search

DC-research / TEMPO

The official code for "TEMPO: Prompt-based Generative Pre-trained Transformer for Time Series Forecasting (ICLR 2024)". TEMPO is one of the very first open source Time Series Foundation Models for forecasting task v1.0 version.
78 stars 12 forks source link

Any Clear Tutorial on How to Use Custom Dataset and What are the Guidelines/HyperParameter tuning? #7

Open Curiosity007 opened 3 weeks ago

Curiosity007 commented 3 weeks ago

Any Clear Tutorial on How to Use Custom Dataset and What are the Guidelines/HyperParameter tuning?

idevede commented 2 weeks ago

Hi Curiosity,

Thanks for your interest in our project! To use a custom dataset and manage hyperparameter tuning, refer to our example notebook (Google Colab Link) and use the configuration template in configs/custom_dataset.yml.

For the Guidelines/HyperParameter tuning: First, define your time series period, then follow the default settings in the training scripts. The recommended approach is to start by loading the pre-trained checkpoint, then incorporate your custom time series data and update the model based on validation dataset performance. This incremental approach allows for optimal model adaptation to your specific data.

image

Best