Run a 1T parameter model on a 32gb Mac by streaming tensors from NVMe
#HackerNews #Run #a #1T #parameter #model #on #a #32gb #Mac #by #streaming #tensors #from #NVMe #https://github.com/t8/hypura #MachineLearning #Tensors #NVMe #Mac #Optimization
Run a 1T parameter model on a 32gb Mac by streaming tensors from NVMe
#HackerNews #Run #a #1T #parameter #model #on #a #32gb #Mac #by #streaming #tensors #from #NVMe #https://github.com/t8/hypura #MachineLearning #Tensors #NVMe #Mac #Optimization
A sharded DuckDB on 63 nodes runs 1T row aggregation challenge in 5 sec
https://gizmodata.com/blog/gizmoedge-one-trillion-row-challenge
#HackerNews #DuckDB #Sharding #Aggregation #Challenge #1T #Rows #BigData #Performance
Getting $#1t DONE today.
Also I am loving my new arc of being able to dispense information and support.
This is what I was [made/built] for.
@LexTenebris I'm in a bit of a bind right now, but before all this $#1t I did comms with volunteer SAR and we used to do all kinds of stuff by/for ourselves.
Ham radio op, and basically all my friends here are from this background. Knew of #Freifunk since forever and wanted to do the same when my small town was still cut from the rest of the globe. And probably will experiment with #LibreMesh soon again.