Optimize PyTorch* Performance on the Latest Intel® CPUs and GPUs

PyTorch* is a favorite among AI developers and researchers because it:

• Is easier to learn and debug
• Interfaces with Python*
• Can distribute tasks among multiple CPUs and GPUs
• Has many extensions

This session introduces the Intel® Extension for PyTorch*—part of Intel® Optimization for PyTorch*—that extends the stock sci-computing framework with optimizations for extra performance on Intel architecture.

Topics include:

• An overview of the Intel optimizations, including installation and performance boost metrics.
• The newest features, including:
  • Intel® Advanced Vector Extensions 512 with Vector Neural Network Instructions (VNNI)
  • Intel® Advanced Matrix Extensions
  • Ease-of-use Python API
  • Vectorization
  • Parallelism
  • Quantization
  • Operator fusion
  • Constant folding
  • Runtime extension
• A live demo that showcases the use and performance boosts on CPUs and GPUs.

Skill level: Intermediate

• Get Intel Optimization for PyTorch and Intel Extension for PyTorch as part of the Intel® AI Analytics Toolkit.
• Or get them as stand-alone versions:
  
  • Intel Optimization for PyTorch
    
  • Intel Extension for PyTorch