I've been impressed with the Paperspace team since Dillon and Dan showed me their first demo at YC. I'm pretty sure that I'm not the only one who has been waiting for them to launch linux. These guys really care about product and I think their take on machine learning will open up the tech to an even broader audience. I can see this really taking off the developer community.
Hey PH! I'm one of the founders of Paperspace and I'm happy to answer any questions you might have :) We have been working on this for a while now and I'm very excited to finally open it up to the public.
With Paperspace you can now run a full desktop Linux environment in the cloud without any of the hassle that normally goes along with setting up a new computer. Let us know what you think!
@guidokok great question! Lots of new machine learning and so-called "deep learning" (as well as more traditional machine learning) use GPUs to exponentially increase the performance over traditional CPUs.
Finally, one of our key differentiators is offering a full desktop environment which is generally favored over working in command line. We also offer headless terminals as well for just kicking off a model training session ie when you're not actively working in the VM.
Another significant pain-point in the ML / data science world is just getting everything configured. Drivers, frameworks, dependencies, interfaces etc. can be a *huge* headache. We offer a pre-built template with all of the most common ML software packages and additionally, have a feature called Templates which makes it dead-simple to spin-up new machines with everything setup.
Wow, this is really cool! Was impressed with my first trial of Paperspace (Nice Windows VM for the occasional PowerShell task), and now a suite of cloud-based VMs, ready for ML experimentation & feature building. Really neat product!
Interesting. About your Windows servers - how responsive are remote connections to them from Europe? I'm mostly interested in graphics work over remote desktop.
@edmarferreira Our machines have a 1Gbps fiber connection to the internet and a 40Gbps connection between machines so getting data in/out should be pretty painless . Usually we recommend downloading directly to the machine itself. Let me know if you have any trouble! We are investigating including some common public data sets so stay tuned :)
Bluesmart