Owlite

Pytorch 2.2 support

OwLite is now compatible with the cutting-edge PyTorch version 2.2, empowering your projects with the latest advancements in deep learning technology.

Web GUI updates

- Layer Selection Enhancements
- Enhanced Layer Search and Selection
- Latency Analysis and Optimization

Please check out GUI updates [here](https://squeezebits.gitbook.io/owlite/updates/1.2.1-update-14th-mar-2024)

Important Notice
Please ensure that you update your OwLite to the latest version to enjoy optimal performance and features.

How to Update to the Latest Version:

bash
pip install git+https://github.com/SqueezeBits/owlite --upgrade --extra-index-url https://pypi.ngc.nvidia.com


New Calibration Option
We have added a new calibration option: **entropy**

OwLite now offers the entropy calibrater, providing users with enhanced flexibility and control over the calibration process directly on https://owlite.ai/.

New Naming Rules:
We've introduced new naming rules for Projects, Baselines, and Experiments. Now, you can utilize letters, numbers, and special characters: ( ) - _ : * &.

Feature Improvements:
Enhanced fake quantizer learning ONNX export speed.

Device Auto Connect:
For users on the free plan, we've introduced device auto-connect functionality. If no devices are found, the system will automatically connect to a free device, ensuring a seamless experience for our free plan users.