From 6ee9471b858031c83c961bf8c11a9ab9105172a4 Mon Sep 17 00:00:00 2001 From: Emmanuel Benazera Date: Wed, 4 Oct 2023 10:01:15 +0200 Subject: [PATCH] chore: release reset --- CHANGELOG.md | 431 --------------------------------------------------- 1 file changed, 431 deletions(-) delete mode 100644 CHANGELOG.md diff --git a/CHANGELOG.md b/CHANGELOG.md deleted file mode 100644 index b9c9eb72a..000000000 --- a/CHANGELOG.md +++ /dev/null @@ -1,431 +0,0 @@ -# joliGEN: Generative AI Toolset (Changelog) - -## 2.0.0 (2023-09-19) - - -### Features - -* add a server endpoint to delete files ([30b2143](https://github.com/jolibrain/joliGAN/commit/30b214341ea565dfe294f5c294f65281dc23d385)) -* add choices for all options ([ed43b82](https://github.com/jolibrain/joliGAN/commit/ed43b82fff25ad39be6f85a351dcf71dd8de1533)) -* add ddim inference ([0196134](https://github.com/jolibrain/joliGAN/commit/01961348b919ce4fab3759685182121383185a01)) -* add DDPM tutorial on the VITON-HD dataset ([c932d73](https://github.com/jolibrain/joliGAN/commit/c932d73698853a5f5760ec7e4d3b75d9b0c7c387)) -* add FastAPI server to run training ([f517462](https://github.com/jolibrain/joliGAN/commit/f5174624b93c2292457d8a315b1812e66a21a172)) -* add lambda for semantic losses ([aab53fe](https://github.com/jolibrain/joliGAN/commit/aab53feab5de2c7791ac1a945c5fa48be68589ed)) -* add LPIPS metric ([f1e0526](https://github.com/jolibrain/joliGAN/commit/f1e05266a96597e8d3ee456907ce55fc7d3511e1)) -* add miou compute to tests ([c0033ef](https://github.com/jolibrain/joliGAN/commit/c0033ef97c78dfc92745885896c074ec776ee5ee)) -* add new metrics ([f3c84cd](https://github.com/jolibrain/joliGAN/commit/f3c84cd8c4cd6e54d7c2c86853d01c713935b899)) -* add palette model ([b7db294](https://github.com/jolibrain/joliGAN/commit/b7db294a14003491dec17f9fb2a8e861ff4411e3)) -* add psnr metric ([7135458](https://github.com/jolibrain/joliGAN/commit/71354583731b55eeaeffd893639042aced6f9ec0)) -* add sampling options to test ([a2958dc](https://github.com/jolibrain/joliGAN/commit/a2958dc5edc8632bef0dee15c35dcdc5c2b97913)) -* add SRC and hDCE losses ([ddfcc97](https://github.com/jolibrain/joliGAN/commit/ddfcc97500fc2b63cff2f6f2e3b321c9e957bc3c)) -* add test for doc generation ([41526f8](https://github.com/jolibrain/joliGAN/commit/41526f80165f074961644154544da9201e4cc530)) -* add test on cycle_gan_semantic_mask ([3eeff76](https://github.com/jolibrain/joliGAN/commit/3eeff76d881a30d43fdf01f79ff72f11b3e11a6c)) -* add tests for reference image dataloaders ([ae6405e](https://github.com/jolibrain/joliGAN/commit/ae6405eb8d47191f86b45eb47abd844645810096)) -* added D noise to CUT with semantics ([31aa4a3](https://github.com/jolibrain/joliGAN/commit/31aa4a3222451cbe99199e0e734e16e11973d7cc)) -* added optimizers and options ([505cac2](https://github.com/jolibrain/joliGAN/commit/505cac22ab91612aa34013093ad2b62bed06159d)) -* allow control of projected discriminator interpolation ([dbffec5](https://github.com/jolibrain/joliGAN/commit/dbffec5a41fd2ace00885e8b8dfc4dd1b8346141)) -* allow ViT custom resolution at D projector init ([82e6e83](https://github.com/jolibrain/joliGAN/commit/82e6e830e84d20f2253740d7c30b755949b22d00)) -* **api:** display current commit at startup ([6f90be8](https://github.com/jolibrain/joliGAN/commit/6f90be83616a4298e5106c16d8fb4bf852ef081a)) -* **aug:** affine transforms for semantics ([170b0f8](https://github.com/jolibrain/joliGAN/commit/170b0f86e991b32b84f56dd9e8e6573c31e5181a)) -* **aug:** configurable online mask delta augmentation by x and y axis ([dfa6459](https://github.com/jolibrain/joliGAN/commit/dfa64595605f7db5b8f4cd6d57b853275194d4d2)) -* **aug:** select bbox category through the path sanitization functionality ([a8d3f48](https://github.com/jolibrain/joliGAN/commit/a8d3f488977271cc77425f4672a4b5f552aee657)) -* auto download segformer weights ([083cc5e](https://github.com/jolibrain/joliGAN/commit/083cc5ee2fe52531b1c3a00017a1b69440c246f3)) -* backward while computing MSE criterion loss ([1b87906](https://github.com/jolibrain/joliGAN/commit/1b87906233f325dd442573b6fbf7f4e43f6585c4)) -* bbox as sam prompt ([a39c5bd](https://github.com/jolibrain/joliGAN/commit/a39c5bdb48ab8e0093b3db04a33493dc672b169d)) -* bbox prompt for sam ([1fa9cae](https://github.com/jolibrain/joliGAN/commit/1fa9caec4180360522edc3930fe68f45e2c6b203)) -* bilinear interpolation of attention heads when dimension does not match, useful for segformer G ([eed9494](https://github.com/jolibrain/joliGAN/commit/eed9494bbae0d5be634afc774b64f75e1c6bda21)) -* bw model export ([8e43efa](https://github.com/jolibrain/joliGAN/commit/8e43efabafbd11010daecc0a4e8c8da169b5f39c)) -* check code format when PR ([eeb56cb](https://github.com/jolibrain/joliGAN/commit/eeb56cb853d3d964fb1ca02d920e027dc59b61d1)) -* choices for canny random thresholds ([9573fc1](https://github.com/jolibrain/joliGAN/commit/9573fc15a77d5fae45f11954cc68bac2b3fd154a)) -* class weights for semantic segmentation network with cross entropy ([4274f1e](https://github.com/jolibrain/joliGAN/commit/4274f1e1401dc649651ae2895c420162193bbfa8)) -* classifier training on domain B ([fa343c0](https://github.com/jolibrain/joliGAN/commit/fa343c00b431c8b938536ece4fd4355b96390652)) -* commandline saving ([6eb503e](https://github.com/jolibrain/joliGAN/commit/6eb503e47c55c195bba73d121e8ec920c857d4df)) -* commandline script for joligan server calls ([48ae23b](https://github.com/jolibrain/joliGAN/commit/48ae23bd108018d50a02e59ce3b4d34487744e71)) -* compute_feats for unet G ([9f1109e](https://github.com/jolibrain/joliGAN/commit/9f1109e5b2dc49370abca391e67ab0a1de845b6e)) -* conditioning for palette ([b9854ee](https://github.com/jolibrain/joliGAN/commit/b9854eed38652c6a89ad07bdd15020b6b4ac789b)) -* config json for client script ([174dce9](https://github.com/jolibrain/joliGAN/commit/174dce9c955534b8e5d386b74b3cf36b8e952b3a)) -* context for D ([b0d3c7b](https://github.com/jolibrain/joliGAN/commit/b0d3c7b554d45e0ddc0363e8a571bfd7f5047688)) -* contrastive classifier noise ([7193e0e](https://github.com/jolibrain/joliGAN/commit/7193e0e3d91879842abe654a335ad3ca3787f748)) -* contrastive loss for D ([deb2ec4](https://github.com/jolibrain/joliGAN/commit/deb2ec4f7388c2c13b7519ac625294c0a11ba27f)) -* cut_semantic model ([b20a943](https://github.com/jolibrain/joliGAN/commit/b20a943c0d7385cfdf8ab3cbaf8ad2d5d9003800)) -* D accuracy ([26ead91](https://github.com/jolibrain/joliGAN/commit/26ead91a7a85aa83c3bf39c29e9022e5da5a7f76)) -* **data:** random noise in images for object insertion ([42cf13d](https://github.com/jolibrain/joliGAN/commit/42cf13d8f2236ff2f68eb03bad64a39aed4292f8)) -* DDP ([68f24da](https://github.com/jolibrain/joliGAN/commit/68f24da0db6c6f2d8ea9c5f9c7b7c99e177c3303)) -* deceiving D for GAN training ([2e2113f](https://github.com/jolibrain/joliGAN/commit/2e2113fdb02bca4731515cc68007083a0df7f6b6)) -* depth model as projector ([10ffc28](https://github.com/jolibrain/joliGAN/commit/10ffc28e1b663510a54f5a7cebb60f2565cc92fa)) -* depth prediction and depth discriminator ([01bc62b](https://github.com/jolibrain/joliGAN/commit/01bc62b5f70aa30be11dec642295a06f20ad6ee7)) -* diff augment ([054509c](https://github.com/jolibrain/joliGAN/commit/054509c47020ff96f95340c330903d69208e7fd8)) -* diffusion inference with old and new models ([9c4c5a9](https://github.com/jolibrain/joliGAN/commit/9c4c5a9f0839b61e2cb21c25460e29e7a39063fb)) -* display augmented images ([2126253](https://github.com/jolibrain/joliGAN/commit/2126253081c8d66daafa0a572a61854a04ebd0fb)) -* display test images ([a1de083](https://github.com/jolibrain/joliGAN/commit/a1de083c78da83da6523bb5f3fbadf9900376b1d)) -* doc options auto update ([1b08f92](https://github.com/jolibrain/joliGAN/commit/1b08f923726d9c31f85b0c7d75c56dc437af441c)) -* **doc:** add JSON config examples ([5332213](https://github.com/jolibrain/joliGAN/commit/5332213d4569dce008bb1e5eb1cc104f22e80fb2)) -* **doc:** basic server REST API ([a757d17](https://github.com/jolibrain/joliGAN/commit/a757d17c3557dc5a4c46c50a4f74ec0bd61587fb)) -* **doc:** datasets ([dfe2343](https://github.com/jolibrain/joliGAN/commit/dfe2343e0adfe5f421459fda399b3e45fca24e81)) -* **doc:** DDPM conditioning training and inference examples ([e694a29](https://github.com/jolibrain/joliGAN/commit/e694a2972fa1527ac7a0b59ceabe00e8f9661d81)) -* **doc:** models ([be1fe34](https://github.com/jolibrain/joliGAN/commit/be1fe341c6050bbcb70be83fe05403988806889b)) -* **doc:** refactored README with links to documentation ([b5bf121](https://github.com/jolibrain/joliGAN/commit/b5bf1216ba68c4b76e4ac5bc9de8146a0ec39870)) -* **doc:** reference image conditioning ([70aeb32](https://github.com/jolibrain/joliGAN/commit/70aeb320f47c841040555345ffee693802bf4ddb)) -* **doc:** remove overview ([2360527](https://github.com/jolibrain/joliGAN/commit/2360527da39936654935f188d2dc2b198eb072c8)) -* **doc:** server, client, docker ([68a5b96](https://github.com/jolibrain/joliGAN/commit/68a5b964de8e8be6e02ce8a1041bd8e32f56d006)) -* **doc:** tips ([3fea9ca](https://github.com/jolibrain/joliGAN/commit/3fea9ca67c2cdf54f6b524bb66d2b089ae340e3d)) -* **doc:** training ([a4b720d](https://github.com/jolibrain/joliGAN/commit/a4b720dc590e20f5b32383f964303e1dae7e3298)) -* **doc:** update inference models and examples ([3c43a7b](https://github.com/jolibrain/joliGAN/commit/3c43a7bc4a343f38a3bbb110c3b5694f88f17f61)) -* **doc:** updated FAQ ([88b417c](https://github.com/jolibrain/joliGAN/commit/88b417cc7d7b415b4ad0d6fa8ba476c42affc9ac)) -* **doc:** updated model export ([e692f78](https://github.com/jolibrain/joliGAN/commit/e692f786c225493638f635c9a882c2cabe6fadb5)) -* edge detection techniques ([78202ea](https://github.com/jolibrain/joliGAN/commit/78202ea4cb9494cd254a60b79080c5f7db5f6654)) -* export for unet_mha ([b4c3cfd](https://github.com/jolibrain/joliGAN/commit/b4c3cfde314f1aef5b59e4d4c4dfec7e8f82ac2b)) -* first recut model ([aaa4069](https://github.com/jolibrain/joliGAN/commit/aaa4069e6a52a3a3ec253b001f77e7bfe4d2cc64)) -* first test ([4ac8cd9](https://github.com/jolibrain/joliGAN/commit/4ac8cd9177e27c2e734af09cd081e5c54bdb07ec)) -* fixed bbox size for online creation and bbox size randomization ([5cd6227](https://github.com/jolibrain/joliGAN/commit/5cd6227cd4ade47e32a1a491d10fcce3f4153ce7)) -* G weights export during traing ([4c045e6](https://github.com/jolibrain/joliGAN/commit/4c045e64da2e5d0642f6fbab424c9e7880220751)) -* generic image augmentation ([d2ceb81](https://github.com/jolibrain/joliGAN/commit/d2ceb8118c06c6ab4a4d7761ff35cf0f2908d29a)) -* get_schema uses default instead of choices (not always available) ([a779b2a](https://github.com/jolibrain/joliGAN/commit/a779b2af429220edc0898ea741b23c30d4d143dd)) -* global models ([3819a40](https://github.com/jolibrain/joliGAN/commit/3819a40621f00cbd8120db0d4a31e44d20376dab)) -* inverted mask for automatic background inpainting ([39f9ff2](https://github.com/jolibrain/joliGAN/commit/39f9ff24ab7c518f7573f346903ee22c66bc8b86)) -* itersize option for cut_model and cut_semantic_mask_model ([b3b9e7a](https://github.com/jolibrain/joliGAN/commit/b3b9e7abca3f02f34de259bf3193f193ec3d9cdd)) -* list all available models in help ([06a6259](https://github.com/jolibrain/joliGAN/commit/06a625901c4abc10cddaf04f4dedf0d5aa3f349c)) -* load segformer torchscript weights ([672b341](https://github.com/jolibrain/joliGAN/commit/672b341c2046cd39212568ec5a7aee1ab7a702a8)) -* loss values saved in json file and file to display it later ([d0fa9a5](https://github.com/jolibrain/joliGAN/commit/d0fa9a5c73b2070b39a2d2c400cb138c114a71a4)) -* madgrad optimizer ([1c410f2](https://github.com/jolibrain/joliGAN/commit/1c410f25aea786c8cf069d2164cc6538b7fdd8b4)) -* metrics for testset ([c875f2b](https://github.com/jolibrain/joliGAN/commit/c875f2bcf965e9cbbec914bcb77c18805f15ea49)) -* miou compute for f_s pred ([5851566](https://github.com/jolibrain/joliGAN/commit/5851566714683e02a29e29f9995f326ef5d5ca6b)) -* **ml:** ability to train in wavelet space, similar to 2102.06108 ([28077d4](https://github.com/jolibrain/joliGAN/commit/28077d4bc7863a6201d8d37546ed9ba17b18f86a)) -* **ml:** added semantic threshold option ([03f33a2](https://github.com/jolibrain/joliGAN/commit/03f33a2e8e3293f1fac7b2ae017488c9b9a16684)) -* **ml:** adding MoNCE contrastive loss to CUT, 2203.09333 ([26f2e9d](https://github.com/jolibrain/joliGAN/commit/26f2e9dc2fc194b6a58ca59295866684ff9f3141)) -* **ml:** classifier-free guidance for image-to-image DDPM ([a6425ef](https://github.com/jolibrain/joliGAN/commit/a6425efcf9027f54f463ed5bd7888731d75f6672)) -* **ml:** controle over image generation with diffusion in-painting model ([0a5ed86](https://github.com/jolibrain/joliGAN/commit/0a5ed8663d960a375c7cd8b69e4945791061d8c9)) -* **ml:** DDIM scheduler ([443e7d7](https://github.com/jolibrain/joliGAN/commit/443e7d7b2ea9d8a61f728a2eb74c6cebc736ccf8)) -* **ml:** diffusion super-resolution for palette ([66c591f](https://github.com/jolibrain/joliGAN/commit/66c591f0fbaf64b15abf7b33f471d3c13f48a091)) -* **ml:** diffusion-based augmentation for D, 2206.02262 ([5cf1f1b](https://github.com/jolibrain/joliGAN/commit/5cf1f1bc6082daccb89b59f75b7de2fefc833e08)) -* **ml:** Discriminator based on SAM ([e38d501](https://github.com/jolibrain/joliGAN/commit/e38d501a71e6b9f65522af1a58010bf183333425)) -* **ml:** efficient UNet / 2205.11487 ([d690ab4](https://github.com/jolibrain/joliGAN/commit/d690ab451f7e0357328e1c89c2657abb0f88a5e9)) -* **ml:** exponential moving average for G ([ad33796](https://github.com/jolibrain/joliGAN/commit/ad33796d2745072dd7f9350f4861893ba81e4514)) -* **ml:** GAN mask generator with sam refined target ([0cd1ee9](https://github.com/jolibrain/joliGAN/commit/0cd1ee915ee649dd0dc40a84de4d7e18970efd0f)) -* **ml:** grayscale support for unaligned with semantics ([a79548f](https://github.com/jolibrain/joliGAN/commit/a79548ffef7634e222a1ad24855aec725ac01dfb)) -* **ml:** initial sam generated masks for unaligned datasets ([5c64440](https://github.com/jolibrain/joliGAN/commit/5c64440e6d6a02fbbd243b1627043ea0774dc8e7)) -* **ml:** ittr generator control over the number of blocks ([ea7390e](https://github.com/jolibrain/joliGAN/commit/ea7390e4d891f2eab7dc74f35dbfc7820c846f3b)) -* **ml:** ITTR transformer arch for image2image - 2203.16015 ([16e2b5e](https://github.com/jolibrain/joliGAN/commit/16e2b5eb318aea1df4761d00a7d532c00000f69c)) -* **ml:** L1 loss for palette diffusion model ([d2bb2a5](https://github.com/jolibrain/joliGAN/commit/d2bb2a5056a77fa5c345bdefd0ec684faadd850b)) -* **ml:** Lion optimizer ([4b99204](https://github.com/jolibrain/joliGAN/commit/4b9920462c0df264838ce1368e9d188b304f8a2f)) -* **ml:** mask conditioning for palette ([8e71bb7](https://github.com/jolibrain/joliGAN/commit/8e71bb7fcab6484f22527e0a16d929b593e3e627)) -* **ml:** mask generation across domains for GAN semantics ([cf4890b](https://github.com/jolibrain/joliGAN/commit/cf4890bc4293b2d5597d12723ba5abc90549175c)) -* **ml:** query-selected attention for contrastive patches, 2203.08483 ([6a13e44](https://github.com/jolibrain/joliGAN/commit/6a13e44b1fb38abfa25d85d890dcd77c18d8ab23)) -* **ml:** torchvision model support for semantic classification ([341205c](https://github.com/jolibrain/joliGAN/commit/341205c447c1363e74cdb20eb2f614e555997d51)) -* **ml:** UNet/UViT resnet block finer control ([9bb979c](https://github.com/jolibrain/joliGAN/commit/9bb979c9d40ff42d76537695a568669deb66149b)) -* **ml:** UViT for GAN ([b1b3607](https://github.com/jolibrain/joliGAN/commit/b1b3607a73ce377fa5555f20c6012a1b38ab2929)) -* **ml:** ViT-14L as projected discriminator ([dab27d2](https://github.com/jolibrain/joliGAN/commit/dab27d259e51011b19b7acb550578dac1e1dfce9)) -* mobile attention resnet ([7b8fd87](https://github.com/jolibrain/joliGAN/commit/7b8fd8719bccc7fa43048bfc2eca7e0d7e73f023)) -* MobileSam implementation ([ea11745](https://github.com/jolibrain/joliGAN/commit/ea117458aa7da231f7448e57d78c5a8115c5d50c)) -* MobileSam implementation ([e0d3a67](https://github.com/jolibrain/joliGAN/commit/e0d3a67da0f1e717df59ac035eff76cba62cc945)) -* more options for image generation with diffusion ([cd08411](https://github.com/jolibrain/joliGAN/commit/cd08411cca233677cfd99a4d56716af05c98ad76)) -* MSE identity loss for cut models ([28bfbfd](https://github.com/jolibrain/joliGAN/commit/28bfbfd2a85c376ba03656e09c5d98ba81995349)) -* multi head attention unet generator ([9e4c232](https://github.com/jolibrain/joliGAN/commit/9e4c232dfcd9604a37dbd2f3ab1c1c95251fd3e8)) -* multimodal generator with cLR-GAN strategy ([c3919b9](https://github.com/jolibrain/joliGAN/commit/c3919b91184b7f9c4deb7cf9a0b0d1e5acfc8ba7)) -* multiple D support ([578f709](https://github.com/jolibrain/joliGAN/commit/578f709d524bb9b9a4b6617e653a6154d58a686f)) -* multiple linearly fitted D via vision aided ensembling, https://arxiv.org/abs/2112.09130 ([9fbca44](https://github.com/jolibrain/joliGAN/commit/9fbca44734e3746d83dac561510a045c8773a5aa)) -* multiscale loss for diff ([0af1a80](https://github.com/jolibrain/joliGAN/commit/0af1a80722c894abe227b1a0ec93307a9a40a6a9)) -* nb of images used for FID computation can be chosen ([233634b](https://github.com/jolibrain/joliGAN/commit/233634bb4dcc0d8413464939d36925b2efde5992)) -* online dataset creation ([9883eac](https://github.com/jolibrain/joliGAN/commit/9883eacd72490a3212dcbb6ce82ead38a7d84604)) -* option for number of inferences (palette model) ([7e8bcc9](https://github.com/jolibrain/joliGAN/commit/7e8bcc9317fe504fef7dc76717c082b2e51d2984)) -* option to choose metrics to compute ([ab1074a](https://github.com/jolibrain/joliGAN/commit/ab1074a16fd15d234a2940fe51d12f8a4acf8185)) -* option to choose norm in unet_mha ([77d0161](https://github.com/jolibrain/joliGAN/commit/77d016178604e7c678e64f2ff8b57faa496bad9f)) -* option to compute semantic G loss on f_s output ([9ef922d](https://github.com/jolibrain/joliGAN/commit/9ef922d652fc4531cbcd8d2df0ddc76858786f1a)) -* option to display real image - fake image ([b8a134a](https://github.com/jolibrain/joliGAN/commit/b8a134a6f95c600daf4954ea165d95e1c1c1c7a8)) -* option to go through resnet blocks twice (backward compatibility) ([1b3330b](https://github.com/jolibrain/joliGAN/commit/1b3330beb8d98a1e24ef1ee9d472c16f4cd51c8f)) -* option to select embedding network for ref conditioning ([2c9933d](https://github.com/jolibrain/joliGAN/commit/2c9933d8340e29c039894c11b2d3258ab959496f)) -* option to use a different f_s for domain B ([6c55c26](https://github.com/jolibrain/joliGAN/commit/6c55c269e362b0c78f748046414e69a4f2f84ff6)) -* options for aspect ratio and augmentation of cond image ([7bf9c95](https://github.com/jolibrain/joliGAN/commit/7bf9c95ec99833495aa7d6cd2ec77c95d0503ced)) -* options.rst gen script ([e13a9d8](https://github.com/jolibrain/joliGAN/commit/e13a9d8ef5a2a2d4a20ced67e8719ed1c0993f00)) -* precommit black formatting ([ef0c883](https://github.com/jolibrain/joliGAN/commit/ef0c8830029dd071ff2a7beef82ace0fa9d55579)) -* pretrained segformer backbone (mit) as feature extractor for projected D ([7566c8a](https://github.com/jolibrain/joliGAN/commit/7566c8a50da92dab2796991c07dd3316ba480f5c)) -* previous frame as y cond in palette ([99de095](https://github.com/jolibrain/joliGAN/commit/99de095108181263637455a3287032c276d5ad9e)) -* projected discriminator ([92c0aa0](https://github.com/jolibrain/joliGAN/commit/92c0aa0315875c112d65c7d5275728f753d691db)) -* quickstart ([0aa50e4](https://github.com/jolibrain/joliGAN/commit/0aa50e4f0212810dfa7f8e666c6f5831107025a0)) -* quickstart update ddpm ([98cef11](https://github.com/jolibrain/joliGAN/commit/98cef114c4fb753fe3c9e4e2bfd937383ddccbd1)) -* random perspective and bilinear interpolation with diffaug ([d2633c8](https://github.com/jolibrain/joliGAN/commit/d2633c80461f908d0b155711953a892eb2c1cf77)) -* readthedocs initial commit ([2a57b6f](https://github.com/jolibrain/joliGAN/commit/2a57b6fde3587787204c882dd461a7012e6907fa)) -* reference image conditioning ([8618129](https://github.com/jolibrain/joliGAN/commit/8618129a28ee8c54577242f5fbf1adecef35d420)) -* rename cur_mask ([53566c6](https://github.com/jolibrain/joliGAN/commit/53566c6fb5765c03aeeed72942f5a5b5c80fbfa4)) -* reorganizing display of images ([bed57bd](https://github.com/jolibrain/joliGAN/commit/bed57bdb3eef3eff60abf7613aa8f2bf4dcf5f65)) -* resnet G for diffusion ([6d1cddc](https://github.com/jolibrain/joliGAN/commit/6d1cddc56ea61c20afdb61cfabe4dea94458746e)) -* resnet_attn for cut model ([7d01972](https://github.com/jolibrain/joliGAN/commit/7d01972fbdd1c81af9f1aa982da4ac8e752471ba)) -* sam for mask refinement and edges ([7e55165](https://github.com/jolibrain/joliGAN/commit/7e5516517a2bfd215cfd4edfe2ced5be32525011)) -* sanitized paths can be saved and loaded ([2f9f42f](https://github.com/jolibrain/joliGAN/commit/2f9f42f688a4f4d9968552d0758c22499dfecd19)) -* save latest images ([1f5fb69](https://github.com/jolibrain/joliGAN/commit/1f5fb69a6e45377ae556222c215db9144fca44d2)) -* saving options json file ([a9452ce](https://github.com/jolibrain/joliGAN/commit/a9452ce2e070f80685bc846d0884026c53f18ef7)) -* script to compute metrics ([f5ab623](https://github.com/jolibrain/joliGAN/commit/f5ab62349326d975457d608759bfdf82e61814b3)) -* script to generate diffusion video ([574e149](https://github.com/jolibrain/joliGAN/commit/574e14997b5595d9565d6c5b225a4c19f546cca4)) -* script to remove useless models weights ([543f280](https://github.com/jolibrain/joliGAN/commit/543f280c1790d7ba685c4b15a13b6a7d2ac299cd)) -* scripts to export generator to JIT and to transform a single image ([f3f5c37](https://github.com/jolibrain/joliGAN/commit/f3f5c3727099ba699dc1c73e185eaa4449fcefe7)) -* second D for all models ([7ec82be](https://github.com/jolibrain/joliGAN/commit/7ec82be31508695cdb2c21612cb60e7826fc404e)) -* segformer can be used for