Composability Meets Performance: The Luminal Approach to Modern Neural Networks

Composability Meets Performance: The Luminal Approach to Modern Neural Networks

Speaker
Joe Fioti of General Cognition

Description:
Luminal is a new machine learning framework focused on speed, simplicity and composability. We take a new approach to ML by focusing on static graphs and leaning heavily on compilers.

Luminal boils all operators down to 11 primitive ops capable of supporting all modern neural networks, and uses compilers to transform them into fast code for various supported platforms. This approach allows immense complexity to be added on as plugins, rather than as a core part of the library.

Global knowledge of the compute graph allows for aggressive optimizations not currently possible on eager-first frameworks like PyTorch, while the library simplicity makes edge deployments far more straightforward.

References:
https://www.luminalai.com
https://github.com/jafioti/luminal

Info:
Spots are limited to keep the discussions organized.
Austin Deep Learning Journal Club is group for committed machine learning practitioners and researchers alike. The group typically meets every first Tuesday of each month to discuss research publications. The publications are usually the ones that laid foundation to ML/DL or explore novel promising ideas. Participants are expected to read the publications to be able to contribute to discussion and learn from others. This is also a great opportunity to showcase your implementations to get feedback from other experts.

Sponsors:
Capital Factory (Austin, Texas)
Antler