Computational Model Studio · Model release
TimesFM 2.0 500M PyTorch
Forecasts future values of a univariate time series from its historical observations.
Developer: Google Research · Last verified:
What it is designed to do
Produce point forecasts from a univariate time-series history and requested horizon.
- Primary contribution
- prediction estimation
- Input
- time series
- Output
- time series (point forecast sequence)
Documented uses: Forecasting future values from contiguous univariate history.
Access and deployment
- TimesFM 2.0 500M PyTorch model files — available. Public publisher repository; review applicable terms before use. Artifact documentation Local hardware requirements have not been verified. The full artifact-specific rights document has not been separately reviewed; consult the publisher repository conditions.
Technical profile and review notes
- Construction
- learned
- Representation
- neural
- Normalized architecture tags
- No normalized architecture tag assigned
- Lifecycle
- Unknown from reviewed sources
Inclusion rationale: Adds time-series prediction and a structured numeric input absent from the seed.
Limitations
- Context and horizon ideally share frequency; missing values are linearly interpolated before inference.
- Experimental quantile heads are not calibrated after pretraining.
Sources and related Studio records
Sources
Connections
- In areaComputational Models
- Supports functionForecast time series
- ArtifactTimesFM 2.0 500M PyTorch model files