IPSL-AID

IPSL-AID is a high-performance research framework for climate data downscaling based on diffusion models, and UNet-based neural networks, designed for GPU clusters and HPC systems.

The framework supports global training, regional or global inference, and multiple diffusion formulations and UNet-based architectures, with a strong emphasis on reproducibility, comprehensive testing, and configurability.

Development Status

IPSL-AID is in active development.

This framework is currently undergoing rapid development and has not yet reached a stable release. Please be aware that:

• APIs and interfaces may change without notice

• Feature names and module structures are subject to modification

• Core architectural decisions may be revised

Recommendations for users:

• Regularly update from the main branch

• Rebase local branches frequently to avoid conflicts

• Submit pull requests for broadly useful enhancements

• Report issues for unexpected behavior or bugs.

---

User Guide

• Overview
  • Key Features
  • Use Cases
  • Background
• Installation
  • Prerequisites
  • Quick Install
  • Verification
  • HPC Configuration
  • Hardware Requirements
  • Dependencies
  • Optional Dependencies
  • Development Installation
  • Troubleshooting
  • Getting Help
• Quickstart
  • Basic Workflow
  • Running the Model
  • Example Configuration
  • Data Preparation
  • Evaluation Metrics
  • Next Steps
• Testing Philosophy (Read This First)
  • Why Testing is Essential
  • Recommended Workflow
  • Test Coverage
  • Debugging Tips
• Project Structure
  • Key Modules Explained
  • Data Flow
  • Configuration Management
  • Testing Strategy
  • HPC Integration
  • Development Workflow
• Pre-Push Workflow
  • Why This Workflow Matters
  • Prerequisites
  • 1. Fetch Latest Changes From Remote
  • 2. Check Branch Status
  • 3. Rebase Onto Latest Remote Branch
  • Conflict Resolution Guide
  • Step 1 — Identify Conflicted Files
  • Step 2 — Examine the Conflict
  • Step 3 — Resolve the Conflict
  • Step 4 — Mark as Resolved
  • Step 5 — Continue the Rebase
  • Abort Rebase (Emergency Option)
  • 4. Standardize Code with Pre-commit Hooks
  • 5. Run the Test Suite
  • 6. Commit Changes (If Needed)
  • 7. Push Your Changes
  • Quick Reference: Daily Workflow
  • Common Pitfalls to Avoid
  • Important Rules Summary
• Cartopy Configuration
  • Installation
  • HPC Considerations
  • Common Issues and Solutions
  • Usage in Diagnostics
  • Performance Tips

Core Concepts

• Diffusion Models
  • EDM Preconditioning
  • Loss Function
  • Sampling
  • Theoretical Background
  • Implementation Details
  • Configuration Example
  • Comparison of Formulations
• Neural Architectures
  • Default Configuration
  • Architecture Components
  • Embedding Layers
  • Conditioning Strategies
  • Input/Output Specification
  • Climate-Specific Adaptations
  • Configuration Examples
  • Performance Considerations
• Training Strategy
  • Random Block Sampling
  • Coarse-Down-Up Procedure
  • Conditioning Inputs
  • Training Schedule
  • Computational Requirements
  • Hyperparameter Tuning
  • Monitoring and Logging
  • Early Stopping
  • Multi-GPU Training
  • Reproducibility
• Inference Modes
  • Global Inference
  • Regional Inference
  • Sampling Procedure
  • Sampling Parameters
  • Ensemble Generation
  • Postprocessing
  • Performance Optimization
  • Real-time Applications
  • Validation and Evaluation
  • Example Usage

API Reference

• IPSL_AID
  • IPSL_AID package

---

Getting Started

• Overview - Framework capabilities and key features

• Installation - Setup instructions using uv

• Quickstart - Basic workflow and example usage

• Testing Philosophy (Read This First) - Essential reading before running experiments

Core Components

• Diffusion Models - VE, VP, EDM, iDDPM implementations

• Neural Architectures - UNet-based architectures for climate data

• Training Strategy - Global training with random block strategy

• Inference Modes - Global, regional, and sampler-based inference

Project Information

• Project Structure - Codebase organization

• Pre-Push Workflow - Development workflow and contribution guidelines

• Cartopy Configuration - Geospatial visualization setup

API Documentation

• IPSL_AID - Complete module reference

Indices and Tables

• Index

• Module Index

• Search Page