Releases · neuraloperator/neuraloperator

Summary

We are excited to release this new version of the neuraloperator library! It brings many improvements, including new architectures (SFNO, GNO, GINO), many improvements to existing ones, out-of-the-box super resolution, super-evaluation and incremental training.

All models can now be easily saved and loaded and we provide a lightweight trainer compatible with all our neuraloperators. Head to the examples for some sample code, and to the API for a full documentation!

What's Changed

Refactor MLP config by @JeanKossaifi in #143
Adds super-resolution to FNO by @JeanKossaifi in #147
Adds ADA_IN norm by @JeanKossaifi in #148
Adds SFNO by @JeanKossaifi in #150
Fix missing parameters output_scaling_factor by @sleepyeye in #159
Finodev by @ashiq24 in #152
quick_avoid by @ashiq24 in #173
Add low-precision to TFNO by @crwhite14 in #172
fix SFNO example by @crwhite14 in #177
fix comma in readme file by @gegewen in #179
Liftproj mod to mlp by @btolooshams in #182
MLP additional statement to check for scenario where n_layers=1 by @btolooshams in #183
marge the update on fno mlp help description by @btolooshams in #191
Update guide on Fourier neural operator by @devzhk in #156
Spectrum analysis of datasets. by @Robertboy18 in #193
adding flag option to only pad the last dim by @btolooshams in #185
Minor error in L#602 in fno.py by @ImanLiao in #194
gino by @kovachki in #195
Fix mlp nonlinearity by @ziqi-ma in #197
Reformat layers/ directory with black by @m4e7 in #199
Use lifting channels in FNO.lifting if it is passed by @dhpitt in #196
removed dead lines in FNOGNO by @dhpitt in #203
Reformat datasets/ directory with black by @m4e7 in #205
docstring for FNOGNO by @dhpitt in #202
Fix Python 3.6 f-string compatibility and condense documentation for FNO classes by @m4e7 in #209
Split preactivation from FNOBlock.forward() by @m4e7 in #214
Padding correction by @ashiq24 in #218
general trainer class for GINO and NO by @dhpitt in #215
Padding correction by @ashiq24 in #220
fix example Trainer API calls by @dhpitt in #219
Refactor rescaling in skips as transform in the Spectral Conv by @JeanKossaifi in #217
Further simplification + UNO fix by @JeanKossaifi in #221
Sht correction by @ashiq24 in #222
Fix loss signatures to build doc by @dhpitt in #224
fix small bug in the WandB logger callback by @dhpitt in #232
Adding 4D prediction only, no nested fno by @gegewen in #225
Revert 4D_FNO changes until they are properly tested by @dhpitt in #235
Remove torch_scatter and torch_cluster from CI pipeline's dependencies by @dhpitt in #233
Model checkpointing by @dhpitt in #234
Updates to documentation and callback docstrings by @dhpitt in #237
index dropout moduleList by @dhpitt in #239
fix syntax error and add index.rst by @dhpitt in #240
Fix typo in checkpoint init by @rybchuk in #241
Bug fixes and unit testing for Callbacks by @dhpitt in #242
Refactors SpectralConv for simpler FNO by @JeanKossaifi in #244
BaseModel: adds checkpointing, versioning, safeguards by @JeanKossaifi in #257
Add Burger's dataset and PINO by @crwhite14 in #256
Enable transform wrappers by @JeanKossaifi in #254
Update tensor_dataset.py by @slanthaler in #260
Update to the checkpoint callback and test by @dhpitt in #258
fix domain_padding to accept list (e.g., [0,0,1]) in addition to sc… by @btolooshams in #263
Move to DataProcessor API by @dhpitt in #262
Fix navier stokes preprocessor bug by @dhpitt in #265
Fixes to make DataProcessor code doc build by @dhpitt in #266
Add AutoML via Optuna by @crwhite14 in #243
fixing the horizontal_skips_map construction, it was not going through by @btolooshams in #267
Updates to saving and loading models by @dhpitt in #268