API Reference

This section contains the complete API reference for pydftracer.

• Core API
  • dftracer Class
  • dft_fn Class
  • Common Utilities
• AI/ML API
  • DFTracerAI Class
  • AI Module
  • AI Categories
  • Utility Functions
  • Constants
  • Module Instances
• Dynamo API
  • Dynamo Class
  • Module Instance
  • Constants
  • Internal Classes
  • Backend Functions
  • Utility Functions
  • Usage Examples
• Environment Configuration
  • Environment Variables
  • Environment Variable Names
  • Logger Setup
  • Configuration Examples
  • Checking Configuration

Main Package

The main package exports all key functions and classes:

Core Tracing

• dftracer - Main tracer class for logging events

• dft_fn - Decorator for tracing functions

• initialize - Initialize the DFTracer profiler

• finalize - Finalize and cleanup the profiler

• get_time - Get current timestamp for profiling

AI/ML Support

• DFTracerAI - Base class for AI/ML tracing

• Dynamo - PyTorch Dynamo integration class

• dynamo - Pre-configured Dynamo instance for tracing

• create_backend - Create a custom torch.compile backend with DFTracer

Configuration

• DFTRACER_ENABLE - Check if DFTracer is enabled

• DFTRACER_INIT_PRELOAD - Check initialization mode

• DFTRACER_LOG_LEVEL - Current log level

Utilities

• TagValue - Represent tagged values for events

• TagType - Enum for tag types (KEY, VALUE, IGNORE)

• TagDType - Enum for data types (INT, FLOAT, STRING)

Modules

• Core API - Core tracing functionality (logger, common utilities)

• AI/ML API - AI/ML specific tracing features

• Dynamo API - PyTorch Dynamo integration

• Environment Configuration - Environment configuration and setup