Use S3mphony to retrieve model metrics and then practical design for automatic model selection based on metrics like R² (higher is better) and RMSE (lower is better), without getting fragile.
Get Started
$ dotnet add package S3mphony.ModelSelectorReadme
S3mphony.ModelSelector
S3mphony.ModelSelection adds an ML-aware decision layer as a post-process for model persistence.
It ranks specified workflow metric using the most appropriate available scoring metric.
- It is generic, algorithm and metric-agnostic, and built for repeated scheduled training runs (like your metrics workers), where every models can belong to different targets, and is safely persisted in S3.
- What it gives you:
- Automatic best-model selection from any number of persisted training runs
- Quality gates to filter out invalid or underperforming candidates
- Composite scoring that prefers high R² and low RMSE
- Stability policy to avoid model flapping or churn
- Recency biasing when metric scores are close
- Works seamlessly in APIs, workers, dashboards, or prototype MLOps flows
Example usage
// e.g.: S3mphony chanel setup to retrieve model metrics from your bucket
var allCandidates = await _s3Channel.ListModelsAsync(
bucket: "my-ml-model-metrics-bucket",
prefix: "regression-models/",
ct: ct);
// Pick the best model without scheduling a meeting with a scientist
var myBestModel = ModelSelector.PickBest(allCandidates);