PaddleScience 1.4.0 Release Notes
PaddleScience 1.4.0 is now available, bringing a series of new features, examples, documentation updates, and bug fixes. This release enhances the overall performance and usability of PaddleScience. Key updates include:
Examples
- Added new examples including Extformer-MoE, adv_cvit, ns_cvit, spinn on Helmholtz equation, brusselator3d, velocitygan, CGCNN, High-resolution Reconstruction of Cardiac MRI, TGCN Model for Traffic Forecasting, and more.
- Introduced examples for AI-aided geometric design of anti-infection catheters, DrivAerNet, Pangu-Weather, and FengWu.
Features
- Added support for amsgrad in Adam/AdamW optimizers.
- Integrated PIR export and inference capabilities.
- Supported ONNX export in PIR mode.
- Introduced leading zero padding in filenames when saving vtu files for better file sorting.
- Enhanced device management by ensuring default device settings and multi-device support.
Improvements
- Refined documentation and code structures for better clarity and maintenance.
- Addressed various fixes related to batch indexing, data types, and device settings.
- Updated URLs and added badges to improve documentation credibility.
- Enhanced loss aggregator functionality within the saving/loading process.
- Improved the evaluation and configuration management across multiple examples.
- Updated requirements and dependencies for consistent environment setup.
- Introduced new modules and refined existing ones for better performance.
- Enhanced compatibility with different modes and devices.
Documentation and Bug Fixes
- Fixed multiple typos and inconsistencies in documentation.
- Resolved issues with solver prediction and device settings.
- Fixed dtype errors and ensured compatibility with various data types.
- Updated documentation to reflect recent changes and improvements.
- Introduced comprehensive guides for new and existing features.
We recommend upgrading PaddleScience to version 1.4.0 for the latest improvements.
New Contributors
- @KennyNH, @KaiCHEN-HT, @niushao12, @banjiuyufen, @zhudequan9, @GoldenStain, @MGWanderer1, @casia-rxwang, @PolaKuma, @xusJs, @ADream-ki, @kylekk01, @LilaKen, @huhaohao525, @BeingGod, @yuchunliu-metax, @AoxiangGeng, and @leeleolay made their first contributions in this release!
Full Changelog: v1.3.0...v1.4.0