This is reproduction code for the paper "TEMPO: Prompt-based Generative Pre-trained Transformer for Time Series Forecasting"
Performance
| dataset | multivariate | seq_len | pred_len | lr | layers | patch_size/stride | mae |
|---|---|---|---|---|---|---|---|
| ETTH1 | False | 384 | 96 | 1e-3 | 6 | 16/8 |
Dataset
ETTh1, ETTh2, ETTm1, ETTm2 are put in the folder "data". PyTorch dataset classes have been implemented, you can check them in src/dataset.py.
Model weights
They would be put in the folder "gpt2" automatically or you can download files from the link: https://huggingface.co/gpt2/tree/main, and put them in folder "gpt2".