Hacker News Reader

>> My day job is primarily ML as well, so I might just go for the 3090. 24 GB of memory is a game changer for what I can do locally. I really just wish Nvidia would get its shit together with Linux drivers.

What is your ML dev environment like?

View on HN

We have a small team of 7, so after a lot of bureaucratic navigation I got us custom built desktops for local experimentation with 2080 Ti's and for the rest we use your typical cloud providers. Personally, I'm partial to GCP, because of the TPU support and BigQuery, but AWS is fine too.

Interesting, thanks.

Reply on news.ycombinator.com