ai-zipper

ai-zipper offers numerous AI model(CNN) compression methods, also it is easy to embed into your own source code

Version 0.0.1

What is ai-zipper?

• ai-zipper offers several AI-Optimization and/or Model-Compression methods for CNN based networks
  • List of methods that will be offered later:
    • Quantization
    • Unstructured Pruning
    • Structured Pruning
    • Knowledge Distillation
    • ...and more

Why Do I Need ai-zipper?

• You need ai-zipper when your model(network) is exceptionally slow, or too big to put into your system
• ai-zipper reduces the size of your convolutional network, and improves inference latency

How does it work?

• List below shows which package is used for specific methods:
  • Quantization:
    • ai-zipper uses Tensorflow Lite to offer Quantization
  • Other methods will be posted later...

How Can I Use ai-zipper?

• Downloading through PyPI will be supported later...
• ai-zipper offers easy-to-import APIs for each methods.
  • It also gives one-shot method for beginners (which will be supported later)