All notable changes to this project will be documented in this file.
The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.
- Fixed HTTP Error 308 when running
examples/super_resolution.py - Fixed deprecated documentation links to the Johns Hopkins Turbulence Database
- Updated package to use
physicsnemov2.0 APIs
- Added Blackwell support
- Improved performance of connectivity calculation. Refactor of Least squares gradients
- Added Incompressible Navier Stokes PDE as a special formulation
- Added ability for
make_nodesfunction to return a dict
-
Fix the zero equation formulation to correctly reflect turbulent kinematic viscosity
-
Made
tiny-cuda-nnoptional in deployment image
- Update
inspect.getargspecusage
- Fix optimizer checkpoint loading errors
- Remove upper bound for Cython
- Add upper bound for numpoly (transitive dependency of chaospy)
- Added devcontainer support.
- Speeds up CUDA extension build by about 3x.
- Fix the area calculated for STL meshes.
- Relaxes versions for several dependencies.
- AMP for derivatives.
- DALI based dataloader for Geometry module.
- Generalized PDE residual computing utility.
- Support for spatial gradients calculations using finite difference, meshless finite difference, spectral and least squares methods.
- Add docs for SDF and relevant features for geometry module.
- Upgrade Sympy and Scikit-Learn versions.
- Upgrade notebook, opencv-python and setuptools versions to fix CVEs GHSA-9q39-rmj3-p4r2, GHSA-qr4w-53vh-m672, and GHSA-cx63-2mw6-8hw5 respectively.
- Added inference notebook for Turbulence Super-resolution example.
- Warp based backed for STL geometry handling
- Update
timmdependency - Update minimum python version to 3.10
- Added reservoir examples using GenAI and CCUS workflows.
- Update OpenCV and Pillow versions to fix security
- Fix bug for
ConvFullyConnectedArch. - Updating
Activation.SILUtest to conform with updated nvFuser kernel generation.
- Added instructions for docker build on ARM architecture.
- Added domain decomposition examples using X-PINN and FB-PINN style.
- Integrated the network architecture layers into PhysicsNeMo-Core.
- Fixed Gradient Aggregation bug.
- Upgrade Pillow and Sympy to their latest versions.
- Upgrade Scikit-Learn version.
- Updated base PyTorch container to 23.10 and Optix version to 7.3.0
- Example for reservoir modeling using PINO and FNO
- Added a CHANGELOG.md
- Accompanying licenses (will provide in the PhysicsNeMo docker image).
- Arch
from_configbug for literal params. - Fixed fused SiLU activation test.
- Update
np.booltonp.bool_. - Added a workaround fix for the CUDA graphs error in multi-node runs
- Updated the base container to latest PyTorch base container which is based on torch 2.0
- Container now supports CUDA 12, Python 3.10
- Updated symengine to >=0.10.0 and vtk to >=9.2.6
- Initial public release.