Intel® AI Analytics Toolkit Release Notes

Overview

This document provides details about new features and known issues for the Intel® AI Analytics Toolkit. The toolkit includes the following components:

• Intel® Optimization for TensorFlow*

• Intel® Optimization for PyTorch*

• Intel® Distribution for Python*

• Intel® Low Precision Optimization Tool

• Model Zoo for Intel® Architecture

• Intel® Distribution for Modin*

Version History

Where to Find the Release

Please check the release page for more information on how to acquire the package.

Compatibility Notes

• Intel® Optimization for TensorFlow* is compatible with version 2.5

• Intel® Optimization for PyTorch * is compatible with version 1.8

• Intel® Distribution for Python* is compatible with cpython version 3.7

• Intel® Optimization of Modin* is compatible with version 0.8.2

• Intel® Low Precision Optimization Tool  is compatible with version 1.4.1

What's New

• Intel® Optimization for TensorFlow* 

  • ​oneDNN - upgraded oneDNN to v2.2
  • The default for Intel TF is now native format, The user will need to set the env-variable TF_ENABLE_MKL_NATIVE_FORMAT=0 to use blocked formats.
  • OneDNN primitive cache enabled. Improved performance of models with batch size 1.
  • Various ops fusions with FP32, BFloat16, and INT8 data format
    • Conv2D+Squeeze+BiasAdd Fusion
    • MatMul+BiasAdd+Add Fusion.
    • Enabled MatMul + Bias + LeakyRelu Fusion.
  • CNMS performance optimization
  • Enabled DNNL CPU dispatch control.
  • Graph pattern match for grappler op fusion optimization
  • Supporting quantized pooling op for signed 8 bits.
  • Enable MklConv/MklFusedConv with explicit padding
  • Remove nGraph build support tensorflow#42870
  • Execute small gemm's single threaded.
  • Removed unnecessary OneDNN dependencies.
  • Removed DNNL 0.x support
  • Bug fixes
    • Issues resolved in TensorFlow 2.5
    • oneDNN resolved issues. 2.2 resolved issues
    • Fixed memory leak in MKLAddN
    • Fixed the bug to duplicate kernel registration of BatchMatMulV2.
    • Fixed unit test failures due to benchmark test API changes
    • incorrect result of _MKLMaxPoolGrad 40122.
• Intel® Optimization for PyTorch * 
  • Upgraded the oneDNN from v1.5-rc to v1.8.1
  • Updated the README file to add the sections to introduce supported customized operators, supported fusion patterns, tutorials and joint blogs with stakeholders
• Intel® Model Zoo

  • One new PyTorch workload containers and model packages that are available on the Intel® oneContainer Portal:

    • DLRM BFloat16 Training

  • Two new TF workload containers and model packages that are available on the Intel® oneContainer Portal:

    • 3D U-Net FP32 Inference
    • SSD ResNet34 BFloat16 Training
• Intel® Low Precision Optimization Tool

  • Extended Capabilities

    • Model conversion from QAT to Intel Optimized TensorFlow model

  • User Experience

    • More comprehensive logging message
    • UI enhancement with FP32 optimization, auto-mixed precision (BF16/FP32), and graph visualization
    • Online document: https://intel.github.io/lpot

  • Model Zoo

    • INT8 key models updated (BERT on TensorFlow, DLRM on PyTorch, etc.)
    • 20+ HuggingFace model quantization

  • Pruning

    • Pruning/sparsity API refinement
    • Magnitude-based pruning on PyTorch
  • Quantization
    • PyTorch FX-based quantization support
    • TensorFlow & ONNX RT quantization enhancement
• Intel® Distribution for Python* 
  • Intel® Distribution for Python* release notes

System Requirements

Please see system requirements.

How to Start Using the Tools

Please reference the usage guides for each of the included tools:

• Intel® Optimization for TensorFlow* - See Performance Considerations
• Intel® Optimization for PyTorch * - See Performance Considerations

Known Limitation

• Intel® Optimization for TensorFlow* 
  • Int8 will only work when environment variable TF_ENABLE_MKL_NATIVE_FORMAT=0 is set.

Notices and Disclaimers

Intel technologies may require enabled hardware, software or service activation.

No product or component can be absolutely secure.

Your costs and results may vary.

© Intel Corporation. Intel, the Intel logo, and other Intel marks are trademarks of Intel Corporation or its subsidiaries. Other names and brands may be claimed as the property of others.

No license (express or implied, by estoppel or otherwise) to any intellectual property rights is granted by this document.

The products described may contain design defects or errors known as errata which may cause the product to deviate from published specifications. Current characterized errata are available on request.

Intel disclaims all express and implied warranties, including without limitation, the implied warranties of merchantability, fitness for a particular purpose, and non-infringement, as well as any warranty arising from course of performance, course of dealing, or usage in trade.