multimodal-timesfm 0.5.0

Multimodal extension of Google's TimesFM for time series forecasting with text

Multimodal TimesFM

A multimodal extension of Google's TimesFM for time series forecasting with text inputs.

Installation

pip install multimodal-timesfm[all]

Quick Start

1. Setup

Clone the Time-MMD dataset:

./scripts/clone_time_mmd.sh

Split the dataset into train / val / test:

PYTHONPATH=. uv run python scripts/split_time_mmd_datasets.py \
    --train-ratio 0.6 \
    --val-ratio 0.2

2. Pre-compute Text Embeddings

PYTHONPATH=. uv run python scripts/cache_time_mmd_datasets.py --text-encoder-type english

3. Hyperparameter Tuning

Run a W&B Sweeps search for the multimodal model:

PYTHONPATH=. uv run python scripts/tune_time_mmd_sweep.py \
    --sweep-config examples/time_mmd/configs/sweeps/multimodal_1layer.yml

To compare against a fine-tuned baseline:

PYTHONPATH=. uv run python scripts/tune_baseline_sweep.py \
    --sweep-config examples/time_mmd/configs/sweeps/baseline.yml

Acknowledgments

We thank the Time-MMD team for providing the multimodal time series dataset used in our examples and experiments.

License

MIT