PaperspaceReviews of paperspace.com

As a computer science student specializing in machine learning, paperspace gradient seems like the ideal service to train and build ML models in the cloud. However, this service somehow manages to drop the ball on nearly every imaginable front. First off, user accounts. You can't remove your credit card data from your account, only replace it. You can also not delete your account from the interface which is a bit frustrating. So is my card data just there in perpetuity? Other reviews already say enough about the random hidden pricing (i.e., how you pay them monthly for the privilege of being able to pay them more money to get a GPU), so I won't go too into detail on that. I will say that they added force-stop limits to notebooks so there's that I guess. The product has really mostly evolved backwards. They added all this random stuff like automatic model deployment and dataset versioning which just doesn't work. I've never, not once, been able to successfully upload a dataset using their dataset system. Not in 2021, not now. Doesn't work via CLI, doesn't work via the web UI. The actually useful cross-notebook persistent storage feature on the other hand is still hidden deep in the docs and has no CLI OR web UI upload. Thanks I guess. And what's up with the CLI anyway? You don't have to try and be the next AWS. Just have a functioning web interface. I know CLIs are cool and all that but the web UI barely works as is. The main gradient product still hardly works perfectly, with notebooks sometimes randomly crashing or browser tabs freezing. And unlike in 2020/2021, now most of the time almost all the free GPUs are occupied anyway, kinda defeating the purpose of having them to begin with. Also I thought it was kinda funny that back in 2020 they spammed me with about 30 e-mails asking me about my experience and if I wanted to have a video chat with the dev team about my experience. These emails abruptly stopped after I said I'd do it for a free month of gradient pro. I guess 8$ is where they draw the line for customer interaction. Fair enough lol. Anyway long rant for me to say, I wish this product just worked. If I could just upload my damn dataset to a persistent storage that works (without having to pay more to put up an S3 container), and have a jupyter notebook with a GPU spun up in a few minutes, I'd rate this thing 7 stars. Instead it's just an exercise in frustration everytime I try to deal with it and I'm left hoping Google Colab eventually evolves past GPU roulette.

Better than google cloud!

Trying to buy their pro plan, my card was declined. Why? They forget to hide error logs, turns out they try to charge me 1000$ directly to Strapi instead of 8$. freaking mad scam.

Paperspace has been a lifesaver for my work. I often have to train multiple machine learning algorithms, and if the supercomputer my company uses goes down, I know I can count on Paperspace to provide an easy-to-use and readily available service to keep my work going.

I’m relatively light user as research assistant tools. Now, I subscribe “Pro”plan; middle class plan.It offers GPU that better than Free plan’s. So far I’m satisfied with this plan and the service. But I have a claim the utility, it is the number of notebooks that can be run same time. Paperspace offers Pro Plan can run 3 notebook simultaneously but I can run only 1 notebook somehow.

Fast onboarding, very stable and reliable, reasonable prices

The service is really good. It has jupyter notebooks with high powered machines that are available at good prices, or even the free machines that come with paying a small monthly fee.

Paperspace is hands down the best cloud GPU platform for ML students like myself. It powered through 3 personal projects of mine with unlimited GPU time limits for just $8 per month. I'll keep using Paperspace unless their pricing changes a ridiculous amount.