Spaces:
Running on Zero
Running on Zero
Lunaron Deployer
feat(refiner): deploy Dual-Domain Neural Refiner v2.0 with Quality Critic and Evolution Architecture
a63c2eb | """ | |
| Lunaron Evolution Engine | |
| Evolutionary continuous improvement architecture: | |
| - LunaronCompositeLoss (Multi-Res STFT, SI-SDR, Leakage, Artifact, Consistency) | |
| - ModelRegistry (Champion / Challenger / Archive model states) | |
| - CandidateEvaluator (Holdout dataset regression testing) | |
| """ | |
| from .training_loss import LunaronCompositeLoss | |
| from .model_registry import ModelRegistry | |
| from .evaluate_candidate import CandidateEvaluator, EvaluationReport, RegressionResult | |
| __all__ = [ | |
| "LunaronCompositeLoss", | |
| "ModelRegistry", | |
| "CandidateEvaluator", | |
| "EvaluationReport", | |
| "RegressionResult", | |
| ] | |