3LC Version 1.0 Release Notes#

  1. March 2023

We’re excited to announce the first release of 3LC!

This is an internal release that demonstrates the capabilities of 3LC to improve the training of machine learning models for computer vision. It will be used by our in-house experts to work with clients. This first version will not be made available to external users.

Major Features#

  • Python Notebook API that can be integrated into Jupyter Notebooks to configure metrics collection and usage of the 3LC-dataformat in the Pytorch Deep Learning Framework.

  • 3LC-dataformat which allows for augmenting existing training data so that subsequent training runs operate on a virtual dataset.

  • Web UI which allows for interactivly disabling and relabeling samples.

  • Web UI with sophisticated plotting and filtering which visualizes the behavior of samples during training.

Availability#

This is an internal release only, which will be made exclusively available through our Artifactory instance.

Supported Platforms#

  • Microsoft Windows 10 and 11 (x86)

  • RedHat 7 (x86)

  • Ubuntu 20.04 (x86)

  • Python 3.8

  • CUDA 11.3 (or above) compatible graphics driver

Known Issues#

  • Training data must be converted into the 3LC-dataformat. This process is cumbersome and future versions will work directly with the native data-loaders of deep learning frameworks.

  • The Python Notebook API is expected to change in the next major release.