attention masks generation ([98a3d6e](https://github.com/jolibrain/joliGAN/commit/98a3d6e50cb0c62d5ab593cc79063322d132405c)) -* semantic loss on identity ([91df963](https://github.com/jolibrain/joliGAN/commit/91df96374c601803e0203b526d39ec2a3935d93e)) -* **server:** add commit number in swagger documentation ([3c96bc0](https://github.com/jolibrain/joliGAN/commit/3c96bc013a65ca7663c30b7ac1932b2f6a139f1c)) -* **server:** add synchronous training ([5e3291e](https://github.com/jolibrain/joliGAN/commit/5e3291e8eefe55628e4f346c81012849af1678ef)) -* **server:** group options into categories ([9358007](https://github.com/jolibrain/joliGAN/commit/9358007063aff13f0da23aa7b2da6ed5aa7f0994)) -* set mininum context via crop/bbox ratio ([094b213](https://github.com/jolibrain/joliGAN/commit/094b21332b8109578dd0781e93820e1817a315e3)) -* single bbox per crop for inpainting while training ([ac2b4d6](https://github.com/jolibrain/joliGAN/commit/ac2b4d6dfd1fda845580e5977a686c318d8f41fd)) -* sphinx rtd theme initial ([6b2ac14](https://github.com/jolibrain/joliGAN/commit/6b2ac145fe27bdbb051ccb8d04d66ce5dc1e97c0)) -* stylegan network from CUT repo ([9e36c55](https://github.com/jolibrain/joliGAN/commit/9e36c555d26f1398114fd39fe92370e14f2e04b4)) -* support for amp on forward ops ([dffbb36](https://github.com/jolibrain/joliGAN/commit/dffbb36cf6bb5b2cf247001acebe4287d5e22e1a)) -* support TensorRT generator models in inference with DeepDetect ([980ad49](https://github.com/jolibrain/joliGAN/commit/980ad49e5b4ba89c1ec3bec5151e01996bc697bb)) -* sync loss only when printing ([ff8de5e](https://github.com/jolibrain/joliGAN/commit/ff8de5e3c7ef215a4a758a4f297bc34d1bcf3c88)) -* temporal criterion ([c0dbf78](https://github.com/jolibrain/joliGAN/commit/c0dbf783b615810acdb129f5761db7b462f2a4ba)) -* temporal discriminator ([154da72](https://github.com/jolibrain/joliGAN/commit/154da72a6058ccf7616047b22c1a7fe3000eb7ed)) -* **temp:** reduce image size during test ([7a7e4a3](https://github.com/jolibrain/joliGAN/commit/7a7e4a3c331a727f35d9eaea09e99a7fd0400744)) -* tests with different f_s net ([8d99c72](https://github.com/jolibrain/joliGAN/commit/8d99c724295d13fb33217ec83709e161e0e0bdb1)) -* tf32 support (with Ampere GPUs) ([c65622c](https://github.com/jolibrain/joliGAN/commit/c65622c02e1b6b0aef3d3ec47866167a058f88a9)) -* timm models as feature extractor for projected D ([498a11c](https://github.com/jolibrain/joliGAN/commit/498a11cd7c649022e0cd07296291c8f3042ad12d)) -* use label 0 in domain B for object removal ([d3e2399](https://github.com/jolibrain/joliGAN/commit/d3e2399695c3a1c78bfd46902658ecc966d1f439)) -* use the trained discriminator to rank generated images ([65d6416](https://github.com/jolibrain/joliGAN/commit/65d6416b2dcc261d016b4f198dc7a9d5f0cb5db4)) -* use torchvision.transforms for differentiable augmentations ([dc46733](https://github.com/jolibrain/joliGAN/commit/dc46733c57b79a4d11168a24e252b34a57a36000)) -* UViT for diffusion ([b427fa8](https://github.com/jolibrain/joliGAN/commit/b427fa89618e37311f0b3922f24c6c44114c311f)) -* working inference tutorials ([23c8ab6](https://github.com/jolibrain/joliGAN/commit/23c8ab60496f2d3983fea508447514f25a8de316)) - - -### Bug Fixes - -* 1 iteration = 1 image ([0b096f8](https://github.com/jolibrain/joliGAN/commit/0b096f8f5adf205aeb5dd33106d7fe9d4efc2dd9)) -* adam/radam beta1 default value ([db0e2dd](https://github.com/jolibrain/joliGAN/commit/db0e2ddae1c52c28d75aa085514dbc5145e83739)) -* add clamping in tensor2im function ([9537eac](https://github.com/jolibrain/joliGAN/commit/9537eac48dc1c9e47232a11d7325c75cf34c2004)) -* add einops to requirements_github_actions ([b30895d](https://github.com/jolibrain/joliGAN/commit/b30895d9e52789e4ded843688f2d5305551de7c0)) -* add exec flag to run.sh ([f6350d7](https://github.com/jolibrain/joliGAN/commit/f6350d78a73a2d705a22f80d97b6565f4372a3db)) -* add missing architectures to options doc ([5c31e1f](https://github.com/jolibrain/joliGAN/commit/5c31e1ffda7c5c7671318156c18ac1d76604a500)) -* add missing test file ([9e520cd](https://github.com/jolibrain/joliGAN/commit/9e520cd1d26a3741029243168895aa3f045d3866)) -* add mmcv to requirements for doc generation ([2c8ef04](https://github.com/jolibrain/joliGAN/commit/2c8ef046789e201b6f44c98b0c3c351580113bd2)) -* add mmseg to requirements for doc generation ([9d83ffe](https://github.com/jolibrain/joliGAN/commit/9d83ffe1d2746eecacac338fbbe2f40ee4c40f0d)) -* add padding type to jit export ([ae3e23c](https://github.com/jolibrain/joliGAN/commit/ae3e23c13b18314088a29b33173bda9ca65454d8)) -* add phase to dataloaders arg ([3992136](https://github.com/jolibrain/joliGAN/commit/39921360d41350df0c1f3fe357e32a6481b62b8c)) -* add reflectionpad to fix resnet ([e702930](https://github.com/jolibrain/joliGAN/commit/e702930d67eeca2c29f9fa7136eba9e31496beec)) -* add tqdm to requirements ([6e34f72](https://github.com/jolibrain/joliGAN/commit/6e34f7250d4745f5df7a9dea405b5ad81ce0e9ee)) -* add type for nb_mask options ([7bce61c](https://github.com/jolibrain/joliGAN/commit/7bce61c38eb0ab9d91b8dff4efc106696df653e2)) -* add vision aided to doc auto requirements ([5fbb13f](https://github.com/jolibrain/joliGAN/commit/5fbb13f291c4ea5842343b1935bf17daaebbe8db)) -* add wget to github action requirements ([ad7b3d7](https://github.com/jolibrain/joliGAN/commit/ad7b3d7892352b82023a2eae699048d8e8445432)) -* all_classes_as_one option ([df01988](https://github.com/jolibrain/joliGAN/commit/df01988ad1494b0bf78db9a56d08c1fc816a65ef)) -* allow for domain B to have no semantic labels ([42604aa](https://github.com/jolibrain/joliGAN/commit/42604aa6ac7998a8c41d29476b96aeb16b10ad52)) -* allow for image reading failures ([d7f5d08](https://github.com/jolibrain/joliGAN/commit/d7f5d08d95a6f38d7e40fdbd84dc710d1f55464b)) -* allow to freeze networks with DDP ([d12bd7e](https://github.com/jolibrain/joliGAN/commit/d12bd7eacbc8ea4d0a52ccfc312a528093f3c876)) -* alternate optimizing G and E ([d76ba1b](https://github.com/jolibrain/joliGAN/commit/d76ba1bf4c5c7328f25313caf02900189907ed5f)) -* APA_img is only added to visual names APA option is true ([309592e](https://github.com/jolibrain/joliGAN/commit/309592ed3249306e020a075ccb79ca9a73278120)) -* attention masks display ([cc5b6b8](https://github.com/jolibrain/joliGAN/commit/cc5b6b80304791e863a065330ddddb54a4470b1d)) -* auto resize of attention masks ([e3bf86a](https://github.com/jolibrain/joliGAN/commit/e3bf86acfb54cf4f52e10642d33ca3b2dafb53b0)) -* avoid lambda in RandomImgAug that prevents DDP ([a206f91](https://github.com/jolibrain/joliGAN/commit/a206f91d6202ca1225df644327c60f15fe0daf6e)) -* B label clamping override + display colormap ([6fc1dd4](https://github.com/jolibrain/joliGAN/commit/6fc1dd4e4d348981e8b533c2e29116ae2ca14a00)) -* batchnorm for single gpu ([93b164c](https://github.com/jolibrain/joliGAN/commit/93b164c31b3451115a232235da91755e2d4410e9)) -* bbox and mask for diffusion generation script ([390fe5d](https://github.com/jolibrain/joliGAN/commit/390fe5de2ef46d8b9e626838ad1b1c2d175e52db)) -* bbox ref idx from crop_image return in inference code ([be0a1b5](https://github.com/jolibrain/joliGAN/commit/be0a1b56c5416444e166ab1f8bc727c3c4ff9936)) -* broken export onnx script ([8080036](https://github.com/jolibrain/joliGAN/commit/80800365d033376555fd460932023de6481f3888)) -* broken