forecast_type: <str> (Optional)

Description

Chooses the type of forecasting model being used. Possible options are:
  • mlp: Uses an MLP to make the final prediction across all future timestamps.
  • implicit: Uses a shared MLP across all future timestamps, conditioned on the future timestamp.
  • autoregressive: Uses an RNN-based forecasting head with support for autoregressive prediction.
The default setting is forecast_type=mlp.

Supported Task Types

  • Forecasting
Experimental Feature