Join us for an AMA with the engineering team behind cuGraph
When: April 12th, 9am (PDT)
Where: Right here in this forum directory
How do I participate?
- Login to your forum account - and follow or “Watch” this post - we will be sure to edit and update this post, and also post reminders as we get close.
- You are welcome to post some questions for the team even in advance of the AMA.
- On the actual day join us online in this category, and you’ll see questions being answered almost real time, for about an hour. Questions we don’t get to in that hour will still be answered - but later.
What is cuGraph?
cuGraph is a robust, full feature, suite of graph analytics libraries. The core cuGraph library includes about three dozen algorithms, with most support scaling to massive GPU clusters. The cuGraph PageRank algorithm, for example has been test on a 4.4 trillion edge graph spread over 2,048 GPU. Even at that scale, a single PageRank iteration only took 1.5 seconds. Recently cuGraph has added supports for GNN with accelerated aggregators, accelerated models, and extensions to both DGL and PyG.