Live data from Hacker News

Intel Optimizes Facebook DLRM, 8x Speedup (Deep Learning Recommendation Model)

Post reply on HN

Print view View original

1–3 of 3 posts

Intel Optimizes Facebook DLRM, 8x Speedup (Deep Learning Recommendation Model)

#1

Post by beluis3d » Tue, Nov 09, 2021, 11:10 PM UTC

Intel Optimizes Facebook DLRM, 8x Speedup (Deep Learning Recommendation Model)
sigopt.com

Re: Intel Optimizes Facebook DLRM, 8x Speedup (Deep Learning Recommendation Model)

#2

Post by Rafuino » Wed, Nov 10, 2021, 1:19 AM UTC

So 16x more sockets provided an 8x speedup?

Re: Intel Optimizes Facebook DLRM, 8x Speedup (Deep Learning Recommendation Model)

#3

Post by alpanka » Wed, Nov 10, 2021, 8:08 AM UTC

I thought this was similar to their scikit optimizations, but after a quick look I still can't figure out what is going on here.

Can someone ELI5 this for me?

Post reply on HN