PyTorch 2.0 has been released: https://pytorch.org/blog/pytorch-2.0-release/
Most of the new features are considered to be in a beta or prototype version.

#pytorch #compiler #quantization #tensor #cuda #graviton3

Our lab will offer the two classes High Performance Computing and Efficient Machine Learning in the upcoming semester. New topics include BF16 support in SVE, recent features of PyTorch 2 and inference on mobile devices: https://scalable.uni-jena.de/teaching/2023/03/15/summer-semester.html

#fsujena #hpc #ml #bfloat16 #sve #arm #graviton3 #pytorch2 #quantization #snapdragon

I've been running image processing benchmarks on #AWS #EC2 #ARM instances and it looks like #Graviton3 (c7g) is almost 40% faster than #Graviton2 (c6g), much better than the advertised 25% improvement.

This is almost on a par with AMD EPYC 3rd gen but with lower power consumption, and pricing between instance types appears to reflect this. The use of DDR5 RAM will also be helping.

If you're already running AWS Graviton2 then the ~5% cost upgrade to use Graviton3 seems to be very much worth it.