How kotserge/pinn-curriculum-learning works
Component Details and Relationships
a) Input/Configuration Layer:
• PDE data generation system (src/data/convection_pde.py)
• Data augmentation component (src/data/augmentation.py)
b) Core Processing Layer:
• Center the curriculum learning system (src/curriculum/curriculum_learning.py)
• To curriculum_scheduler
• To curriculum_trainer
• To curriculum_evaluator
• PINN model architecture (src/model/pinn.py)
• Optimizer components (Adam, L-BFGS, SGD)
c) Evaluation Layer:
• Loss computation modules (src/loss/)
• Curriculum evaluation feedback loop
• Performance metric collectors
d) Output/Monitoring Layer:
• WandB integration component
• Artifact storage system
Additional Elements
• External dependencies (PyTorch, WandB)
• File storage systems (model checkpoints, artifacts)
• Configuration management system
The configuration management system
Use a hierarchical layout with:
• Top: Input/Configuration
• Middle: Core Processing
• Bottom: Output/Monitoring
• Right side: Evaluation feedback loop
Remember to maintain clarity while showing the complexity of the system, particularly the curriculum learning aspects and the PDE solving capabilities.
This structure will provide a clear understanding of how the system implements curriculum learning for PINNs while maintaining traceability of experiments through WandB integration.