with structural guidance 2020-11-14 20:16:07 -07:00
James Betker
ec621c69b5 Fix train bug 2020-11-14 09:29:08 -07:00
James Betker
a07e1a7292 Add separate Evaluator module and FID evaluator 2020-11-13 11:03:54 -07:00
James Betker
fc55bdb24e Mods to how wandb are integrated 2020-11-12 15:45:25 -07:00
James Betker
db9e9e28a0 Fix an issue where GPU0 was always being used in non-ddp 
Frankly, I don't understand how this has ever worked. WTF.
 2020-11-12 15:43:01 -07:00
James Betker
88f349bdf1 Enable usage of wandb 2020-11-11 21:48:56 -07:00
James Betker
6a2fd5f7d0 Lots of new discriminator nets 2020-11-10 16:06:54 -07:00
James Betker
0cf52ef52c latent work 2020-11-06 20:38:23 -07:00
James Betker
74738489b9 Fixes and additional support for progressive zoom 2020-10-30 09:59:54 -06:00
James Betker
607ff3c67c RRDB with bypass 2020-10-29 09:39:45 -06:00
James Betker
da53090ce6 More adjustments to support distributed training with teco & on multi_modal_train 2020-10-27 20:58:03 -06:00
James Betker
2a3eec8fd7 Fix some distributed training snafus 2020-10-27 15:24:05 -06:00
James Betker
ff58c6484a Fixes to unified chunk datasets to support stereoscopic training 2020-10-26 11:12:22 -06:00
James Betker
9c3d059ef0 Updates to be able to train flownet2 in ExtensibleTrainer 
Only supports basic losses for now, though.
 2020-10-24 11:56:39 -06:00
James Betker
8636492db0 Copy train.py mods to train2 2020-10-22 17:16:36 -06:00
James Betker
e9c0b9f0fd More adjustments to support multi-modal training 
Specifically - looks like at least MSE loss cannot handle autocasted tensors
 2020-10-22 16:49:34 -06:00
James Betker
76789a456f Class-ify train.py and workon multi-modal trainer 2020-10-22 16:15:31 -06:00
James Betker
3e3d2af1f3 Add multi-modal trainer 2020-10-22 13:27:32 -06:00
James Betker
5753e77d67 ChainedGen: Output debugging information on blocks 2020-10-21 16:36:23 -06:00
James Betker
d8c6a4bbb8 Misc 2020-10-20 12:56:52 -06:00
James Betker
331c40f0c8 Allow starting step to be forced 
Useful for testing purposes or to force a validation.
 2020-10-19 15:23:04 -06:00
James Betker
981d64413b Support validation over a custom injector 
Also re-enable PSNR
 2020-10-19 11:01:56 -06:00
James Betker
9ead2c0a08 Multiscale training in! 2020-10-17 22:54:12 -06:00
James Betker
d856378b2e Add ChainedGenWithStructure 2020-10-16 20:44:36 -06:00
James Betker
6f8705e8cb SSGSimpler network 2020-10-15 17:18:44 -06:00
James Betker
24792bdb4f Codebase cleanup 
Removed a lot of legacy stuff I have no intent on using again.
Plan is to shape this repo into something more extensible (get it? hah!)
 2020-10-13 20:56:39 -06:00
James Betker
17d78195ee Mods to SRG to support returning switch logits 2020-10-13 20:46:37 -06:00
James Betker
2bc5701b10 misc 2020-10-12 10:21:25 -06:00
James Betker
b2c4b2a16d Move gpu_ids out of if statement 2020-10-06 20:40:20 -06:00
James Betker
0e3ea63a14 Misc 2020-10-05 18:01:50 -06:00
James Betker
ffd069fd97 Lots of SSG work 
- Checkpointed pretty much the entire model - enabling recurrent inputs
- Added two new models for test - adding depth (again) and removing SPSR (in lieu of the new losses)
 2020-10-04 20:48:08 -06:00
James Betker
fc396baf1a Move loaded_options to util 
Doesn't seem to work with python 3.6
 2020-10-03 20:29:06 -06:00
James Betker
3cbb9ecd45 Misc 2020-10-03 16:15:42 -06:00
James Betker
21d3bb83b2 Use tqdm reporting with validation 2020-10-03 11:16:39 -06:00
James Betker
6c9718ad64 Don't log if you aren't 0 rank 2020-10-03 11:14:13 -06:00
James Betker
19a4075e1e Allow checkpointing to be disabled in the options file 
Also makes options a global variable for usage in utils.
