I am very interested in participating in the Intrinsic Challenge. However, I noticed that the recommended setup requires a computer with a GPU and 32 GB of RAM. Unfortunately, I currently do not have access to hardware with these specifications.
I was wondering if there are any cloud-based resources, shared compute environments, or recommended platforms that participants without such hardware could use to run the required tools.
Thank you very much for your time and for organizing this challenge.
I am also interested to know if there are any options for credits to access GPUs cloud credits?
@Atieh_Merikh_nejadas @rosricard,
Unfortunately, we don’t provide compute or credits for the qualification phase, so you’ll need a setup that meets those specs for now. That said, teams that qualify will get access to Flowstate, our cloud robotics platform, for the next stage of the challenge!
If anyone else in the community has tips for budget-friendly cloud setups or shared hardware, please jump in and share here.
1 Like
"I think 32GB is overkill; 16GB is fine. I have a laptop with 32GB of RAM, but running a modified ‘cheatcode’ with the default Gazebo and RViz setup only used 16GB.
As for GPU cloud credits, students can get $300 in free credits for Google Colab. LeRobot has documentation on Google Colab training; I haven’t reached that step for this competition yet, but I plan to use it."
2 Likes
I decided to use an AWS EC2 instance. I was able to get everything set up environment-wise on the free tier (m7i-flex.large), but it can’t really run Gazebo and RViz. Does any one have any recommendations on what instance type would be good to upgrade to? I’m looking to balance cost and performance. Claude recommended g4dn.xlarge, but wondering if anyone has any other recommendations. Thanks!
