GiftEvalParquet

# GiftEval Parquet Collection This repository hosts the parquet formatted GiftEval test data for ease of evaluating with LLM backboned models. Each dataset in the original GiftEval dataset can be loaded separately using the config names: `datasetName_freq_term`. Each row is a sample window from the test split of data, generated using the original GiftEval [proressing script](https://github.com/SalesforceAIResearch/gift-eval). Each entry contains the following fields: - **item_id** (string): e.g. "item_0_dim0_window0/2018-04-12 20:00:00" - **frequency** (string): e.g. "15T" - **history_start** (string): e.g. "2016-07-01 00:00:00" - **history_end** (string): e.g. "2018-04-12 19:45:00" - **history_value** (list): e.g. [1,3,3,4,5,6,7,...] - **future_start** (string): e.g. "2018-04-12 20:00:00" - **future_end** (string): e.g. "2018-04-20 07:45:00" - **future_value** (list): e.g. [8,9,10,...] - **config** (string): e.g. "ett1/15T/long" Note: that multivariate datasets are split into univariate form so each row is a univariate sample. The dimension information is saved within item_id. ## Example usage ```python from datasets import load_dataset ds = load_dataset( "Salesforce/GiftEvalParquet", "bitbrains_fast_storage_5T_long", split="train" ) print(len(ds)) print(ds[0].keys()) ``` # Citation If you find this benchmark useful, please consider citing: ``` @article{aksu2024giftevalbenchmarkgeneraltime, title={GIFT-Eval: A Benchmark For General Time Series Forecasting Model Evaluation}, author={Taha Aksu and Gerald Woo and Juncheng Liu and Xu Liu and Chenghao Liu and Silvio Savarese and Caiming Xiong and Doyen Sahoo}, journal = {arxiv preprint arxiv:2410.10393}, year={2024}, } ```