Gregory Stoner

Love building out apps for Science, Data Science/Deep Learning, and VFX and CG production. Music is a passion via an electric guitar. All opinions are my own

[Ryan Keisler @RyanKeisler]

I'm excited to finally open-source the model from my 2022 paper, “Forecasting Global Weather with Graph Neural Networks”.

Highlights:
• 10-day forecast in <1 min
• Initialize forecasts from ERA5 or IFS analysis
• Scripts for eval, sensitivities, & Hurricane Sandy

[Gregory Stoner @angstroms]

@yochze Hope your doing great

[Gregory Stoner @angstroms]

@andrey_cheptsov The scheduler is headache, but you left out a whole host of issue for building out high performance clusters. Let alone if you want it multi-tenant another rat hole

[Gregory Stoner @angstroms]

@reidatcheson Or a seizure, which happens post a stroke.

[Gregory Stoner @angstroms]

@NikonUSA To San Juan Mountains

[Gregory Stoner @angstroms]

@CUDAHandbook Do not forget the HW Architecture Issue that drive the software arch issues

[Gregory Stoner @angstroms]

@ChapelLanguage Really like this new website https://t.co/9ZLtu5HivQ

[Gregory Stoner @angstroms]

@RajaXg When you think about it and look at the past shift in Materials we all ways hit same wall Power and Thermals, now we have Geometry as well to contend with. I remember looking at what IBM would do to cool ECL logic with water-cooling and ceramic packaging. We at this juncture

[Gregory Stoner @angstroms]

@PatrickMoorhead Total inspiration, my wife also worked with Dr Ube

[Gregory Stoner @angstroms]

@PytorchToAtoms @cdnamz That can happen when I retire..

[Gregory Stoner @angstroms]

@PytorchToAtoms @cdnamz You missed all the conversation I seen about support Gloo.

[Gregory Stoner @angstroms]

@HPC_Guru @AMD @DylanOnChips Well deserved Victor.

[Gregory Stoner @angstroms]

When you find a majestic view - Canyon Lands Green River Overlook

[Gregory Stoner @angstroms]

@j_c_inPDX @exascaleproject @intelhpc Intel OpenMP Compiler with Offload Support currently is one of best I have worked with. It is great team who is working on hardening and driving performance improvements everyday.

[Gregory Stoner @angstroms]

@ChapelLanguage @JoshMilthorpe @simonmcs Looking forward to the talk

[Gregory Stoner @angstroms]

@colindaven @HPC_Guru @nvidia @FT The Google TPU has been around for long time it been running XLA for more then 8 years, NVIDA use XLA compiler in Tensorflow for it Gordon Bell run on Summit. note a lot of this is hidden in DL frameworks

[Gregory Stoner @angstroms]

@RajaXg Bits and Bytes library for Quantization is example of cross over from Consumer to Enterprise, it where early innovation happens on your consumer hardware. Academic need access to low cost hardware.

[Gregory Stoner @angstroms]

@RajaXg Second thing, you can not fracture your hardware platform ISA features, like not supporting Float64 and/or tensor math. You need both the consumer and enterprise hardware to be in market with core software, also not make It enterprise exclusive.