What is LightOn Cloud?
LightOn Cloud is a service offered by LightOn, the company bringing AI to you, at the speed of light. LightOn Cloud allows users around the globe to perform Machine Learning experiments using LightOn OPUs (Optical Processing Units) through a dedicated Data Centre infrastructure, hosted in Paris.
LightOn OPUs are the first and only optical AI chips available today through a cloud platform. For a detailed presentation of LightOn’s technology, you may visit LightOn.ai.
Just like any other cloud service ... with a twist
All LightOn Cloud users have access to our proprietary Optical Processing Units alongside powerful hardware and support for the most popular packages for ML and Data Science:
As simple as importing a Python library
Machine Learning at the speed of light
- The OPU is a specialized co-processor that uses photonics to perform Machine Learning tasks at unprecedented speed, power efficiency, and virtually no limitation in data-size.
- More precisely, the OPU allows you to perform a specific kind of matrix-vector multiplication, one of the most important elementary computing blocks in Machine Learning.
- Deep Learning schemes stack such matrix-vector multiplications with non-linearities, and an OPU does just that: it multiplies the input data by a fixed matrix, passes it through an element-wise non-linearity, and outputs the result.
- The elements of this matrix are drawn from a random distribution, at a size that can reach a stunning 1 million rows and 1 million columns. These random matrices are a perfect match to the statistical nature of modern data-driven machine learning.
- While notoriously slow to compute at large sizes on standard silicon-based chips, they become blazingly fast on the OPU.
- Integrated within a whole computing pipeline, in smart combination with CPUs and GPUs, this opens endless possibilities for accelerated and energy-savvy hybrid computations.
Harness the speed of light
Let us walk you through the first steps on the LightOn Cloud and with the LightOn Aurora OPUs.
Access the platform, book a time slot and build a classifier from scratch.
Typical use-cases
- Anomaly detection
- Image/video classification
- Reinforcement Learning
- Natural Language Processing
- Graph Analysis
- Molecular Dynamics
- NN training
Deeper insight
- Through LightOn Cloud you can access a set of sample Jupyter notebooks for every use-case.
- LightOn Cloud users can modify these open source examples to their specific use-cases. And of course, we can help you identify if LightOn OPUs can accelerate your ML task.
- We would love to hear about your project – please let us know on the LightOn community website.
And we are just getting started
LightOn’s team is already prototyping novel OPUs, targeting a wider variety of Machine Learning applications. LightOn Cloud users will be the first to be informed about their release.