v0.0.5

What's Changed

• Update LVAE DataType list by @veegalinova in #251
• 3D model by @CatEek in #240
• ci(pre-commit.ci): autoupdate by @pre-commit-ci in #250
• Feature: Predict to disk (outerloop implementation) by @melisande-c in #253
• refac: modularized + cleaned the code for LVAE losses by @federico-carrara in #255
• refac: reorganization of pydantic configs in LVAE model by @federico-carrara in #256
• Fix(BMZ export): torchvision version; model validation by @melisande-c in #257
• Fix: Enforce dataloader params to have shuffle=True by @melisande-c in #259
• Performance test induced fixes by @CatEek in #260
• ci(pre-commit.ci): autoupdate by @pre-commit-ci in #262
• fix: Allow singleton channel in convenience functions by @jdeschamps in #265
• feat: Add convenience function to read loss from CSVLogger by @jdeschamps in #267
• Refac: Rename config file to careamics.yaml in BMZ export by @melisande-c in #271
• Feature: Load model from BMZ using URL by @melisande-c in #273
• feat: Enable prediction step during training by @jdeschamps in #266
• refac: added possibility to pick kl_restricted loss & other loss-related refactoring by @federico-carrara in #272
• A new enum for a new splitting task. by @ashesh-0 in #270
• Fix(BMZ): Relax model output validation kwargs; extract weights and config file following new spec and core release by @melisande-c in #279
• Fix(dependencies): Set bioimageio-core version greater than 0.7.0 by @melisande-c in #280
• fix: fixed a bug in KL loss aggregation (LVAE) by @federico-carrara in #277
• fix: ONNX exportability compatibity test and fix by @nimiiit in #275
• refac: updated calibration code + moved it into a separate script by @federico-carrara in #283
• MicroSplit: adding data type for a new dataset. by @ashesh-0 in #284
• Feature: Add predict command to CLI by @melisande-c in #281
• Noise models for paper notebooks by @CatEek in #285
• Fix: fix device array handling in NoiseModels by @CatEek in #288
• chore: Add templates by @jdeschamps in #291
• Add dependabot and cap dependencies by @jdeschamps in #287
• refac: optimize WelfordStatistics update and finalization by @federico-carrara in #325
• Update BMZ README by @jdeschamps in #322

New Contributors

• @nimiiit made their first contribution in #275

Full Changelog: v0.0.4...v0.0.5

v0.0.4.2

Full Changelog: v0.0.4.1...v0.0.4.2

v0.0.4.1

Full Changelog: v0.0.4...v0.0.4.1

v0.0.4

What's Changed

• feat: added ADAMAX optimizer among supported ones by @federico-carrara in #227
• Remove old examples by @CatEek in #228
• refac: removed noise model parameters from LadderVAE module by @federico-carrara in #233
• fix: making range parameter mandatory in psnr() by @federico-carrara in #236
• ci(pre-commit.ci): autoupdate by @pre-commit-ci in #237
• fix: set upper Pydantic version to 2.8 by @jdeschamps in #241
• refac: refactored serialization of arrays for noise model and likelihood configs by @federico-carrara in #232
• fix: fixing various bugs to enable microSplit training by @federico-carrara in #242
• Feature: Function that computes CAREamics TileInformation with GridIndexManager by @melisande-c in #217
• Simplify AlgorithmConfig discrimination by @jdeschamps in #222
• feat: added more metrics for evaluation of Splitting results by @federico-carrara in #230
• Chore: Split test CIs into CAREamics official and LVAE by @jdeschamps in #245
• Feature: Transform multiple arrays identically to patch and target by @melisande-c in #238
• Added several params to training model by @CatEek in #225
• bug fix (LC + noise set) by @ashesh-0 in #246
• Feature: Update BMZ export by @melisande-c in #248
• Add custom transforms to convenience functions, and other API improvement by @jdeschamps in #239
• LVAE dataset: add MultiFileDset by @veegalinova in #223
• Feature: CLI work in progress by @melisande-c in #130
• Add stop training method by @jdeschamps in #249

New Contributors

• @ashesh-0 made their first contribution in #246

Full Changelog: v0.0.3...v0.0.4

v0.0.3

What's Changed

• Algorithm split by @CatEek in #187
• Refactor: Noise Models by @federico-carrara in #205
• Feature: LVAE prediction functions by @melisande-c in #206
• General refactoring and testing of LVAE loss functions. by @federico-carrara in #208
• Renaming pydantic models for noise models and likelihood to SomethingConfig by @federico-carrara in #209
• Feature: Add tiling method from Disentangle repository by @melisande-c in #207
• Fixing circular import error arising from importing UNet in models/__init__.py by @federico-carrara in #213
• Masked percentage value upper limit by @CatEek in #215
• docs: fix typos, and fix building of docs by @tlambert03 in #214
• LVAE dataset: replace ml_config with pydantic by @veegalinova in #212
• Feature: LVAE prediction compatible with CAREamics tiling pipeline by @melisande-c in #211
• Fix: serialize data statistics as strings by @jdeschamps in #197
• Update authors list by @jdeschamps in #220
• Add custom algorithm to FCNAlgorithmConfig by @jdeschamps in #221
• Noise model training by @CatEek in #210
• Implemented training_step and validation_step, fixed all the configs needed to instantiate the lightning module by @federico-carrara in #218

