codewithavra/ionospheric-tec-forecasting ? reverse-engineered prompt

Build me a notebook driven project that forecasts next day ionospheric TEC from GNSS data for the IISC Bangalore station.

I want to be able to drop the raw dataset folders into the repo, run the notebook from top to bottom, and get a full workflow that sorts the daily TEC files, builds the 1440 point day to day training pairs, trains both an LSTM and a Transformer, and compares their predictions on the test days. Please also include plots for the actual versus predicted TEC, training and validation loss, and the ionospheric delay estimates for L1, L2, and L5. Save the prediction results and delay values into CSV files too.

If anything is unclear, look up current docs online if you need to. Please keep the code clean and easy to follow so someone can rerun the whole experiment without having to dig through every cell.