🔖 Release Note: v1.3.0

Feature

• Workflow now accepts a list of opt_direction.

Fixes

• Several fixes about indexing and other bugs by @sses7757 in #242
• Fix win installer by @ZaberKo in #243
• reconstruct cec2022, fix and add funcs of basic.py by @XU-Boqing in #240
• use_state when there are multiple additional arguments

Full Changelog: v1.2.2...v1.3.0

🔖 Release Note: v1.2.2

This is a minor release focused solely on bug fixes:

• Removed unused imports to improve code cleanliness.
• Fixed unintended behavior in certain Differential Evolution (DE) algorithms where step was being called within init_step.
• Various fixes in the documentation.

Full Changelog: [v1.2.1...v1.2.2](v1.2.1...v1.2.2)

🔖 Release Note: v1.2.1

This is a minor release focused on improving stability and addressing bugs, with a few quality-of-life enhancements.

✨ New Features

• New Benchmark Functions: Added single-objective numerical functions: Ellipsoid and Griewank.

🐛 Bug Fixes

• Fixed an issue where StdWorkflow failed to work with algorithms that inherit from other algorithms.
• Corrected a bug in latin_hypercube_sampling_standard function.
• Resolved an issue with non_dominate failing under torch.compile.
• Corrected a problem where PSO did not use the default device properly in certain cases.

🧹 Refactoring & Maintenance

• Re-exported commonly used utilities to the top level for convenience, for example:
  • evox.compile instead of evox.core.compile
  • evox.vmap instead of evox.core.vmap.
• Removed deprecated or redundant code.

Full Changelog: v1.2.0...v1.2.1

🎉 EvoX v1.2.0 Release! 🎉

We’re excited to announce the release of EvoX v1.2.0, packed with new features, improvements, and important bug fixes! This update enhances the framework's flexibility and performance across key modules.

🔧 Highlights:

• 🚀 Added final_step() and updated hpo_wrapper and std_workflow for smoother execution workflows.
• 🤖 Introduced the Mujoco Playground – a new environment for reinforcement learning experiments.
• 📘 Released a brand-new tutorial to help users get up to speed quickly.
• 🔌 Added new EvoX extension functions, enabling better integration with sister projects like EvoMO.
• 🐞 Various bug fixes and documentation updates for better stability and clarity.

Related pull requests

• Update Docstring of MOEAs by @Zhenyu2Liang in #230
• Fix the vmap used in the hpo_wrapper.py by @starquakee in #232
• Add RVEAa and final_step, and modify DTLZ, IGD and RVEA by @XU-Boqing in #231
• Update hpo_wrapper and std_workflow (final_step()) by @XU-Boqing in #233
• Add mujoco_playground problem. by @Nam-dada in #234
• Update docstring by @Zhenyu2Liang in #235
• Fix bugs of mujoco_playground by @Nam-dada in #236
• Tutorial by @BillHuang2001 in #237

🔗 Full Changelog: v1.1.2 → v1.2.0

What's Changed

• Update README with usage notes and parameter adjustments by @JiangTao97 in #225
• Some vmap and HPO-related features and bug fixes by @sses7757 in #226
• Some documentation fixes @JiangTao97 in #227 and #228
• Some fixes about benchmark code by @sses7757 in #229
• Initial distributed workflow support has landed!

Full Changelog: v1.1.1...v1.1.2

What's Changed

This minor release primarily includes bug fixes and improvements:

• Fixed an issue where torch.compile did not properly capture workflow.step.
• Fixed an issue where use_state leads to a graph break.
• Fixed some incorrect model buffer usage.
• Fixed an issue where monitor.plot was not functioning as intended in certain situation.
• Introduced a new wrapper, evox.compile, to work around certain limitations with torch.compile and torch.vmap.
• Resolved various BatchedTensor-related issues:
  • Fixed a bug where vmapping a workflow with EvalMonitor could cause BatchedTensor leaks.
  • Fixed an issue preventing HPOProblem from working with BraxProblem.
• Enhanced the implementations of RVEA and CSO for better performance and reliability.
• Enhanced the implementation of BraxProblem.
• Various small fixes and improvements.

Related Pull requests

• Add compile & vmap support for while loops and external functions by @sses7757 in #221
• Update RVEA by @Zhenyu2Liang in #222
• Fix std_workflow._evaluate cannot be captured by torch.compile bug by @sses7757 in #223
• BatchedTensor fixes by @BillHuang2001 in #224

Full Changelog: v1.1.0...v1.1.1

Major Changes

Dynamo Integration 🚀

We have transitioned to torch.compile as our compiler backend, replacing TorchScript. This update simplifies the compilation workflow, making it more user-friendly while maintaining high performance and efficiency.

Why Move to torch.compile?

Previously, EvoX used tracing to obtain the computation graph—JAX tracing in pre-1.0.0 and TorchScript in v1.0.0. However, tracing modes were difficult for users, lacked compatibility with many Python functions (especially non-PyTorch functions), and required manual intervention to handle dynamic workflow. While tracing was fast and produced a full computation graph, its limitations made it less flexible and hard to use.

With torch.compile, EvoX can now provide a smoother and robust compilation process:

✅ Ease of Use – No need for explicit scripting or tracing; simply use torch.compile(workflow.step), and your code gets optimized automatically.
✅ Better Performance – The torch.compile backend is actively maintained and optimized, delivering better performance than previous tracing methods, and works on more hardware backends.
✅ Full Python Compatibility – Works directly with Python code and non-PyTorch functions, eliminating compatibility issues.
✅ Future-Proof – Aligns with PyTorch’s long-term strategy, making torch.compile the recommended approach for graph capture.

