GitHub - google-research/timesfm: TimesFM (Time Series Foundation Model) is a pretrained time-series foundation model developed by Google Research for time-series forecasting.
5 days ago
- #time-series
- #forecasting
- #machine-learning
- TimesFM is a pretrained time-series foundation model by Google Research.
- Latest version is TimesFM 2.5 with 200M parameters and 16k context length support.
- Supports continuous quantile forecast up to 1k horizon.
- Model upgrades include removal of frequency indicator and new forecasting flags.
- Inference API upgraded with upcoming Flax version for faster inference.
- Installation involves cloning the repository and setting up a virtual environment.
- Supports PyTorch and Jax backends for different accelerators.
- Example usage includes forecasting with dummy inputs and quantile outputs.