Skip to content

NXP/eiq-olive

 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

1,994 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

eiq-olive

Latest version Rebased olive-ai version

eiq-olive is an extension of Microsoft's Olive optimization framework, tailored to support additional workflows and deployment targets. This fork introduces new optimization passes, enhanced support for TensorFlow Lite (TFLite), and other improvements aimed at streamlining model conversion and deployment to NXP's edge devices.

Key Features

  • Extended optimization passes for NXP edge devices
  • Enhanced TensorFlow Lite (TFLite) support
  • Integration with NXP's Neutron SDK and Vela compiler
  • ONNX to TFLite conversion pipeline
  • Quantization support for edge deployment

Added passes

Pass name Implementation Dependencies Input type Output type Example
NeutronConversion Code uv (eiq-neutron-sdk packages downloaded on demand) TFLite TFLite Config
ONNX2Quant Code eiq-onnx2tflite ONNX ONNX Config
TFLiteConversion (ONNX2TFLite) Code eiq-onnx2tflite ONNX TFLite Config
VelaConversion Code nxp-ethos-u-vela TFLite TFLite Config

Installation

eiq-olive and all pass-related dependencies are available in the NXP eIQ PyPI repository at https://eiq.nxp.com/repository/.

pip install --extra-index-url https://eiq.nxp.com/repository/ eiq-olive

⚠️ Security Notice: When using --extra-index-url, be aware of potential dependency confusion attacks. Packages may be hijacked via pypi.org if an attacker uploads a malicious package with the same name and a higher version number. Pip will prioritize the package with the highest version across all configured indexes. Consider using additional security measures such as pinning package versions or using hash verification for production environments.

Examples

The examples directory contains various optimization workflows demonstrating different passes and use cases:

Each example includes its own README with specific setup instructions and required dependencies. Please refer to the individual example directories for detailed usage information.

License

This project maintains the same license as the upstream Microsoft Olive project. See LICENSE for details.

Support

For issues related to:

Changelog

See CHANGELOG.md for version history and updates.

About

Olive: Simplify ML Model Finetuning, Conversion, Quantization, and Optimization for CPUs, GPUs and NPUs.

Resources

License

Contributing

Security policy

Stars

Watchers

Forks

Packages

 
 
 

Contributors

Languages

  • Python 98.2%
  • Jupyter Notebook 1.6%
  • Other 0.2%