 2020-10-03 11:03:28 -06:00
James Betker
c9a9e5c525 Prompt user for gpu_id if multiple gpus are detected 2020-10-01 17:24:50 -06:00
James Betker
0b5a033503 spsr7 + cleanup 
SPSR7 adds ref onto spsr6, makes more "common sense" mods.
 2020-09-29 16:59:26 -06:00
James Betker
eb12b5f887 Misc 2020-09-26 21:27:17 -06:00
James Betker
254cb1e915 More dataset integration work 2020-09-25 22:19:38 -06:00
James Betker
f211575e9d Save models before validation 
Validation often fails with OOM, wasting hours of training time.
Save models first.
 2020-09-16 08:17:17 -06:00
James Betker
c833bd1eac Misc changes 2020-09-15 20:57:59 -06:00
James Betker
747ded2bf7 Fixes to the spsr3 
Some lessons learned:
- Biases are fairly important as a relief valve. They dont need to be everywhere, but
  most computationally heavy branches should have a bias.
- GroupNorm in SPSR is not a great idea. Since image gradients are represented
   in this model, normal means and standard deviations are not applicable. (imggrad
   has a high representation of 0).
- Don't fuck with the mainline of any generative model. As much as possible, all
   additions should be done through residual connections. Never pollute the mainline
   with reference data, do that in branches. It basically leaves the mode untrainable.
 2020-09-09 15:28:14 -06:00
James Betker
c04f244802 More mods 2020-09-08 20:36:27 -06:00
James Betker
e6207d4c50 SPSR3 work 
SPSR3 is meant to fix whatever is causing the switching units
inside of the newer SPSR architectures to fail and basically
not use the multiplexers.
 2020-09-08 15:14:23 -06:00
James Betker
a18ece62ee Add updated spsr net for test 2020-09-07 17:01:48 -06:00
James Betker
e8613041c0 Add novograd optimizer 2020-09-06 17:27:08 -06:00
James Betker
6657a406ac Mods needed to support training a corruptor again: 
- Allow original SPSRNet to have a specifiable block increment
- Cleanup
- Bug fixes in code that hasnt been touched in awhile.
 2020-09-04 15:33:39 -06:00
James Betker
886d59d5df Misc fixes & adjustments 2020-09-01 07:58:11 -06:00
James Betker
0a9b85f239 Fix vgg_gn input_img_factor 2020-08-31 09:50:30 -06:00
James Betker
4b4d08bdec Enable testing in ExtensibleTrainer, fix it in SRGAN_model 
Also compute fea loss for this.
 2020-08-31 09:41:48 -06:00
James Betker
623f3b99b2 Stupid pathing.. 2020-08-26 17:58:24 -06:00
James Betker
80aa83bfd2 Try copytree for tb_logger again. 2020-08-26 17:55:02 -06:00
James Betker
b593d8e7c3 Save tb_logger to alt_path 2020-08-26 17:45:07 -06:00
James Betker
f35b3ad28f Fix val behavior for ExtensibleTrainer 2020-08-26 08:44:22 -06:00
James Betker
19487d9bbd Fix distributed launch for large distributed runs 2020-08-25 15:42:59 -06:00
James Betker
a65b07607c Reference network 2020-08-25 11:56:59 -06:00
James Betker
f9276007a8 More fixes to corrupt_fea 2020-08-23 17:52:18 -06:00
James Betker
dffc15184d More ExtensibleTrainer work 
It runs now, just need to debug it to reach performance parity with SRGAN. Sweet.
 2020-08-23 17:22:45 -06:00
James Betker
afdd93fbe9 Grey feature 2020-08-22 13:41:38 -06:00
James Betker
40bb0597bb misc 2020-08-18 08:50:24 -06:00
James Betker
0c98c61f4a Enable start_step to be specified 2020-08-15 18:34:59 -06:00
James Betker
2d205f52ac Unite spsr_arch switched gens 
Found a pretty good basis model.
 2020-08-12 17:04:45 -06:00
James Betker
bdaa67deb7 Misc 2020-08-12 08:46:15 -06:00
James Betker
1d5f4f6102 Crossgan 2020-08-07 21:03:39 -06:00
James Betker
3ab39f0d22 Several new spsr nets 2020-08-05 10:01:24 -06:00
James Betker
328afde9c0 Integrate SPSR into SRGAN_model 
SPSR_model really isn't that different from SRGAN_model. Rather than continuing to re-implement
everything I've done in SRGAN_model, port the new stuff from SPSR over.

This really demonstrates the need to refactor SRGAN_model a bit to make it cleaner. It is quite the
beast these days..
