Fix common runtime errors¶

This guide covers common errors seen when running the app.

Unable to resolve TimeSHAP local APIs¶

What to do:

1. Open Environment diagnostics in the app.
2. Confirm Python executable and timeshap import status.
3. Run Streamlit with the same interpreter used for package install:

uv run python -m streamlit run timeshap_app.py

SHAP Kernel import errors¶

Symptom includes import failure under shap.explainers._kernel.

What to do:

uv run python -m pip install "shap<0.50"

Then fully restart Streamlit.

No windows could be created¶

What to do:

• lower Sequence length, or
• provide more timesteps per entity

Need more entities/windows to form train/validation split¶

What to do:

• upload data with more entities, and
• ensure multiple entities produce valid windows

Invalid scaler file¶

What to do:

• ensure .npz includes arrays named mean and std
• ensure both arrays have length equal to feature count

Checkpoint compatibility errors¶

What to do:

• confirm checkpoint and selected features use the same input dimension
• confirm loaded model returns expected output structure
• if using Lightning class path, verify module import path and class name