pydftracer

Links:

  • DFTracer Documentation
  • DFTracer GitHub

Contents:

  • Installation
    • Basic Installation
    • Development Installation
    • Optional Dependencies
      • PyTorch/Dynamo Support
      • Development Tools
    • Requirements
    • Verifying Installation
  • Quick Start
    • Basic Usage
      • Enabling DFTracer
      • Initialize DFTracer Logger
      • Function Tracing with dft_fn
      • Using Iterators
      • Environment Configuration
    • AI/ML Tracing
      • PyTorch Dynamo Integration
      • Dynamo Class
      • AI Tracing Features
    • Advanced Usage
      • Custom Tags
      • Metadata Events
    • Next Steps
  • Examples
    • Python Examples
      • Application Level Example
      • LD_PRELOAD Example
      • Hybrid Mode Example
    • Deep Learning Example: ResNet50 on ALCF Polaris
      • Environment Setup
      • Application Instrumentation
      • Job Submission Script
      • Understanding the Output
    • Integrated Applications
    • Additional Resources
  • Type Safety
    • Overview
    • Type-Preserving Decorators
      • Basic Example
      • Complex Type Signatures
      • Generic Types
      • Method Decorators
    • MyPy Integration
      • Configuration
      • Running MyPy
      • Example: Full Type Checking
    • Common Patterns
      • Pattern 1: Preserving Generic Return Types
      • Pattern 2: Multiple Return Values
      • Pattern 3: Optional Parameters
    • Benefits
    • Implementation Details
      • How It Works
      • Comparison
    • Best Practices
    • Summary
  • AI/ML Tracing Guide
    • Motivation
    • Overview
    • Basic Setup
    • Data Operations
      • Tracing Data Loading
      • Dataloader Integration
      • Data Preprocessing
    • Device Operations
      • Tracing GPU Transfers
    • Compute Operations
      • Forward and Backward Passes
      • Optimization Steps
    • Communication Tracing
      • Distributed Training
    • Checkpointing
      • Model Checkpoints
    • Training Pipeline
      • Complete Training Loop
      • Metadata and Custom Tags
    • Advanced Features
      • Custom Categories
      • Disabling Specific Categories
    • AI/DL Logging Conventions
    • Flexible API Styles
      • Decorator Style
      • Context Manager Style
      • Iterable Style
      • Constructor Hooking
    • Updating Arguments
    • Force Enable or Disable Specific Events
    • Hook/Checkpoint Style
    • Derivation
    • Metadata / Streaming Style
    • Init Events
    • Caveats
      • Call Ordering Matters
    • Summary
  • PyTorch Profiler Integration
    • Overview
    • Setup
    • Basic Usage
    • Training Loop Example
    • What Gets Traced
    • Example Output
    • Combining with I/O Tracing
    • API Reference
      • dftracer.python.torch.trace_handler()
  • PyTorch Dynamo Integration
    • Overview
    • Prerequisites
    • Setup
    • Using the Dynamo Decorator
      • Basic Model Tracing
    • Advanced Configuration
      • Using create_backend with torch.compile
      • Backend Parameters
      • Custom Dynamo Tracer
    • Training Loop Integration
      • Complete Training Example with create_backend
      • Using the Dynamo Decorator
    • What Gets Traced
      • Operation Details
      • Traced Information
    • Troubleshooting
      • Common Issues
    • Performance Considerations
      • Overhead
      • Best Practices
    • Example Output
      • Trace File Contents
    • Summary
    • API Reference
  • Developer Guide
    • Development Setup
      • Prerequisites
      • Clone the Repository
      • Create Development Environment
    • Project Structure
      • Repository Layout
      • Key Modules
    • Running Tests
      • Run All Tests
      • Run Specific Tests
      • Test with Environment Variables
    • Code Quality
      • Linting
      • Type Checking
      • Formatting
      • Configuration
    • Building Documentation
      • Build HTML Docs
      • Clean Build
      • Check for Broken Links
      • Build Other Formats
    • Contributing
      • Development Workflow
      • Commit Message Guidelines
    • Adding New Features
      • Adding a New Tracer Category
      • Adding New Environment Variables
    • Debugging
      • Enable Debug Logging
      • Use Debug Logger
      • Common Issues
    • Performance Considerations
      • Profiling Overhead
      • Reducing Overhead
    • Release Process
      • Versioning
      • Creating a Release
    • Resources
    • Getting Help
    • License
    • Contributing to Documentation
  • API Reference
    • Core API
      • dftracer Class
      • dft_fn Class
      • Common Utilities
        • Tag Types
        • Profiler Protocol
    • AI/ML API
      • DFTracerAI Class
      • AI Module
      • AI Categories
        • ProfileCategory
        • AI Components
      • Utility Functions
      • Constants
      • Module Instances
    • Dynamo API
      • Dynamo Class
      • Module Instance
      • Constants
      • Internal Classes
        • CallStackRecord
      • Backend Functions
        • create_backend
      • Utility Functions
        • create_detailed_op_name
      • Usage Examples
    • Environment Configuration
      • Environment Variables
      • Environment Variable Names
        • DFTRACER_ENABLE_ENV
        • DFTRACER_INIT_ENV
        • DFTRACER_LOG_LEVEL_ENV
      • Logger Setup
        • setup_stream_logger()
      • Configuration Examples
        • Enable DFTracer
        • Set Log Level
        • Set Initialization Mode
      • Checking Configuration
    • Main Package
      • Core Tracing
      • AI/ML Support
      • Configuration
      • Utilities
    • Modules
pydftracer
  • Search

© Copyright 2026, Hariharan Devarajan, Ray Andrew Sinurat.

Built with Sphinx using a theme provided by Read the Docs.