Version 0.2.0 - Transformers and Anomaly Detection

What's Changed

• Adds likelihood computation by @marksgraham in #122
• Add classifier-free guidance tutorial by @Warvito in #131
• Add missing scale by @Warvito in #147
• Fix 2d-ldm-tutorial. by @JessyD in #151
• Add 3d percetual loss by @Warvito in #158
• Adds inpainting tutorials by @marksgraham in #161
• Fix set_timesteps by @Warvito in #163
• Add stable diffusion v2.0 x4 upscaler tutorial by @Warvito in #148
• Added the MMD Metric and tests by @danieltudosiu in #152
• Add v_prediction and update docstrings by @Warvito in #165
• Add RadImageNet to Perceptual Loss by @Warvito in #153
• Add option num_head_channels as Sequence by @Warvito in #172
• Fix kernel_size in quant_conv and post_quant_conv layers by @Warvito in #170
• Fix "medicalnet_..." network_type used with spatial_dims==2 by @Warvito in #167
• Adds context error by @marksgraham in #175
• Remove ununsed predict_epsilon by @Warvito in #184
• Fix TestDiffusionSamplingInferer by @Warvito in #180
• Add option to use residual blocks for updownsampling by @Warvito in #176
• Optimise Attention Mechanisms by @Warvito in #145
• Fix F.interpolate usage with bfloat16 by @Warvito in #157
• Pretrained DDPM by @marksgraham in #177
• Add full precision attention by @Warvito in #189
• Add FID by @Warvito in #40
• Update tests, CI and pre-commit by @Warvito in #193
• Fix missing init.py by @Warvito in #200
• Replace FeedForward with MLPBlock by @Warvito in #201
• Remove python3.8 as default_language_version by @Warvito in #209
• Refactor code with new pre commit configuration by @Warvito in #207
• Fixes by @marksgraham in #211
• Modify sample function to divide by scale factor before passing to th… by @virginiafdez in #214
• Addition of is_fake_3d setting condition to error in PerceptualLoss by @virginiafdez in #215
• Add verification code by @Warvito in #221
• Suspend CI by @Warvito in #224
• Sequence Ordering class by @danieltudosiu in #168
• Add AutoregressiveTransformer by @Warvito in #225
• Remove ch_mult from AutoencoderKL by @Warvito in #220
• Fix print messages for MS-SSIM by @JessyD in #230
• 228 update pretrained ddpm by @marksgraham in #233
• WIP by @OeslleLucena in #202
• Add annotations from future by @Warvito in #239
• Change num_res_blocks to Sequence[int] | int by @Warvito in #238
• Change num_res_channels and num_channels to Sequence[int] | int by @Warvito in #237
• Initialise inference_timesteps to train_timesteps by @marksgraham in #240
• Fix eval mode for stage_2 by @Warvito in #246
• Fix format by @Warvito in #250
• Fix format by @Warvito in #251
• Use no_grad decorator for sample method by @Warvito in #244
• Add VQVAE + Transformer inferer by @Warvito in #242
• Fix TypeError by @Warvito in #254
• Harmonise VQVAE with AutoencoderKL by @Warvito in #248
• Adds transformer get_likelihood by @marksgraham in #257
• Changed PatchAdversarialLoss to allow for least-squares criterion to … by @virginiafdez in #262
• 258 fix diffusion inferer by @marksgraham in #265
• Fixes type by @marksgraham in #264
• Change default values by @Warvito in #267
• 106 vqvae transfomer tutorial by @Ashayp31 in #236
• 203 add scale factor to the ldm training tutorials by @virginiafdez in #271
• Mednist Bundle by @ericspod in #263
• Add cache_dir by @Warvito in #278
• Add option to use flash attention by @Warvito in #222
• Set param.requires_grad = False by @Warvito in #273
• Update loss imports by @marksgraham in #279
• Add Brain LDM to model-zoo by @Warvito in #188
• Fix 3D DDPM tutorial by @Warvito in #277
• Add use_flash_attention argument by @Warvito in #284
• Fixes transformer max_sequence_length training/inference by @marksgraham in #282
• Add MIMIC pretrained model by @Warvito in #286
• Fix validation data in tutorials by @Warvito in #291
• Fix prediction_type="sample" by @Warvito in #300
• Moving DiffusionPrepareBatch by @ericspod in #305
• Change transformer number of tokens by @Warvito in #303
• Add tutorial performing anomaly detection based on likelihoods from generative models by @Warvito in #298
• Add causal self-attention block by @Warvito in #218
• Remove TODOs by @Warvito in #310
• Fix num_head_channels by @Warvito in #316
• Update AutoencoderKL and add more content by @Warvito in #315
• fixed typo in anomaly detection tutorial by @vacmar01 in #321
• Fix generate function by @Warvito in #322
• Clip image_data values before casting to uint8 by @Warvito in #324
• 314 fix transformer training by @marksgraham in #318
• Fix xtransformer error by @Warvito in #327
• Update README.md by @Warvito in #328
• 150 - Diff-scm by @SANCHES-Pedro in #306
• Update tutorial by @Warvito in #329
• Add README file to tutorial dir by @Warvito in #330
• Add more content to tutorial by @Warvito in #331
• Fix dependencies and license by @Warvito in #332
• Update README.md by @marksgraham in #333
• Update anomaly detection notebook by @marksgraham in #334
• initial commit anomaly detection with gradient guidance by @JuliaWolleb in #190
• Fix formatting by @Warvito in #340
• fix transformer max_sequence_length by @marksgraham in #335
• Fix typo by @Warvito in #341
• Remove dependency on x transformers by @Warvito in #325
• Add flash attention to Transformers by @Warvito in #342
• Makes use of inferer classes for training and sampling by @marksgraham in #347
• Fix MS-SSIM by @Warvito in #348
• initial commit for segmentation with diffusion models by @JuliaWolleb in #292
• Fix format by @Warvito in #349
• Use double in FID computation by @Warvito in #350
• Adopt exact computation of FID by @Warvito in #355
• Fixing vulnerable version of setuptools by @ericspod in #360
• Diffusion Autoencoder tutorial by @SANCHES-Pedro in #361
• Fix typo by @Warvito in #367
• Fix number of attention heads by @Warvito in #369
• Update CXR model by @Warvito in #336
• Add sample figures by @Warvito in #370
• Fix CUDA device mismatch by @ycremar in #371
• Fix grad strides warning when using ddp by @marksgraham in #375
• Fix MMD Metric by @Warvito in #368
• Add ControlNet by @Warvito in #358
• Evaluate the performance of generative models (realism and diversity) by @JessyD in #227
• Remove asserts by @Warvito in #383
• Bump to version 0.2 by @Warvito in #384

New Contributors

• @OeslleLucena made their first contribution in #202
• @Ashayp31 made their first contribution in #236
• @ericspod made their first contribution in #263
• @vacmar01 made their first contribution in #321
• @JuliaWolleb made their first contribution in #190
• @ycremar made their first contribution in #371

Full Changelog: 0.1.0...0.2.0