flowmatic-astana-autoencoder-anomaly-repair

Repair noisy/missing smart-city streams and detect reconstruction anomalies.

Model summary

Field	Value
Architecture	autoencoder
Dataset	astana
Source checkpoint	astana_autoencoder_anomaly_repair
Test metrics	{"test_loss": 0.9881124693483662, "test_mae": 0.8410043120384216, "test_rmse": 0.994038462638855}

Files

• model.safetensors — primary weights
• model.torchscript.pt — TorchScript export (when present)
• metadata.json — schema, scaler, metrics
• model_card.md — training card

Usage in Flowmatic

Deploy in the Core Unit with model ID flowmatic-astana-autoencoder-anomaly-repair (prefixed by your Hugging Face username in the UI).

Inference is routed through Flowmatic's model-inference service using your organization Hugging Face token.