Deep Ivy - A mission to unify all ML frameworks
- 13:00 15th February 2022 ( week 5, Hilary Term 2022 )
The number of open-source ML projects, libraries and codebases have exploded in recent years, with all of these written in many different incompatible ML frameworks. Wouldn’t it be nice if you could take a DeepMind author's JAX code and then run it straight in your PyTorch pipeline without any issue? Ivy makes this possible. Ivy is a thin templated and purely functional framework, which wraps the functional APIs of existing ML frameworks to provide consistent call signatures and syntax. Higher level functions, layers and libraries can then be built on top of Ivy’s functional API, for users of all frameworks. With the use of framework-specific frontends (currently in development), Ivy will also enable automatic conversion between any two different frameworks. No need to “back a horse” with your framework selection, Ivy enables you to back all horses simultaneously, and mix and match libraries for all frameworks in a single project!
We are in talks with developers from Facebook, Nvidia,The Allen Institute for AI, and other top software companies who would like to use Ivy in their popular open-source projects, to instantly support all frameworks. A few examples are: Kornia, PyG, BoTorch, and Tonic.
Please use the following teams link to access the talk at 1pm on Tuesday 15th February
Microsoft Teams meeting
Join on your computer or mobile app
Speaker bio
Daniel Lenton is a PhD student at Imperial College London, working in Robotics and 3D Vision under the supervision of Prof. Andrew Davison at the Dyson Robotics Lab. He currently serves as a reviewer for NeurIPS, CVPR, IROS, ICRA and others. He is also the creator of Ivy, which has just raised a round of pre-seed VC funding to hire a team of developers. Ivy is on a mission to unify all Machine Learning (ML) frameworks. Daniel has also interned at Facebook Reality Labs and Amazon Prime Air, working on real-time 3D vision and robotic systems at the intersection of Deep Learning. Prior to his PhD, Daniel completed his MEng Mechanical Engineering also at Imperial College, attaining 1st class honors and deans list. More information can be found at danlenton.comShare this:
