In this article, we introduce TimesFM, a decoder-only foundation model for time-series forecasting. Time-series forecasting is widely used in retail, finance, healthcare, and more, and improving accuracy in demand forecasting can have significant benefits. Deep learning models have shown promise in this field, and large language models used for natural language processing tasks can also be powerful tools. However, DL-based forecasters face challenges, such as long training and validation cycles. TimesFM aims to provide out-of-the-box forecasts on unseen time-series data with no additional training, making it a valuable tool for retail demand planning. We demonstrate that TimesFM performs well on various datasets, outperforming traditional methods and even matching the performance of larger models.
https://blog.research.google/2024/02/a-decoder-only-foundation-model-for.html