I'm wondering if anyone else is dealing with the same issue I'm facing. We don't have any service quota problems, but we can't seem to spin up any P5 type EC2 instances for our model training. It's quite frustrating since we checked all the regions and there seems to be a major shortage. Are there any tips or workarounds people can suggest? Also, we've noticed that Trainium instances are unavailable too!
1 Answer
It's odd that you can't find any P5 instances. You might want to reach out to your business development manager at AWS since they should have the most updated info. I'm hearing that they're dealing with pretty tight capacity right now, though. Just keep an eye on it!
Yeah, we did reach out, but it sounds like it's a capacity issue rather than a quota thing. They mentioned that they just don’t have the P5 servers available. It's been a hassle because training on G6 and G5 instances is taking forever, like 21 hours for a single run!