New Contributors

• @tlambert03 made their first contribution in #214

Full Changelog: v0.0.2...v0.0.3

v0.0.2

What's Changed

• Chore: Exclude lvae training from coverage by @jdeschamps in #178
• Feature: Prediction writer callback initialisation by @melisande-c in #179
• Feature: Prediction writer callback by @melisande-c in #181
• Refactor: Stitch predictions – enforce singleton C axis in TileInformation by @melisande-c in #183
• Fix: Set bioimage.core<=0.6.5 to stop tests from failing by @melisande-c in #193
• Fix: Use the experiment-name in config as WandB logger name by @melisande-c in #191
• Fix: Use Codecov CLI version v0.7.3 in github action by @melisande-c in #200
• Revert: Set bioimageio.core>=0.6.0 as before by @melisande-c in #199
• ci(pre-commit.ci): autoupdate by @pre-commit-ci in #198
• Feat: Allow passing folder in BMZ by @jdeschamps in #184
• Fix: Remove nested function to allow num_workers > 0 to work by @conradkun in #194
• Chore: Update README by @jdeschamps in #203

New Contributors

• @conradkun made their first contribution in #194

Full Changelog: v0.1.0rc8...v0.0.2

v0.1.0rc8

What's Changed

• ci(pre-commit.ci): autoupdate by @pre-commit-ci in #173
• Chore: Remove outdated notebooks by @jdeschamps in #163
• Feature: Add an autocorrelation utility function by @jdeschamps in #172
• Doc: Add descriptions to the Pydantic models by @jdeschamps in #171
• Feature: Use current weights for prediction by @jdeschamps in #166
• Fix: Inconsistent prediction outputs by @melisande-c in #170
• Refactoring: lightning API package and smoke tests by @jdeschamps in #161
• Feature: Add File IO package by @melisande-c in #174

Full Changelog: v0.1.0rc7...v0.1.0rc8

v0.1.0rc7

What's Changed

• Refactor prediction pipeline by @jdeschamps in #131
• Make normalization mandatory by @jdeschamps in #139
• Drop support for Python3.8 by @jdeschamps in #137
• Add seeds and random number generator by @jdeschamps in #138
• Improve CAREamist docs by @jdeschamps in #142
• Simplify CAREamist BMZ export by @jdeschamps in #144
• Channel-dependent normalization by @CatEek in #134
• (Fix) Upper versions for numpy, zarr and skimage deps by @jdeschamps in #149
• Fix: Load checkpoint on the correct device by @jdeschamps in #146
• Refactor: Rename image means and stds with plural by @jdeschamps in #147
• Fix: add missing test fail template by @jdeschamps in #155
• Feature: API parameter for custom callbacks by @jdeschamps in #150
• Refactor: Tiling applied post prediction (in contrast to during loop) by @melisande-c in #141
• Feature: Allow Noise2Noise with different in and out channels by @jdeschamps in #152
• Feature: Include LVAE model and training code by @federico-carrara in #154
• Fix: stitch multiple tiled prediction outputs by @melisande-c in #165
• Fix: Fix TTA and add tests by @jdeschamps in #164
• Feature: Welford std approximation by @CatEek in #153

New Contributors

• @federico-carrara made their first contribution in #154

Full Changelog: v0.1.0rc6...v0.1.0rc7

v0.1.0rc6

What's Changed

• Fix some pre-commit errors by @jdeschamps in #123
• Remove prediction transform by @jdeschamps in #124
• Fix pre-commit errors by @jdeschamps in #126
• ci(pre-commit.ci): autoupdate by @pre-commit-ci in #129
• Fix transforms by @jdeschamps in #128
• (chore): add py12 to CI by @jdeschamps in #133
• Inputs to BMZ should not be normalized by @jdeschamps in #132

Full Changelog: v0.1.0rc5...v0.1.0rc6

v0.1.0rc5

What's Changed

• ci(pre-commit.ci): autoupdate by @pre-commit-ci in #116
• Fix patching by @jdeschamps in #113
• N2V2 fix by @jdeschamps in #114
• Fix bmz by @jdeschamps in #117
• Feature: Add independent channel option in UNet – Conv group implementation by @melisande-c in #118
• (chore): New PR template by @jdeschamps in #120
• (refactor): Remove Albumentations by @jdeschamps in #119
• Update README.md by @jdeschamps in #122
• Enforce specific tile size in UNet to reduce artifacts by @jdeschamps in #121

New Contributors

• @melisande-c made their first contribution in #118

Full Changelog: v0.1.0rc4...v0.1.0rc5