rllm.nn.encoder.FTTransformerPreEncoder¶

class rllm.nn.encoder.FTTransformerPreEncoder(out_dim: int, metadata: Dict[ColType, List[Dict[str, Any]]], in_dim: int = 1)[source]¶

Bases: TablePreEncoder

The FTTransformerPreEncoder class is a specialized pre-encoder for the FTTransformer model. It initializes a column-specific encoder dict for categorical and numerical features based on the provided metadata. Specifically, it uses EmbeddingEncoder for categorical features and LinearEncoder for numerical features.

Parameters:

out_dim (int) – The output dimensionality.
metadata (Dict[rllm.types.ColType, List[Dict[str, Any]]]) – Metadata for each column type, specifying the statistics and properties of the columns.
in_dim (int, optional) – The input dimensionality for numerical features (default: 1).

Example

>>> from rllm.nn.encoder import FTTransformerPreEncoder
>>> encoder = FTTransformerPreEncoder(out_dim=32, metadata={})