TimesFM 1.0
TimesFM (Time Series Foundation Model) is a pretrained time-series foundation model developed by Google Research for univariate time-series forecasting.TimesFM (Time Series Foundation Model) is a pretrained time-series foundation model developed by Google Research for univariate time-series forecasting.
The 1.0 release contains a 200M-parameter checkpoint along with its inference code. It’s a transformer based model, and was trained in the decoder only fashion on a pretrain dataset containing over 100 billion real-world timepoints. It performs univariate time series forecasting for context lengths up to 512 timepoints and any horizon lengths, with an optional frequency indicator input.
This model can be deployed into an endpoint and be used in a notebook. Click OPEN NOTEBOOK to use the model in Colab Enterprise.
TimesFM is trained with a pretrain dataset of over 100 billion real-world timepoints. Please refer to our paper for further details.
Taking as input context a univariate time series with its frequency, the model forecasts
Here we provide a user-friendly endpoint to query TimesFM. Please refer to our GitHub page for alternative ways to access the model forecast.
An endpoint prediction request looks like, with the input
field mandatory and everything else optional.
Each entry in the instances
list is a forecast task. The input
key maps to
the context time series, and the freq
key maps to the frequency (default: 0).
In particular regarding the frequency, TimesFM expects a categorical indicator valued in {0, 1, 2}:
Notice you do NOT have to strictly follow our recommendation here. Although this is our setup during model training and we expect it to offer the best forecast result, you can also view the frequency input as a free parameter and modify it per your specific use case.
Please refer to the notebook to see how to use other fields. Click OPEN NOTEBOOK to use the model in Colab Enterprise.
TimesFM 1.0 focuses on point forecasts:
For each forecast task, the endpoint will return
Please refer to the notebook to see how to access these forecasts. Click OPEN NOTEBOOK to use the model in Colab Enterprise.
Please refer to the notebook for end-to-end examples. Click OPEN NOTEBOOK to use the model in Colab Enterprise.
Resource ID | Release date | Release stage | Description |
---|---|---|---|
google/timesfm-v20240828 | 2024-08-28 | GA | Inference only. Timestamp and covariate support. |
google/timesfm | 2024-05-30 | GA | Inference only |
Google Cloud Console has failed to load JavaScript sources from www.gstatic.com.
Possible reasons are: