Google's 200M-parameter time-series foundation model with 16k context

TimesFM (Time Series Foundation Model) is a pretrained time-series foundation model developed by Google Research for time-series forecasting.

• Paper: A decoder-only foundation model for time-series forecasting, ICML 2024.
• All checkpoints: TimesFM Hugging Face Collection.
• Google Research blog.
• TimesFM in BigQuery: an official Google product.

This open version is not an officially supported Google product.

Latest Model Version: TimesFM 2.5

Archived Model Versions:

• 1.0 and 2.0: relevant code archived in the sub directory v1. You can pip install timesfm==1.3.0 to install an older version of this package to load them.

Added back the covariate support through XReg for TimesFM 2.5.

TimesFM 2.5 is out!

Comparing to TimesFM 2.0, this new 2.5 model:

• uses 200M parameters, down from 500M.
• supports up to 16k context length, up from 2048.
• supports continuous quantile forecast up to 1k horizon via an optional 30M quantile head.
• gets rid of the frequency indicator.
• has a couple of new forecasting flags.

Along with the model upgrade we have also upgraded the inference API. This repo will be under construction over the next few weeks to add support for an upcoming Flax version of the model (faster inference), add back covariate support, and populate more docstrings, docs and notebook.

Installation:

git clone https://github.com/google-research/timesfm.git
cd timesfm
uv venv
source .venv/bin/activate
uv pip install -e .[torch]

Quick Start:

import torch
import numpy as np
import timesfm
torch.set_float32_matmul_precision("high")
model = timesfm.TimesFM_2p5_200M_torch.from_pretrained("google/timesfm-2.5-200m-pytorch")
model.compile(
    timesfm.ForecastConfig(
        max_context=1024,
        max_horizon=256,
        normalize_inputs=True,
        use_continuous_quantile_head=True,
        force_flip_invariance=True,
        infer_is_positive=True,
        fix_quantile_crossing=True,
    )
)
point_forecast, quantile_forecast = model.forecast(
    horizon=12,
    inputs=[
        np.linspace(0, 1, 100),
        np.sin(np.linspace(0, 20, 67)),
    ],
)
# point_forecast.shape: (2, 12)
# quantile_forecast.shape: (2, 12, 10)