API Documentation

RecIS provides rich API interfaces. This section details the usage of each module.

• IO Module
  • Core Features
  • Dataset Classes
  • Best Practices
  • Common Questions
• Feature Processing Module
  • Core Features
  • Operator Fusion Optimization
  • Core Components
  • Feature Operations
  • Advanced Usage
• Neural Network Module
  • Components
  • Component Overview
  • Best Practices
  • Common Questions
• Evaluation Metrics Module
  • Core Metrics
  • Metrics Integration
  • Custom Metrics
  • Frequently Asked Questions
• Hook System
  • Basic Hooks
  • Custom Hooks
• Optimizer Module
  • Sparse Optimizers
  • TensorFlow Compatible Optimizers
  • Usage Guide
  • Performance Optimization Recommendations
  • Frequently Asked Questions
• Training Framework Module
  • Core Components
  • Advanced Usage
• Parameter Saving and Loading
  • Save Format
  • Save and Load Interface
• Utilities Module
  • Logging System
  • Data Processing Tools

Core Module Overview

Core Module Description

Module	Description
IO Module	Data reading and preprocessing, supports local Orc file data sources
Feature Processing Module	Feature engineering module, supports feature processing and transformation
Neural Network Module	Model-related modules, including dynamic embedding and special operators provided by recis
Evaluation Metrics Module	Evaluation metrics, including AUC, GAUC and other commonly used metrics in recommendation systems
Hook System	Hooks, including log printing, performance analysis tools and other hooks
Optimizer Module	Optimizer module, provides sparse parameter optimization
Training Framework Module	Training framework, provides trainer and checkpoint management
Parameter Saving and Loading	Serialization module, includes Saver, Loader and Checkpoint direct reading tools
Utilities Module	Utility functions, including logging, data processing and other auxiliary functions

Quick Index

Common Classes and Functions

• recis.io.OrcDataset - ORC Dataset

• recis.features.FeatureEngine - Feature Engine

• recis.nn.EmbeddingEngine - Embedding Engine

• recis.metrics.auroc.AUROC - AUC Calculation

• recis.optim.sparse_adamw.SparseAdamW - Sparse AdamW Optimizer

• recis.framework.trainer.Trainer - Trainer

Important Configuration Classes

• recis.nn.EmbeddingOption - Embedding Configuration

• recis.framework.trainer.TrainingArguments - Training Parameter Configuration

Checkpoint Reading Tools

• recis.serialize.checkpoint_reader.CheckpointReader - Read Checkpoint Files

Version Compatibility

Note

RecIS requires Python 3.10+ and PyTorch 2.4+. APIs may differ between versions, please refer to the documentation for the corresponding version.