Skip to content

Input Encoders

This folder contains various input encoder implementations for processing and embedding different types of data in machine learning models.

Files

1. triplet_encoder.py

This file implements a TripletEncoder class that encodes triplets of (time, code, numeric_value) data. It uses separate embedders for each component of the triplet:

  • CVE (Continuous Value Encoder): For encoding times and numeric values.
  • nn.Embedding: For encoding codes.

2. triplet_prompt_encoder.py

This file contains a TripletPromptEncoder class, which is similar to the TripletEncoder but specifically designed for prompt-based tasks. It encodes (code, numeric_value) pairs without the time component.