Skip to content

Latest commit

 

History

History
67 lines (36 loc) · 3.54 KB

File metadata and controls

67 lines (36 loc) · 3.54 KB
graph LR
    Predictor["Predictor"]
    Vector_Retrieve["Vector Retrieve"]
    CSV_Predictor["CSV Predictor"]
    Hive_Parquet_Predictor["Hive Parquet Predictor"]
    Hive_Predictor["Hive Predictor"]
    Predictor -- "delegates input/output handling to" --> CSV_Predictor
    Predictor -- "delegates input/output handling to" --> Hive_Parquet_Predictor
    Predictor -- "delegates input/output handling to" --> Hive_Predictor
    Predictor -- "directs prediction results to" --> Vector_Retrieve
Loading

CodeBoardingDemoContact

Details

The Deployment & Serving subsystem is encapsulated within the easy_rec.python.inference package. This package is responsible for loading trained models, performing real-time or batch inference, and outputting predictions, aligning with the project's focus on an end-to-end recommendation framework.

Predictor

Serves as the core inference engine. It orchestrates the entire prediction workflow, including loading the trained model, constructing the TensorFlow graph for inference, managing input/output tensors, executing the model, and handling prediction results. It acts as the primary interface for users to perform predictions.

Related Classes/Methods:

Vector Retrieve

Specializes in post-processing model outputs into numerical vector representations. This is crucial for recommendation systems that rely on vector similarity search for candidate generation or retrieval tasks.

Related Classes/Methods:

CSV Predictor

Handles the specific logic for performing predictions when the input data is provided in CSV format. It manages CSV-specific data reading and parsing, and output writing.

Related Classes/Methods:

Hive Parquet Predictor

Manages the prediction workflow for input and output data stored in Hive tables using the Parquet format. This component is optimized for large-scale, columnar data storage prevalent in big data ecosystems.

Related Classes/Methods:

Hive Predictor

Provides the general prediction logic for interacting with Hive tables, handling various Hive table inputs and outputs beyond just Parquet.

Related Classes/Methods: