ETSformer: Time-Series Transformer with Exponential Smoothing

To use ETSformer, you must have Python 3.8 installed. You can install the necessary dependencies using pip from the provided requirements.txt file.

pip install -r requirements.txt

Training is performed using shell scripts located in the ./scripts directory. Each script corresponds to a specific benchmark (e.g., ./scripts/ETTm2.sh).

Before running the scripts, ensure they have execution permissions by running:

chmod u+x scripts/*

Then, execute the desired benchmark script from your terminal.

ETSformer uses pre-processed datasets (originally from the Autoformer repository).

1. Download the datasets from Tsinghua Cloud or Google Drive.
2. Place the downloaded files into a dataset/ directory in the project root.

Example structure: dataset/ETT-small/ETTm2.csv

ETSformer provides a dedicated script for performing a grid search. Run the following command to initiate the process:

./grid_search.sh