UI controls¶

This page lists user-facing controls in timeshap_app.py.

Control	Type	Purpose
`Dataset`	radio	Choose demo data or uploaded file
`Upload long-format data`	file uploader	Accepts `.csv` and `.parquet`
`Entity column`	selectbox	Sequence identifier column
`Time column`	selectbox	Time ordering column
`Target column`	selectbox	Binary label column

Control	Type	Purpose
`Sequence length`	slider	Window length used for sequence construction
`GRU hidden size`	slider	Hidden dimension for built-in GRU
`Model source`	radio	Choose `Train in app` or `Load Lightning checkpoint`

When Model source = Train in app:

Control	Type	Purpose
`Epochs`	slider	Number of training epochs
`Batch size`	slider	Training batch size
`Learning rate`	number input	Adam optimizer learning rate
`Train / Retrain`	button	Triggers model training

When Model source = Load Lightning checkpoint:

Control	Type	Purpose
`Lightning checkpoint`	file uploader	Upload `.ckpt`, `.pt`, or `.pth`
`Lightning class path (optional)`	text input	Optional `package.module.ClassName` for `load_from_checkpoint`
`Model attribute`	text input	Attribute containing `nn.Module` on loaded object
`Scaler .npz (optional: mean/std)`	file uploader	Optional normalization stats
`Load checkpoint`	button	Triggers checkpoint loading and validation

Main panel controls¶

Control	Type	Purpose
`Validation sample`	slider	Select validation window index
`Pruning tolerance (tol)`	slider	TimeSHAP pruning tolerance
`Kernel SHAP samples`	slider	Number of SHAP samples
`Run TimeSHAP`	button	Compute local TimeSHAP explanations