tcfuse-inference

name

tcfuse-inference

description

TC-Fuse inference and prediction pipeline — running a trained checkpoint over a split with `scripts/inference/infer.py` (Hydra, `conf/inference.yaml`), the single-GPU `trainer.predict` flow with task-specific target masking, the `PredictionWriter` callback, and the `SamplePrediction` / `PredictionRun` output format (per-window HDF5 + `index.parquet` + `manifest.yaml`) with its `compute_metrics` evaluation (bias, RMSE, MAE, R2 per channel). Use when running inference, evaluating a checkpoint, changing the prediction output format, or working in `src/tcfuse/data/predictions/` or `src/tcfuse/lightning/prediction_writer.py`.

Content has moved. The full skill documentation is in .agents/inference.md. Read that file for the inference invocation, the end-to-end flow, the on-disk PredictionRun format, the read/evaluation API, and the predict-time masking caveat.