v1.5.6

Version 1.5.5 Release Notes

• New feature: DPC with preview horizon using new class SystemPreview that acts as drop-in replacement for System class
• New example: Neural DAEs via operator splitting method
• New example: Mixed-Integer DPC for thermal system
• New example: Grid-responsive DPC for building energy systems

What's Changed

• fix: Update outdated neuromancer dependency in linux_env.yml by @Srindot in #239
• Grid response by @EladMichael in #241
• MI-DPC and SystemPreview Class by @xboldocky in #249
• Update README.md by @drgona in #251
• Update README.md by @drgona in #253

New Contributors

• @Srindot made their first contribution in #239
• @EladMichael made their first contribution in #241
• @xboldocky made their first contribution in #249

Full Changelog: v1.5.5...v1.5.6

v1.5.5

Version 1.5.5 Release Notes

• New feature: DPC with preview horizon using new class SystemPreview that acts as drop-in replacement for System class
• New example: Neural DAEs via operator splitting method
• New example: Mixed-Integer DPC for thermal system
• New example: Grid-responsive DPC for building energy systems

What's Changed

• Develop by @drgona in #254

Full Changelog: v1.5.4...v1.5.5

v1.5.4

Version 1.5.4 Release Notes

• New feature: Function encoders for function approximation and zero-shot generalization in neural ODEs
• Bug fix: plotting issues in several examples,
• Bug fix: computing constraints violations and objective values in L2O examples
• Bug fix: CSTR dynamics model in psl submodule

This release was supported by the Ralph O’Connor Sustainable Energy Institute at Johns Hopkins University.

v1.5.3

Version 1.5.3 Release Notes

New feature: Script to generate documents for offline configuration of LLM/RAG-based systems
New feature: In-depth notebook comparing and contrasting RL vs DPC for building controls
New feature: Library now supports Python 3.11
New feature: Updated Node class than can accept instantiated Variables in its constructor

This research was partially supported by the Energy Efficiency and Renewable Energy, Building Technologies Office under the “Dynamic decarbonization through autonomous physics-centric deep learning and optimization of building operations” and the “Advancing Market-Ready Building Energy Management by Cost-Effective Differentiable Predictive Control” projects. This project was also supported from the U.S. Department of Energy, Advanced Scientific Computing Research program, under the Uncertainty Quantification for Multifidelity Operator Learning (MOLUcQ) project (Project No. 81739).
PNNL is a multi-program national laboratory operated for the U.S. Department of Energy (DOE) by Battelle Memorial Institute under Contract No. DE-AC05-76RL0-1830.

Part of this work was supported by the research group of Ján Drgoňa in the Department of Civil and Systems Engineering and the Ralph S. O’Connor Sustainable Energy Institute (ROSEI) at Johns Hopkins University (JHU).

v1.5.2

Version 1.5.2 Release Notes

• New feature: Multi-fidelity Kolgomorov Arnold Networks for SOTA function approximation
• New feature: Load forecasting on building energy systems tutorials
• New feature: Transformer block

This research was partially supported by the Energy Efficiency and Renewable Energy, Building Technologies Office under the “Dynamic decarbonization through autonomous physics-centric deep learning and optimization of building operations” and the “Advancing Market-Ready Building Energy Management by Cost-Effective Differentiable Predictive Control” projects. This project was also supported from the U.S. Department of Energy, Advanced Scientific Computing Research program, under the Uncertainty Quantification for Multifidelity Operator Learning (MOLUcQ) project (Project No. 81739).
PNNL is a multi-program national laboratory operated for the U.S. Department of Energy (DOE) by Battelle Memorial Institute under Contract No. DE-AC05-76RL0-1830.

v1.5.1

Version 1.5.1 Release Notes

• Enhancement: Now supports integration of all Lightning hooks into the Neuromancer Lightning trainer. Please refer to Lightning examples README for more information
• Deprecated WandB hyperparameter tuning via LitTrainer for now
• New feature: TorchSDE integration with Neuromancer core library, namely torchsde.sdeint(). Motivating example for system ID on stochastic process found in examples/sdes/sde_walkthrough.ipynb
• New feature: Stacked physics-informed neural networks
• New feature: SINDy -- sparse system identification of nonlinear dynamical systems

v1.5.0

Version 1.5.0 Release Notes

New Feature: PyTorch Lightning Integration with NeuroMANCER core library. All these features are opt-in.

• Code simplifications: zero boilerplate code, increased modularity
• Added ability for user to define custom training logic
• Easy support for GPU and multi-GPU training
• Easy Weights and Biases (https://wandb.ai/site) hyperparameter tuning and Tensorboard Logging

NeuroMANCER v1.4.2

Version 1.4.2 Release Notes

• New feature: Update violation energy for projected gradient #110 (based on idea #86).
• Reverted psl.nonautonomous.TwoTank (umin, umax) bounds to (0.5, 0.5) for numerical stability #105
• Added new unit tests for problem.py and system.py #107
• Automated docs build from master -> gh-pages #107
• Fixed positional arg error and added support for Time data in file_emulator.py #119
• Fixed a bug in System which caused incorrect visualization of the computational graph
• Minor updates to examples