OwLite

OwLite is the easiest AI compression toolkit.

Why OwLite?

Solve AI compression with a toolkit

Just a Few Lines
of Code Away

Simply add a few lines of OwLite code into your existing PyTorch training scripts to unlock the full potential of your AI with model compression.

Start with Powerful Recommended Settings

No expertise in model compression? No problem.

With SqueezeBit's specialized algorithms, you can easily apply advanced compression techniques to your models. Start with just one click and achieve expert-level compression effortlessly.

Worried About the Security & Privacy?

Securing your team’s data is our foremost priority. Training data and the model weight fully remain on the user’s server. OwLite only uses ONNX model structure information throughout the whole process.

Never Compromise on performance

Fully customizable Quantization Aware Training, only on OwLite

Performance Recovery via Quantization Aware Training

Thanks to our unique compiler technology, seamless back-and-forth compatibility between the PyTorch model and TensorRT engine is now possible, making Quantization-Aware Training incredibly easy.

With the enablement of QAT, performance recovery in lightweight models becomes much easier.