 2020-08-02 12:55:08 -06:00
James Betker
c8da78966b Substantial SPSR mods & fixes 
- Added in gradient accumulation via mega-batch-factor
- Added AMP
- Added missing train hooks
- Added debug image outputs
- Cleaned up including removing GradientPenaltyLoss, custom SpectralNorm
- Removed all the custom discriminators
 2020-08-02 10:45:24 -06:00
James Betker
f894ba8f98 Add SPSR_module 
This is a port from the SPSR repo, it's going to need a lot of work to be properly integrated
but as of this commit it at least runs.
 2020-08-01 22:02:54 -06:00
James Betker
eb11a08d1c Enable disjoint feature networks 
This is done by pre-training a feature net that predicts the features
of HR images from LR images. Then use the original feature network
and this new one in tandem to work only on LR/Gen images.
 2020-07-31 16:29:47 -06:00
James Betker
e37726f302 Add feature_model for training custom feature nets 2020-07-31 11:20:39 -06:00
James Betker
a7541b6d8d Fix illegal tb_logger use in distributed training 2020-07-23 09:14:01 -06:00
James Betker
dbf6147504 Add switched discriminator 
The logic is that the discriminator may be incapable of providing a truly
targeted loss for all image regions since it has to be too generic
(basically the same argument for the switched generator). So add some
switches in! See how it works!
 2020-07-22 20:52:59 -06:00
James Betker
8a9f215653 Huge set of mods to support progressive generator growth 2020-07-18 14:18:48 -06:00
James Betker
47a525241f Make attention norm optional 2020-07-18 07:24:02 -06:00
James Betker
8d061a2687 Add u-net discriminator with feature output 2020-07-16 10:10:09 -06:00
James Betker
1b1431133b Add DualOutputSRG 
Also removes the old multi-return mechanism that Generators support.
Also fixes AttentionNorm.
 2020-07-14 09:28:24 -06:00
James Betker
812c684f7d Update pixgan swap algorithm 
- Swap multiple blocks in the image instead of just one. The discriminator was clearly
  learning that most blocks have one region that needs to be fixed.
- Relax block size constraints. This was in place to gaurantee that the discriminator
  signal was clean. Instead, just downsample the "loss image" with bilinear interpolation.
  The result is noisier, but this is actually probably healthy for the discriminator.
 2020-07-10 15:56:14 -06:00
James Betker
5f2c722a10 SRG2 revival 
Big update to SRG2 architecture to pull in a lot of things that have been learned:
- Use group norm instead of batch norm
- Initialize the weights on the transformations low like is done in RRDB rather than using the scalar. Models live or die by their early stages, and this ones early stage is pretty weak
- Transform multiplexer to use u-net like architecture.
- Just use one set of configuration variables instead of a list - flat networks performed fine in this regard.
 2020-07-09 17:34:51 -06:00
James Betker
b2507be13c Fix up pixgan loss and pixdisc 2020-07-08 21:27:48 -06:00
James Betker
26a4a66d1c Bug fixes and new gan mechanism 
- Removed a bunch of unnecessary image loggers. These were just consuming space and never being viewed
- Got rid of support of artificial var_ref support. The new pixdisc is what i wanted to implement then - it's much better.
- Add pixgan GAN mechanism. This is purpose-built for the pixdisc. It is intended to promote a healthy discriminator
- Megabatchfactor was applied twice on metrics, fixed that

Adds pix_gan (untested) which swaps a portion of the fake and real image with each other, then expects the discriminator
to properly discriminate the swapped regions.
 2020-07-08 17:40:26 -06:00
James Betker
8a4eb8241d SRG3 work 
Operates on top of a pre-trained SpineNET backbone (trained on CoCo 2017 with RetinaNet)

This variant is extremely shallow.
 2020-07-07 13:46:40 -06:00
James Betker
a47a5dca43 Fix pixdisc bug 2020-07-05 21:57:52 -06:00
James Betker
188de5e15a Misc changes 2020-07-04 13:22:50 -06:00
James Betker
77d3765364 Fix new feature loss calc 2020-07-03 22:20:13 -06:00
James Betker
da4335c25e Add a feature-based validation test 2020-07-03 15:18:57 -06:00
James Betker
703dec4472 Add SpineNet & integrate with SRG 
New version of SRG uses SpineNet for a switch backbone.
 2020-07-03 12:07:31 -06:00
James Betker
ea9c6765ca Move train imports into init_dist 2020-07-02 15:11:21 -06:00