HN Explorer

5 comments

[ 5.3 ms ] story [ 21.1 ms ] thread
smoldesu
(by paying us to run it for you)
naderkhalil
The GPU described in the blog post is $1.45/hr. You can also deploy it in your AWS account
smoldesu
Most people can also deploy it locally for free just as fast. If you've already got Docker installed, it probably won't even take 4 minutes.
capableweb
The model weights around 4GB + various dependencies end up at least 6GB. So in order for people to get it running in 4 minutes, you need to have at least a 400Mbit/s connection (would download in around 3 minutes, leaving one minute for entering the commands for installing + running).

It is very easy to get started, average technical person with OK connection could get it up and running in 15 minutes with something like https://github.com/invoke-ai/InvokeAI probably

alecfong
This is a neat fork! Running benchmarks on the machines seems to show ~1426.01 Mbit/s download