How ai-tanzil811/green-ai-decision-support-frameworks works
GreenPEFT is a research and decision-support framework for sustainable parameter-efficient fine-tuning. Its user-facing workflow loads surrogate artifacts and a candidate catalog, builds pre-run features, predicts outcomes, applies resource constraints, and ranks feasible strategies with Pareto filtering and GEI. A separate empirical-analysis branch organizes benchmark measurements and examines feasibility and trade-offs; validation and export support the surrogate lifecycle. Links among CLI internals and analysis modules follow excerpts or SOURCE INDEX references; broader benchmark-to-surrogate relationships are documented in the README.
Main entry points: recommender.py Surrogate Prediction · greenpeft_data.py Shared Data Loader · metrics Benchmark Results · cli.py CLI
Components 22
Recommendation Engine
- CLI
green_peft_cli/green_peft_pkg/green_peft/cli.py - CLI Banner
green_peft_cli/green_peft_pkg/green_peft/banner.py - Model Catalog
green_peft_cli/green_peft_pkg/green_peft/data - Pre-run Features
green_peft_cli/green_peft_pkg/green_peft/features.py - Surrogate Prediction
green_peft_cli/green_peft_pkg/green_peft/recommender.py - Confidence Scope
green_peft_cli/green_peft_pkg/green_peft/confidence.py - Constraint Filtering
green_peft_cli/green_peft_pkg/green_peft/recommender.py - Pareto and GEI
green_peft_cli/green_peft_pkg/green_peft/recommender.py
Benchmark Analysis
- Benchmark Data
new update works/greenpeft_ml_ready_dataset.csv - Shared Data Loader
analysis/greenpeft_data.py - Benchmark Analysis
analysis/benchmark_analysis.py - Frontier Analysis
analysis/pareto_analysis.py - Research Figures
analysis/figures.py - Recommendation Scenarios
analysis/recommendation_scenarios.py
Surrogate Lifecycle
- Training Envelope
analysis/build_training_envelope.py - Surrogate Artifacts
new update works/greenpeft_surrogate_models.joblib - Model Metadata
models/model_metadata.json - Model Export
analysis/build_hf_export.py
Results and Validation
- Prediction Validation
experiments/validate_recommendation.py - Benchmark Results
results/canonical_benchmark/metrics - Recommendation Results
results/recommendations
Other components
- Practitioner
Connections 27
- Practitioner → CLI: requests recommendation
readme.md - CLI → CLI Banner: prints banner
green_peft_cli/green_peft_pkg/green_peft/cli.py - CLI → Surrogate Prediction: invokes engine
green_peft_cli/green_peft_pkg/green_peft/cli.py - Surrogate Prediction → Model Catalog: loads catalog
green_peft_cli/green_peft_pkg/green_peft/recommender.py - Surrogate Prediction → Surrogate Artifacts: loads models
green_peft_cli/green_peft_pkg/green_peft/recommender.py - Surrogate Prediction → Pre-run Features: builds features
green_peft_cli/green_peft_pkg/green_peft/recommender.py - Surrogate Prediction → Confidence Scope: annotates predictions
green_peft_cli/green_peft_pkg/green_peft/recommender.py - Surrogate Prediction → Constraint Filtering: filters candidates
green_peft_cli/green_peft_pkg/green_peft/recommender.py - Surrogate Prediction → Pareto and GEI: scores candidates
green_peft_cli/green_peft_pkg/green_peft/recommender.py - Benchmark Data → Shared Data Loader: provides measurements
analysis/greenpeft_data.py - Shared Data Loader → Benchmark Analysis: supplies benchmark data
analysis/benchmark_analysis.py - Shared Data Loader → Frontier Analysis: supplies measured runs
analysis/pareto_analysis.py - Shared Data Loader → Research Figures: supplies analysis data
analysis/figures.py - Shared Data Loader → Recommendation Scenarios: supplies analysis data
analysis/recommendation_scenarios.py - Benchmark Analysis → Benchmark Results: writes summaries
readme.md - Frontier Analysis → Pareto and GEI: uses decision engine
analysis/pareto_analysis.py - Frontier Analysis → Benchmark Results: writes frontiers
readme.md - Research Figures → Benchmark Results: renders results
analysis/figures.py - Recommendation Scenarios → Pareto and GEI: uses decision engine
analysis/recommendation_scenarios.py - Recommendation Scenarios → Recommendation Results: writes scenarios
readme.md - Benchmark Data → Training Envelope: defines measured scope
readme.md - Training Envelope → Model Metadata: records envelope
readme.md - Surrogate Artifacts → Model Export: packages model
analysis/build_hf_export.py - Model Metadata → Model Export: supplies metadata
analysis/build_hf_export.py - Model Catalog → Model Export: packages configs
analysis/build_hf_export.py - Prediction Validation → Pareto and GEI: checks predictions
experiments/validate_recommendation.py - Prediction Validation → Benchmark Results: backtests against
experiments/validate_recommendation.py