Cannot run Demand forecasting with the Temporal Fusion Transformer

Demand and forecasting with the Temporal Fusion Transformer

Hi everyone,
I hope you’re doing well!

  1. I’m new to both PyTorch and this forum, but I’ve been drawn to the Temporal Fusion Transformer (TFT) model because of its potential and usefulness.
  2. For the past two weeks, I’ve been trying to run the basic example from the tutorial here: Demand forecasting with the Temporal Fusion Transformer — pytorch-forecasting documentation.
  3. Unfortunately, despite my best efforts, I haven’t been able to get it to work.
  4. I’m using Windows, CUDA 12.4, Python, and Visual Studio Code.
  5. I’ve tried different combinations of Python versions (3.8, 3.10, 3.12) with Conda and Pip, and even Google Colab, but faced issues at various points.
  6. I’m starting to wonder if I’m missing something fundamental or if the example code has issues.
  7. Could someone kindly provide a simple, step-by-step guide for setting up the environment (specific versions of packages, installation commands, etc.)? Detailed instructions would be really helpful!
  8. I’ve encountered multiple errors, mostly package collisions and import issues, so I’ve been unable to get a clean run of the example.
  9. I’ve searched this forum and online resources, but while I’ve found fixes for individual issues, new ones keep arising with each step.
  10. I’m feeling stuck at this point, so any advice or suggestions would be greatly appreciated!

Thank you so much in advance for your help!
Regards from New York