Hello,
Thank you for providing this benchmark -- it is a very valuable resource for the community to test methodologies across diverse datasets and models. It would be great if you could also open-source the notebooks for TFT and PatchTST. I am also curious about how features were selected for TFT in your experiments. Specifically, how were datasets with different types of covariates (past, future, auxiliary) handled such as in the case of TFT?
Hello,
Thank you for providing this benchmark -- it is a very valuable resource for the community to test methodologies across diverse datasets and models. It would be great if you could also open-source the notebooks for TFT and PatchTST. I am also curious about how features were selected for TFT in your experiments. Specifically, how were datasets with different types of covariates (past, future, auxiliary) handled such as in the case of TFT?