vikbehal@gmail.com is my email ID. I signed up last week and total GPU utilization is ~2 Hours. I’m still asked to add billing to continue the services.
Yes me too, 6 hours, 20 minutes, 47 seconds = > CPU, 2 hours, 17 minutes, 58 seconds => GPU, 411.14 MB => Storage (before the training was 41MB), and for 10 days I used 5GB of storage, and even if it would have been 30 days does not justify the $1.76 you charged me. yet the service is still very inexpensive, and is very easy to use, lack of improved clarity regarding storage, is not very clear. and I fully understand that your profit for added value is in storage, but it must be very clear, because the documentation is really very poor. in general the service is good and for rapid development is very good.
@anurag - Good luck with Crestle. I only used it sparingly last week, but so far looked good. It hits a really nice sweet spot of minimal setup and very reasonable prices. Once I run out the credits, I will surely come back to explore more.
One suggestion - Interactive Jupyter is good, but many times we want to run in the background as jobs. If you provide a way to Schedule and Run Deep Learning Tasks as Jobs that we could submit to Crestle and collect the results later, it might be a useful platform to run several experiments.
For some reason, if you remove an AMI from AWS and have a limit of one AMI, you can’t create another one right away, AWS persists in that the image exists (it was my fault I wasn’t patient in waiting for the official image), but Crestle solves you quickly.
You can save the notebook and other relevant files that you want to preserve, to a GitHub repo. And then pull/push later once you have the setup issues resolved.