This transition enhances usability while ensuring compatibility with PyTorch's evolving ecosystem.

What's Changed

• Translations update from Hosted Weblate by @weblate in #209
• Change cec2022.py to avoid accuracy blowout in the extreme by @starquakee in #210
• Fix possible shape bug in DE_differential_sum function by @starquakee in #211
• Brax now supports Single-layer vmap (Hpo Problem) by @starquakee in #212
• Fix bug of APD calculation by @Zhenyu2Liang in #213
• Some vmap related improvements by @sses7757 in #214
• Fix brax problem not evaluating in jit.trace and vmap by @sses7757 in #215
• Add NSGA3 by @LiHao-MS in #216
• Dynamo by @BillHuang2001 in #217
• Add HypE by @Zhenyu2Liang in #218

Full Changelog: v1.0.1...v1.1.0

Minor release, bug fixes.

What's Changed

• update win-install.bat by @ZaberKo in #204
• fix(DE, ODE, JaDE): implement true random selection for base vectors by @starquakee in #205
• Fixed an array out of bounds bug in SHADE when pop_size>100 by @starquakee in #206

Full Changelog: v1.0.0...v1.0.1

Released EvoX 1.0.0a1 🎉 — EvoX 🤝 PyTorch for full compatibility! For users seeking the previous JAX-based version, please refer to the v0.9.0 branch.

What's Changed

• Fix typo in the core docstring by @dandansamax in #153
• Add 6 PSO variants to EvoXTorch by @XU-Boqing in #158
• Add SupervisedLearningProblem and ParamsAndVector components with tests for neuroevolution process on PyTorch based EvoX. by @liyc5929 in #159
• EvoXTorch project structure changes by @BillHuang2001 in #157
• Dev bh docs by @BillHuang2001 in #163
• Evoxtorch dev doc by @BillHuang2001 in #166
• Add RVEA algorithm by @BillHuang2001 in #161
• Add HPO developer guide by @generaldi in #167
• Add 4 doc files. by @XU-Boqing in #171
• Fixed CEC2022 by @LYinMX in #173
• Modify HPO，picture.Add containers. by @generaldi in #176
• multi-objective algorithm example docs by @Nam-dada in #175
• Add cec2022 (12 so problems), and the test file. by @XU-Boqing in #162
• Add CMA-ES algorithm by @XinyaoLi0212 in #177
• fix de.py selection bug and add ode.py by @starquakee in #164
• Modify 4 docs. by @XU-Boqing in #172
• Try to make docs' link goes well. by @XU-Boqing in #180
• Add JaDE by @starquakee in #181
• Add HPO for user gd, modify HPO for developer gd. by @generaldi in #179
• Add NSGA2 and MOEA/D by @Zhenyu2Liang in #182
• Upgrade Multi-Objective Algorithm examples page by @Nam-dada in #178
• Optimizing text. by @Nam-dada in #183
• Add jit wsd part by @generaldi in #184
• Add hpo in index by @generaldi in #185
• Add hv and gd metric by @Zhenyu2Liang in #186
• Finish document of numerical optimization in example and fix PSO. by @XinyaoLi0212 in #189
• Evoxtorch-dev-lh by @LiHao-MS in #190
• Add link. by @Nam-dada in #193
• Add part of the jit-vmap doc. by @XU-Boqing in #188
• Add EvoX installation guide by @JiangTao97 in #191
• Add some ES and DE by @LYinMX in #187
• Enable brax visualize by @BillHuang2001 in #194
• Monitor plot by @BillHuang2001 in #195
• Add win installer script and update install doc by @ZaberKo in #192
• Update win installer by @ZaberKo in #196
• Update the numerical optimization document and the comments of CMAES. by @XinyaoLi0212 in #197
• Update README by @Zhenyu2Liang in #199
• Fix some spelling mistakes of docs. by @XU-Boqing in #198
• Add doc "Transformation from MATLAB to PyTorch and EvoX" by @sses7757 in #200
• Add jit-vmap and supervised-learning tutorial documents. by @liyc5929 in #201
• Fix some mistakes inside the JIT doc. by @XU-Boqing in #202
• EvoXTorch Staging by @BillHuang2001 in #160

New Contributors

• @dandansamax made their first contribution in #153
• @XU-Boqing made their first contribution in #158
• @liyc5929 made their first contribution in #159
• @generaldi made their first contribution in #167
• @LYinMX made their first contribution in #173
• @Nam-dada made their first contribution in #175
• @XinyaoLi0212 made their first contribution in #177
• @JiangTao97 made their first contribution in #191
• @sses7757 made their first contribution in #200

Full Changelog: v0.9.0...v1.0.0a1

What's Changed

• Translations update from Hosted Weblate by @weblate in #139
• Update README.md by @ranchengcn in #140
• Fix bugs of MOEA/D by @Zhenyu2Liang in #141
• Translations update from Hosted Weblate by @weblate in #142
• Bug Fix & Distributed Training Improvement by @ZaberKo in #143
• Update evoxbench.py by @AlpAcA0072 in #147
• State based IO by @BillHuang2001 in #145

New Contributors

• @AlpAcA0072 made their first contribution in #147

Full Changelog: v0.8.1...v0.9.0