Docs /Templates/ sinapsis-timesfm

TimesFM

sinapsis-timesfmTimesFMTimesFMforecastingtime_seriesdataframes

Python template · sinapsis_timesfm.templates.timesfm

Overview

Template for time series forecasting using the TimesFM model.

Attributes: model_name (str): Hugging Face model identifier for TimesFM. Default: 'google/timesfm-2.5-200m-pytorch'. forecasting_config (ForecastConfigBM): Forecasting behavior configuration passed to TimesFM.compile(). Default: factory=ForecastConfigBM. forecast_horizon (int): Number of future steps to predict for each input series. Default: 12. validation_mode (bool): If true, reserve the last `forecast_horizon` points for validation. Default: False. device (str): Torch device used for inference, for example 'cpu' or 'cuda:0'. Default: 'cpu'. time_column_name (str): Name of the time column in dataframe content. Default: 'Date'. output_format (Literal['dataframe', 'darts_time_series']): Format of the output forecasts. 'dataframe' returns a pandas DataFrame, while 'darts_time_series' returns a Darts TimeSeries object. Default: 'darts_time_series'.

Capabilities

Capabilities: not set

Installation

Install the published package using pip or uv.

Terminalbash
pip install sinapsis-timesfm==1.0.0
uv pip install sinapsis-timesfm==1.0.0
uv add sinapsis-timesfm==1.0.0

Configuration

Effective fields from the inspector schema, including nested Pydantic model fields.

device

Torch device used for inference, for example 'cpu' or 'cuda:0'.

str
Optional
Default: "cpu"
metadata
sinapsis_core.template_base.base_models.TemplateMetadata | None
Optional
Default: null
model_name

Hugging Face model identifier for TimesFM.

str
Optional
Default: "google/timesfm-2.5-200m-pytorch"
output_format

Format of the output forecasts. 'dataframe' returns a pandas DataFrame, while 'darts_time_series' returns a Darts TimeSeries object.

typing.Literal['dataframe', 'darts_time_series']
Optional
Default: "darts_time_series"
validation_mode

If true, reserve the last `forecast_horizon` points for validation.

bool
Optional
Default: false
forecast_horizon

Number of future steps to predict for each input series.

int
Optional
Default: 12
time_column_name

Name of the time column in dataframe content.

str
Optional
Default: "Date"
forecasting_config

Forecasting behavior configuration passed to TimesFM.compile().

ForecastConfigBM
Optional
Default: null
max_context
int
Optional
Default: 0
max_horizon
int
Optional
Default: 0
window_size
int
Optional
Default: 0
return_backcast
bool
Optional
Default: false
normalize_inputs
bool
Optional
Default: false
infer_is_positive
bool
Optional
Default: true
per_core_batch_size
int
Optional
Default: 1
fix_quantile_crossing
bool
Optional
Default: false
force_flip_invariance
bool
Optional
Default: true
use_continuous_quantile_head
bool
Optional
Default: false

Configuration example

Generated from documented defaults. Replace any placeholders and complete required nested settings before running.

yaml
templates:
  - template_name: "TimesFM"
    class_name: "TimesFM"
    attributes:
      device: "cpu"
      model_name: "google/timesfm-2.5-200m-pytorch"
      output_format: "darts_time_series"
      validation_mode: false
      forecast_horizon: 12
      time_column_name: "Date"
      forecasting_config:
        max_context: 0
        max_horizon: 0
        window_size: 0
        return_backcast: false
        normalize_inputs: false
        infer_is_positive: true
        per_core_batch_size: 1
        fix_quantile_crossing: false
        force_flip_invariance: true
        use_continuous_quantile_head: false

Inheritance

Immediate parent: sinapsis_core.template_base.template.Template

Environment variables

Environment definitions belong to their canonical Python modules.

No directly related environment variables listed.