test + bbox_ref_id option in diffusion single image generation script ([179b3ca](https://github.com/jolibrain/joliGAN/commit/179b3ca044dbd747345ca62bbf32004e35e7b7b5)) -* build docker file ([c055385](https://github.com/jolibrain/joliGAN/commit/c0553853f052df289b3dc435ac47ac4e28d541b8)) -* catch image reading errors ([0153eb4](https://github.com/jolibrain/joliGAN/commit/0153eb4a702aeeb8cfb6ade6aa7f40d0e6f26b2b)) -* circumvent isTrain in base options ([2eb905e](https://github.com/jolibrain/joliGAN/commit/2eb905e18a33277ae7e7128fde86a98da04f5937)) -* clean code ([b83282b](https://github.com/jolibrain/joliGAN/commit/b83282b0fffc58a17da919041377d4967a70d266)) -* cls net was created with f_s options ([70e9009](https://github.com/jolibrain/joliGAN/commit/70e9009eb74fa3d4a3b58e3407bfec5e07eeb4d2)) -* cls_class_weights default to [] ([ad719f6](https://github.com/jolibrain/joliGAN/commit/ad719f6966a8af74be06ef84c9f94a1cc8ed1acc)) -* computation of out_mask loss when multigpu ([31b7294](https://github.com/jolibrain/joliGAN/commit/31b729429b43f949dfba39e2fe9b2dd16bf0fc5c)) -* compute_temporal_fake defined twice ([83e0d2d](https://github.com/jolibrain/joliGAN/commit/83e0d2d498b98cdddf77b7c07cf5ab4aa5a1d7e7)) -* cond inference ([77c0444](https://github.com/jolibrain/joliGAN/commit/77c0444177039512033495a4ef1acfb6a8621f59)) -* context pixel for sanitize paths ([d18cee9](https://github.com/jolibrain/joliGAN/commit/d18cee993de1ff49cacbcd9f9de065e59b2e11f0)) -* correct urls in quickstart_gan.rst ([580353f](https://github.com/jolibrain/joliGAN/commit/580353f3b377b1f437f723e8b99e50224b45833b)) -* cpu device for single image generation script ([bb3c70c](https://github.com/jolibrain/joliGAN/commit/bb3c70c776e160a5ee907e86c501518372f13e3d)) -* criterion temporal loss display ([48d4c27](https://github.com/jolibrain/joliGAN/commit/48d4c276d6f15c16be3fa8a99f9786effe9f8598)) -* criterionIdt is not used by cut ([3665410](https://github.com/jolibrain/joliGAN/commit/3665410682329e6ce9fae51b84b55a2f92fe8b09)) -* crop_image for non square images and large bboxes ([afdde75](https://github.com/jolibrain/joliGAN/commit/afdde75fe61e376a845cee49ff004fb5d9951bee)) -* cyclegan semantic mask failure with f_s ([0688974](https://github.com/jolibrain/joliGAN/commit/06889742b5779266a5c9eed65613d69570678338)) -* D global requires grad were not set true before backward ([6a55914](https://github.com/jolibrain/joliGAN/commit/6a559146621f3f8ddae172ddad3156c7668e5b44)) -* D noise in cut base model ([c6a2114](https://github.com/jolibrain/joliGAN/commit/c6a21145dad2cabcff06f8b82c852930f78b825c)) -* D_global optimization ([284d8ae](https://github.com/jolibrain/joliGAN/commit/284d8ae30efd290ba082a014a93df4e6fbe22a8e)) -* data dept init ([8a0707e](https://github.com/jolibrain/joliGAN/commit/8a0707e8ca33eec220e7ae9bcaa30cca958b7695)) -* dataaug_D_diffusion option ([70628d1](https://github.com/jolibrain/joliGAN/commit/70628d1c1917ed5a51d01c10017c6c0e852f943c)) -* dataset created once, one dataloader per gpu ([70348ff](https://github.com/jolibrain/joliGAN/commit/70348ff8d41a266cd7fe36cc29b9c15fd4a60252)) -* DDIM restoration when batch_size > 1 ([ccb445b](https://github.com/jolibrain/joliGAN/commit/ccb445b6067344722e49ceabbf88b7b9b775e10e)) -* ddpm inference with sam masks + inverted masks ([3055910](https://github.com/jolibrain/joliGAN/commit/3055910cd9fe24edbdccacf09febaff00687b9a0)) -* default --cls_config_segformer value ([caadf7a](https://github.com/jolibrain/joliGAN/commit/caadf7ac9a53f9bec6f37d90bd547fb3ef8307db)) -* default nb_mask_input value ([632a0f6](https://github.com/jolibrain/joliGAN/commit/632a0f6d458ce806f4f61c244454343d11d54f49)) -* dependencies to docker build file ([76eb760](https://github.com/jolibrain/joliGAN/commit/76eb76039e329c98d82388eed8da8a581cc7343c)) -* diff aug options ([a337990](https://github.com/jolibrain/joliGAN/commit/a337990de4b58a54154abdf165f931c533132b7d)) -* diffusion generation script is already multimodal ([d49e66a](https://github.com/jolibrain/joliGAN/commit/d49e66a236b178f022b1911acbdc488bf0bd7ac6)) -* diffusion inference with mask_in ([dea27ac](https://github.com/jolibrain/joliGAN/commit/dea27acb4e9a820761976b7add2fa09b022ae84a)) -* diffusion video generation fixes to generate API changes ([2d74f7a](https://github.com/jolibrain/joliGAN/commit/2d74f7a1edd1d4be644a24c9a400b245583a86dc)) -* diffusion with temporal data loader ([c328acb](https://github.com/jolibrain/joliGAN/commit/c328acb1aeeefa1d6914a02a678dab9e70dbde5f)) -* display for palette model ([2a98cf7](https://github.com/jolibrain/joliGAN/commit/2a98cf7a31147521afa7f201b16d54bd793caca3)) -* doc_gen for palette model ([32ef0fa](https://github.com/jolibrain/joliGAN/commit/32ef0fa939e6c1f4795cd05e66657d4fbe801dff)) -* **doc:** badly formatted title ([8b26c87](https://github.com/jolibrain/joliGAN/commit/8b26c8729cd363c6928c38df473befc883b0794b)) -* docker server source image ([b29eb4f](https://github.com/jolibrain/joliGAN/commit/b29eb4f31c8d2498fc5835d978df9e744426a5b3)) -* **doc:** missing losses ([f74c373](https://github.com/jolibrain/joliGAN/commit/f74c373e644b407d622171d20ddcc0d3793d35b8)) -* **doc:** missing models ([a9bddd4](https://github.com/jolibrain/joliGAN/commit/a9bddd4147b4a24f1ab4bba47b19cfba49de0984)) -* **doc:** title format ([f580d88](https://github.com/jolibrain/joliGAN/commit/f580d889aef69484ac2923a940ac42e0233a4d85)) -* export & inference scripts for new encoder/decoder architectures ([7993544](https://github.com/jolibrain/joliGAN/commit/79935442739a6a219da6763adcaef22e5c88acf7)) -* f_s backward when multigpu ([f8b6067](https://github.com/jolibrain/joliGAN/commit/f8b6067664cb84bd17b46b7a5e6de29b25db3e9b)) -* f_s zero_grad in cut_semantic_mask ([0e1c2b3](https://github.com/jolibrain/joliGAN/commit/0e1c2b3443711749f9fe2d4314e2df8723a94b40)) -* feats went through resnet blocks 2 times and rm old code ([7d768b3](https://github.com/jolibrain/joliGAN/commit/7d768b39235320ef9ccdbf2beab71aed95743110)) -* fid for cut and for relative path ([57dbabf](https://github.com/jolibrain/joliGAN/commit/57dbabf6b070cbf789d50b52b03a77741945b9af)) -* flush print() in joligan_api.py ([0880d85](https://github.com/jolibrain/joliGAN/commit/0880d857c7507ab7a8a67e62c9d3655125c552a8)) -* force RGB in online creation ([15efa68](https://github.com/jolibrain/joliGAN/commit/15efa683287dafc78f0db092783f961a7e76f6ed)) -* force torch to stable 1.13.1 ([7fffaf0](https://github.com/jolibrain/joliGAN/commit/7fffaf0c343c0fc8818e9370e4b09ffabf239ae9)) -* gan_networks import for inference ([09f2969](https://github.com/jolibrain/joliGAN/commit/09f2969efb8044d891574d0a987c4b02675a4cfe)) -* generate doc script ([01f25af](https://github.com/jolibrain/joliGAN/commit/01f25af4481f1aeb1ebe131b5510e96d485fbe94)) -* generation script for non multimodal models ([a2ff20b](https://github.com/jolibrain/joliGAN/commit/a2ff20b03d072c2f50d08504533f99c28ebb0f98)) -* generation scripts help ([d1102ff](https://github.com/jolibrain/joliGAN/commit/d1102ff5e3da388d9ca298c3fe1fd6a3dc8d0e4a)) -* get schema for * options ([d7e33eb](https://github.com/jolibrain/joliGAN/commit/d7e33ebd14f4b5349b3c6f7661f0ef011210bc4f)) -* get schema should not compute base_gan_model ([5dd6662](https://github.com/jolibrain/joliGAN/commit/5dd66625331ae6b330c0e8afb9a19336caec48ae)) -* get_feats feature extractor for projected D ([212751d](https://github.com/jolibrain/joliGAN/commit/212751d580b468fb0030f404dab80540888ada23)) -* guidance scale option for diffusion inference ([846e4f5](https://github.com/jolibrain/joliGAN/commit/846e4f58ba2c0d29b49bc77eb45fa3837253dbde)) -* help command ([727fc15](https://github.com/jolibrain/joliGAN/commit/727fc157d17e5dc0c8edc9e5903b5eee09d77bb8)) -* help for diffusion inference script ([d85fcc9](https://github.com/jolibrain/joliGAN/commit/d85fcc9bfe51db50005da0986ee60ca7215d1318)) -* image generation script with torch model ([4b67c05](https://github.com/jolibrain/joliGAN/commit/4b67c05eba52f22fe606b8ab8623e4c059587786)) -* image size control in generation script ([01ad67d](https://github.com/jolibrain/joliGAN/commit/01ad67d6ebe0cfa028d9d43d461019512a9d6695)) -* image size is unused in gen_single_image ([b7ea7f4](https://github.com/jolibrain/joliGAN/commit/b7ea7f4d48058f600faf5c007bdeee441e37b003)) -* img visu for palette models ([668b253](https://github.com/jolibrain/joliGAN/commit/668b253682918b1cabb2f4363ea2a65576d9355b)) -* imgaug option with no mask ([f09d484](https://github.com/jolibrain/joliGAN/commit/f09d48450eb422de3afac11989a20a34751c0b02)) -* import signal module ([e0d1c52](https://github.com/jolibrain/joliGAN/commit/e0d1c523ee70a3699197b9b8bb26d6485924e1a1)) -* improved diffusion inference scripts, including video generation ([485fad9](https://github.com/jolibrain/joliGAN/commit/485fad90aa0fb444d08e952114c7f582cc002c3d)) -* in dataloader, warning when class > nclasses ([e9231eb](https://github.com/jolibrain/joliGAN/commit/e9231eb5af16f6f208100a43cc20134fbc3ef613)) -* in place gradient error with f_s ([283cb85](https://github.com/jolibrain/joliGAN/commit/283cb8561800c9fe6dd3e2b9777c5f8f8dc94134)) -* inference beta_schedule location ([ca910d8](https://github.com/jolibrain/joliGAN/commit/ca910d858f4d96f59b5fdcc547348f71148d1231)) -* inference when no mask cond ([e19505d](https://github.com/jolibrain/joliGAN/commit/e19505dd3d5e793f627da7bb174fc28ab79177cb)) -* inference_num for reference img ([e444ba1](https://github.com/jolibrain/joliGAN/commit/e444ba1c9e84164beb28aeecf9be77e77f7b8afd)) -* input_B_label loading when available ([83490b3](https://github.com/jolibrain/joliGAN/commit/83490b3b21c3e37e29157bfa5a91db42cfdff265)) -* int are considered as float if needed and warning if remaining keys in options ([ad6d99e](https://github.com/jolibrain/joliGAN/commit/ad6d99e0cef458ddae6e8bb1455a80fdcfa8b682)) -* inverted BtoA direction labels, moving direction to datasets ([4ac2845](https://github.com/jolibrain/joliGAN/commit/4ac2845faf4d1445d66932d0b63fcd662283ef26)) -* jenkins docker perms ([ebad13f](https://github.com/jolibrain/joliGAN/commit/ebad13f2bfab4617223d798be66ecd7e5e4731a5)) -* jit and onnx export scripts post refactor ([32e0912](https://github.com/jolibrain/joliGAN/commit/32e0912b72cf80d2e7b6c4c6e2364030658d0115)) -* joligan api launch training ([5ea68a4](https://github.com/jolibrain/joliGAN/commit/5ea68a464b7700ce8224f91c68e83cbefefdcee5)) -* lambda for GAN losses ([020435a](https://github.com/jolibrain/joliGAN/commit/020435af7831e8f7cf96d07bd50ba5a311b6fb30)) -* license year ([2424de0](https://github.com/jolibrain/joliGAN/commit/2424de03596ea334e991b00c65718176e0114147)) -* loading image with transparency ([3bcc7fd](https://github.com/jolibrain/joliGAN/commit/3bcc7fd15f83f503f7e6ce7d23565cfdd5b0d699)) -* loss values averaged accross gpus and compute last step ([f941595](https://github.com/jolibrain/joliGAN/commit/f94159589e843979b58f246421e7809dc70f9c8b)) -* macs to flops ([69a02db](https://github.com/jolibrain/joliGAN/commit/69a02dbf23e5539ac33f530d9d6597b2fe82aad2)) -* make aligned dataset work again ([61741cc](https://github.com/jolibrain/joliGAN/commit/61741cc9b62a5bede6b4f336df094d24987de770)) -* mask and class conditioning in generation script ([38631b8](https://github.com/jolibrain/joliGAN/commit/38631b82aa93e472936a6bab74514791fca44654)) -* mask clamping and display for diff models ([4b24fe4](https://github.com/jolibrain/joliGAN/commit/4b24fe41f947bf27065223ede893c2b525e34730)) -* mask display when more than one class ([fa21b2e](https://github.com/jolibrain/joliGAN/commit/fa21b2e18e29c2d3396675da9264f03acc6e4819)) -* mask_delta checks ([ef095f1](https://github.com/jolibrain/joliGAN/commit/ef095f1f004218deafce3017429ca55dd9498bf6)) -* mask_delta for inference diffusion ([69bf8ef](https://github.com/jolibrain/joliGAN/commit/69bf8eff9376774ae39435b0f7943499b97f7e77)) -* mask_delta ratio online check ([341dd73](https://github.com/jolibrain/joliGAN/commit/341dd73da288e177e2e4cdb8ab5c76b1b5c4089d)) -* missing get_weights for f_s ([835ab4f](https://github.com/jolibrain/joliGAN/commit/835ab4fe1e55fec52d09ea85c62f6018f3c31f5a)) -* missing help on projected D arch ([c65f2d2](https://github.com/jolibrain/joliGAN/commit/c65f2d2864bc047fdd7589a7826f1f65477d197c)) -* missing input_nc in f_s segformer constructor ([309f75f](https://github.com/jolibrain/joliGAN/commit/309f75fc8358c258e6703791aca242b840a607b0)) -* missing shape attribute in conditional ([9e3d365](https://github.com/jolibrain/joliGAN/commit/9e3d3659fdba9cde692a28058fba474730051799)) -* **ml:** allow semantic loss weight control ([8e552be](https://github.com/jolibrain/joliGAN/commit/8e552beb80a1a2804df0cf2d1111bc08eb3242a3)) -* **ml:** allow torchvision semantic model backbones to work with bw images ([d112532](https://github.com/jolibrain/joliGAN/commit/d112532a64ddc1141c033e4fa2a664436fe1c779)) -* **ml:** control of cls semantic classifier learning rate ([be566c6](https://github.com/jolibrain/joliGAN/commit/be566c6728d72e572d2303f46de42b5338de5a08)) -* **ml:** DDIM with reference image ([f06e47c](https://github.com/jolibrain/joliGAN/commit/f06e47c187f775743670c282230f4d535b5c8f17)) -* **ml:** diffusion schedule with generation script ([4c26b09](https://github.com/jolibrain/joliGAN/commit/4c26b099481f5bb0c1a20108ab61996fa58c7ef3)) -* **ml:** semantic regression loss tensor dims ([1ddedc6](https://github.com/jolibrain/joliGAN/commit/1ddedc6c288801688c2eb26bd959b8f49b3b73bb)) -* mobile netG option ([4c3db79](https://github.com/jolibrain/joliGAN/commit/4c3db79302893ce24232753f161f76d73b75a43e)) -* modify command line for palette model ([2287054](https://github.com/jolibrain/joliGAN/commit/2287054b85c76d6024827c266d818486f1e067c3)) -* multigpu for cut model ([ad95e67](https://github.com/jolibrain/joliGAN/commit/ad95e67e76678a862309eb1a52c44a67c7af0fd6)) -* multigpu for cut model failed when batch_size < nb_gpus ([8d2fd1a](https://github.com/jolibrain/joliGAN/commit/8d2fd1a486fa0e3c16a15f6f7dbcd9739290ea37)) -* multigpu for cut_semantic_mask model ([4fe3ee2](https://github.com/jolibrain/joliGAN/commit/4fe3ee2584111f38c70d1bd88accc74116824361)) -* multimodal GAN requires retain_graph for now ([7be1628](https://github.com/jolibrain/joliGAN/commit/7be16288c32178c1815951853bd40b6789fb1a8c)) -* nb attn variable after refactor ([3e33fcc](https://github.com/jolibrain/joliGAN/commit/3e33fcc8e845a296eefc5a85e4c9ceda88a388b6)) -* nce layers for ittr generator ([916b475](https://github.com/jolibrain/joliGAN/commit/916b47581dba37764d0a0a98613a7c3d269125e3)) -* NCE with segformer architecture ([c037912](https://github.com/jolibrain/joliGAN/commit/c03791223019e06c08c014207cdaa17becad849d)) -* netF is initialized on the right gpu ([7856bd9](https://github.com/jolibrain/joliGAN/commit/7856bd9c0180d2d0b63a4f5f4e8c180c32f14af4)) -* netF weights updating ([a75544e](https://github.com/jolibrain/joliGAN/commit/a75544ec00fe67efb7970fb66e61252be8bc1429)) -* network loading ([dcb25d7](https://github.com/jolibrain/joliGAN/commit/dcb25d781b81f6676c1187ddf7c2214104661d07)) -* networks groups inherit from other models ([f988002](https://github.com/jolibrain/joliGAN/commit/f988002ef75637f7bca5c58d9c2c3d26aecd07f3)) -* new options names for unaligned_labeled_mask ([6fed55c](https://github.com/jolibrain/joliGAN/commit/6fed55cbe1d4e5d88700513349aac5c567c1b4f9)) -* no need for retain_graph to be True on losses ([b715638](https://github.com/jolibrain/joliGAN/commit/b7156388ebb03768f2c2a8da2b8726d9b2ee5ee4)) -* nvidia key rotation ([eadd702](https://github.com/jolibrain/joliGAN/commit/eadd70293016d9f77e07ed8c557955b9e1ca8cb6)) -* online creation and selfsupervised ([42dd435](https://github.com/jolibrain/joliGAN/commit/42dd435731339a71f17a3fe0eb4343110be70e4c)) -* online creation for temporal dataloader ([9cf3d0a](https://github.com/jolibrain/joliGAN/commit/9cf3d0a571607395e15462d8c73b1b351dd6d93c)) -* online creation when cropsize > imgsize ([c2fc34e](https://github.com/jolibrain/joliGAN/commit/c2fc34ed9a21881f0f47f5f98a57c0ae4b354536)) -* online mask loading with multiple bboxes ([1d7a6e1](https://github.com/jolibrain/joliGAN/commit/1d7a6e12e7f4e34c1fd525efe9f772be8ff4574a)) -* online mask padding ([90d7444](https://github.com/jolibrain/joliGAN/commit/90d74446b548d67931cb802854975da27b0db880)) -* only load testset on gpu 0 ([abcb4c4](https://github.com/jolibrain/joliGAN/commit/abcb4c4a92604e3a7178dfa5a40162c032b5f335)) -* only one gpu visible in each process ([4954717](https://github.com/jolibrain/joliGAN/commit/4954717dd5b322f1f1bff91494a5166d16a38616)) -* only onnx export fort segformer G ([543ce28](https://github.com/jolibrain/joliGAN/commit/543ce280dd364873a5737960cb3cf499c5f51173)) -* onnx export ([e9d36a8](https://github.com/jolibrain/joliGAN/commit/e9d36a83a820f3c800784fdaa2b939bd33cba21b)) -* option to not set device when parsing json config file ([092b5f8](https://github.com/jolibrain/joliGAN/commit/092b5f85e9fb570d51d74fc37898381e89e69a9f)) -* option type and max int value ([d7f60f2](https://github.com/jolibrain/joliGAN/commit/d7f60f2686aa67b0d504fb49668459cba4ff460e)) -* out mask images can be computed with more than 2 semantic classes ([36d54b9](https://github.com/jolibrain/joliGAN/commit/36d54b968b2398d6e8cf4f87609785c18a4e90af)) -* output_display_type default value should be a list ([d694d29](https://github.com/jolibrain/joliGAN/commit/d694d298fbc7d2649f153fdbf5c7ecb510a12a00)) -* palette inference without ref image ([02d3f76](https://github.com/jolibrain/joliGAN/commit/02d3f76abb6db51407a7f46d5d7b9e788bc792c4)) -* projection interpolation at init ([2aecf46](https://github.com/jolibrain/joliGAN/commit/2aecf46fe78e91da56efbd410d85aca1c14ff6ab)) -* proper categories titles in schema + improve option saving ([bbc41c6](https://github.com/jolibrain/joliGAN/commit/bbc41c6b8047c1f6b44fed58d2e9f870b7ac095e)) -* python html lib override ([be2b089](https://github.com/jolibrain/joliGAN/commit/be2b08933b7ccbca5f12579db5f6f9185ada5074)) -* random image aug class argument ([2d8f686](https://github.com/jolibrain/joliGAN/commit/2d8f6864fc1fef8bad58b115f0bd41e20e8eac1f)) -* random offset missing in crop_image ([1fc4bf4](https://github.com/jolibrain/joliGAN/commit/1fc4bf477d6605dd5fe2ffcabb9ae15c5a3e9597)) -* real image f_s pred name ([d40109c](https://github.com/jolibrain/joliGAN/commit/d40109c0cc7681b81f4fff43bae5c5f4d777e447)) -* remove environment.yml ([783cae8](https://github.com/jolibrain/joliGAN/commit/783cae8c296693494d1672a1c79b85f79f64242f)) -* remove eval mode for export ([9bdae8d](https://github.com/jolibrain/joliGAN/commit/9bdae8d935945d223eed86d950456834d8815ef6)) -* remove last batch norm to cls for batch size 1 ([d7fe727](https://github.com/jolibrain/joliGAN/commit/d7fe7278c9658c0380d171d4c3f3ce51bbdfdd4f)) -* remove use_resize from segformer forward ([c71d4de](https://github.com/jolibrain/joliGAN/commit/c71d4ded952224f894817a9e21e0d4ba90c3a7b9)) -* requirements for github actions ([6227c06](https://github.com/jolibrain/joliGAN/commit/6227c06e285e3c103c91e594120064578553fffb)) -* requirements for github actions ([4fe34ce](https://github.com/jolibrain/joliGAN/commit/4fe34cece2247a03c3a02facd5c486daa323ef99)) -* resnet attn class call ([17f907f](https://github.com/jolibrain/joliGAN/commit/17f907f23847d06301a135f003b2498122a2b741)) -* resnet default number of blocks with diffusion ([3d34368](https://github.com/jolibrain/joliGAN/commit/3d34368887df51c7b6d69bda0a5ab8f48a7c8341)) -* resnet with attention class name ([950237d](https://github.com/jolibrain/joliGAN/commit/950237de16a303adfa1852baae8880a6b79de39b)) -* resnet_attn model name ([eb4216e](https://github.com/jolibrain/joliGAN/commit/eb4216e8e59bdbcad125d35dc429e2026e73a1c3)) -* resnet_attn option ([08b61ea](https://github.com/jolibrain/joliGAN/commit/08b61eabfca0d9bd02a7adbc0fa9c8cc3d50aa36)) -* reverse DDIM schedule ([17ba96a](https://github.com/jolibrain/joliGAN/commit/17ba96a3c3bc0e5ebf0346e7918f8bb24ab4df5a)) -* round pixel gap before offset sampling ([05e8959](https://github.com/jolibrain/joliGAN/commit/05e8959aeef9ead7938486ec4d481b79663293b3)) -* rst autodoc ([5309cc2](https://github.com/jolibrain/joliGAN/commit/5309cc2eb04bbef62e86c858a5ab203add7269df)) -* rst format and typos ([f20f264](https://github.com/jolibrain/joliGAN/commit/f20f264eb622437e0f36c976e24c475927bbe188)) -* sam compatible with mask inversion in palette ([4294679](https://github.com/jolibrain/joliGAN/commit/42946798de144ae477b5f94cdd8ecbf6988d0f3c)) -* sam inference as discriminator ([5708832](https://github.com/jolibrain/joliGAN/commit/57088323771c1bb0a9540e567f9fb55c3c416c4c)) -* sample runs and default options ([f08d132](https://github.com/jolibrain/joliGAN/commit/f08d13248c5ac47931e77ae9995226915a714e58)) -* save generated bbox only when useful in diffusion inference ([4b93cf3](https://github.com/jolibrain/joliGAN/commit/4b93cf34703c8f8eec8ca31c47fd19c65f220f1f)) -* save model every epoch is default ([51c5d34](https://github.com/jolibrain/joliGAN/commit/51c5d34ad8b01ab11eca942fa1486f3c5974b78c)) -* save networks img for diffusion ([090df33](https://github.com/jolibrain/joliGAN/commit/090df33f8fad5f90237f9a617629fa1609b792b8)) -* saving paths_sanitized in checkpoints dir ([49c5139](https://github.com/jolibrain/joliGAN/commit/49c51394522ebf8472b23071813817c6b6f5e66f)) -* script for image gen using diffusion ([e909525](https://github.com/jolibrain/joliGAN/commit/e909525de69a3ad668d428e45dab98e44cff6818)) -* segformer as f_s net ([d55d63d](https://github.com/jolibrain/joliGAN/commit/d55d63d5e478aea94a06ee987d5d4a90caf26d13)) -* segformer for semantics with optional partially pre-trained model ([3078e51](https://github.com/jolibrain/joliGAN/commit/3078e51152d7a382508fdb05b51ab18391961bdb)) -* segformer G and segformer feature extractor ([85171e1](https://github.com/jolibrain/joliGAN/commit/85171e1c44ef8c6d5e7bfa7791603f8e36494b4b)) -* segformer ONNX export and image generation script ([c5a700b](https://github.com/jolibrain/joliGAN/commit/c5a700b493d70724e2aa4a60440a497a9b4a643b)) -* selfsupervised dataloaders don't need domain B anymore ([257a056](https://github.com/jolibrain/joliGAN/commit/257a056716bc6046f56f802a918278976fc98f6e)) -* sem compute for cls on identity ([6a87858](https://github.com/jolibrain/joliGAN/commit/6a87858ee327ded475db2d8ac13a082330976313)) -* semantic losses was initialized twice ([3fb85ab](https://github.com/jolibrain/joliGAN/commit/3fb85ab0f8dec7cedcc6f4663ec097482767a060)) -* server launching in generate api doc ([f459842](https://github.com/jolibrain/joliGAN/commit/f459842ad98e6b066f1cae2731ec5cadb18e1012)) -* **server:** wrong import ([cb4087e](https://github.com/jolibrain/joliGAN/commit/cb4087e773ace7781bf80705b96f14384bcb09e0)) -* smaller batchsize at end of epochs ([c49ddd1](https://github.com/jolibrain/joliGAN/commit/c49ddd1143f9cebefbc56e6a8a4a4427e6dc8c8b)) -* smaller miou interval to trigger computing ([76382d0](https://github.com/jolibrain/joliGAN/commit/76382d07f3eaed6bdc63b5b8889e5a13ff61b2a6)) -* softmax G semantic was applied twice ([5c4c3ec](https://github.com/jolibrain/joliGAN/commit/5c4c3ec2626f1c9582ad1715031d89ac2642dd35)) -* stylegan2 feature extraction ([e442021](https://github.com/jolibrain/joliGAN/commit/e44202180d93ce5b087f19b3389eed74e27edf08)) -* support for no mask in B domain ([6dd98d6](https://github.com/jolibrain/joliGAN/commit/6dd98d6bd3800af5c1c9d5a36677e2b017fd7b41)) -* sv network as latest even with train_save_by_iter ([9016d3f](https://github.com/jolibrain/joliGAN/commit/9016d3fd5ee49ae19c71036c6c6c504f15d34f56)) -* **temp:** norm default ([d95d10f](https://github.com/jolibrain/joliGAN/commit/d95d10f03c7d726d355a0d0245237d59b4691fe1)) -* temporal criterion loss compute ([4ddce8b](https://github.com/jolibrain/joliGAN/commit/4ddce8b3631c4c42c29cb4506e4becbee9912555)) -* temporal D and context pixels compatibility ([421f458](https://github.com/jolibrain/joliGAN/commit/421f458dd3d51251ed6bee8f6ac0f27dbfa25044)) -* temporal dataloader, with added path sanitize ([6451db2](https://github.com/jolibrain/joliGAN/commit/6451db2acca4979e394d717af55ab825dd65a649)) -* temporal dataloading and loss computing ([07706d3](https://github.com/jolibrain/joliGAN/commit/07706d3dfb303030e69403d0a262970822da448c)) -* temporal end of sequence ([e08ffba](https://github.com/jolibrain/joliGAN/commit/e08ffbadece35144a77f7b96341324347d909e30)) -* temporal mse with GAN models ([79b0d14](https://github.com/jolibrain/joliGAN/commit/79b0d141a0bfddab3d4c51be2d56211638f22537)) -* test functions names ([2247f18](https://github.com/jolibrain/joliGAN/commit/2247f18cb2981d23e62000028a41444d11ee2ee5)) -* tests ([fabd8b6](https://github.com/jolibrain/joliGAN/commit/fabd8b65f9febb3c6ebf09e55d6c9e75012b64e1)) -* tests for all f_s net ([144d7fd](https://github.com/jolibrain/joliGAN/commit/144d7fd91ab5822a7adc8c06818118920c8cc0e7)) -* tests with no cache ([787f27b](https://github.com/jolibrain/joliGAN/commit/787f27b88b5b170396709fa2ac20e01bee0b9338)) -* training examples with mask_delta ([9b14339](https://github.com/jolibrain/joliGAN/commit/9b14339a5777dfb08d82c42973685c22bd67dd22)) -* training examples with mask_delta ([8d08839](https://github.com/jolibrain/joliGAN/commit/8d0883995a153a4f7d6357fb107925150917fec8)) -* typo ([4274784](https://github.com/jolibrain/joliGAN/commit/4274784d1e5e2ab7e5c37a2b436f95c80be225ba)) -* typo ([f1dcff7](https://github.com/jolibrain/joliGAN/commit/f1dcff7498d406c7c91acef4cbadaa52ae530896)) -* typo in README.md ([f5ef15e](https://github.com/jolibrain/joliGAN/commit/f5ef15e09b2ae6af4571cb6d8e561ef01b69a0e7)) -* typos ([5b061f2](https://github.com/jolibrain/joliGAN/commit/5b061f2d952dc24e4578cca40577277ad85b5797)) -* typos and class name ([6462374](https://github.com/jolibrain/joliGAN/commit/646237406979672bc99fbe7363380f4080614e78)) -* unaligned data_dataset_mode in README ([5e7fbd9](https://github.com/jolibrain/joliGAN/commit/5e7fbd930132a87364c9a63550358ca02ea2e7be)) -* underflow in custom cut l2norm, replaced by torch built-in ([ff42852](https://github.com/jolibrain/joliGAN/commit/ff4285267230d5b60c4e16e4b34b68f1aef05d92)) -* UNet/UViT layers for cut NCE ([8459876](https://github.com/jolibrain/joliGAN/commit/84598761018df13a0a294c9d9ba237ed8f64665a)) -* unset index_B for unaligned mask offline datasets ([8ba5e12](https://github.com/jolibrain/joliGAN/commit/8ba5e1271bd4db89bc76ef6a77de24094220e646)) -* Update README.md ([cb9031d](https://github.com/jolibrain/joliGAN/commit/cb9031daae2ac0d4b2553e6def29642ac9278621)) -* Update README.md ([7e5fe82](https://github.com/jolibrain/joliGAN/commit/7e5fe82089477aa1d4d95bbc35a1d728e4dcabc3)) -* use only mask from selected bbox at inference time ([dd4fe3b](https://github.com/jolibrain/joliGAN/commit/dd4fe3bb0ce0494d6af6ad78179ed7ae678017e3)) -* use signal handler to kill all the processes ([fab93ef](https://github.com/jolibrain/joliGAN/commit/fab93ef95aea6d17b105a29c44135bd9183c670f)) -* using cut with F as pure sampling function ([887b9a5](https://github.com/jolibrain/joliGAN/commit/887b9a5178a76b16e4cfd1a9b4b074854245432c)) -* validation with no label, e.g. simple cut model ([53984f4](https://github.com/jolibrain/joliGAN/commit/53984f4d6cdf165a8b4bec0205d1b0142dee7410)) -* visdom autostart and no display ([fbc9942](https://github.com/jolibrain/joliGAN/commit/fbc994267b5d05b706ebec46228606549a11250e)) -* visdom port for server launching ([747d18a](https://github.com/jolibrain/joliGAN/commit/747d18ae1f2823521acdc61a92c05fbfc23607b5)) -* visuals for cyclegan ([b7cabaa](https://github.com/jolibrain/joliGAN/commit/b7cabaa773f4bb2de4d19feade47524dbf57c8c5)) -* vitclip16 requires 224 input size ([8b8bb67](https://github.com/jolibrain/joliGAN/commit/8b8bb677b4d806f6bfcb9928540591c71380ef55)) -* wrong D_noise option ([eaea415](https://github.com/jolibrain/joliGAN/commit/eaea415295588c04d4a8fdf824b97f931851a387)) -* wrong feature network in projected ([3318c77](https://github.com/jolibrain/joliGAN/commit/3318c7722bb224ac24b471b77586a62834833da4)) -* wrong indentation ([bac6ce2](https://github.com/jolibrain/joliGAN/commit/bac6ce2ff82efecde36e922161f82feb792c8c2c)) -* wrong option for data sanitize ([479242e](https://github.com/jolibrain/joliGAN/commit/479242e62e4391a8e229075e9059ac57355f9bbd)) -* wrong option name in test and add comment in json option loading ([1a7abd6](https://github.com/jolibrain/joliGAN/commit/1a7abd6bb17d85298933331de7c7b687eb238bc8)) -* wrong options ([4298071](https://github.com/jolibrain/joliGAN/commit/42980714a7b768b670b145db10dae753b258b8a5)) -* wrong options for validation loading and fid computing ([a9d2516](https://github.com/jolibrain/joliGAN/commit/a9d251615b751f563107743b7d0df79df751c517)) -* wrong train_compute_fid option ([893c0e2](https://github.com/jolibrain/joliGAN/commit/893c0e29bae60ee6c7ece5257661192fe52dfcd9)) -* wrong values for bbox ref coordinates ([dc2a00f](https://github.com/jolibrain/joliGAN/commit/dc2a00f0a0215d631ff65faf4379cbcd827d1a7f))