FloydHub

…a platform for data scientists building deep learning and AI applications.

I was looking for another option to test code using GPU. I used machine from Hetzner earlier, but then I realized that I don’t spend lot of time working / learning on AI stuff so it was a bit of wasting resources. The machine from Hetzner also quite powerful, it would be beneficial for larger or more serious project but not suitable to my situation. So I am looking for a lighter GPU powered solution.

FloydHub’s documentation is good enough to let me start using their service immediately. Go to Quick Start for standard python environment, or go to Jupyter Notebook to get jupyter’s environment. I am more interested on Jupyter’s environment at the time so I cloned previous project’s submission for Udacity’s CarND about traffic’s sign classifier, and set it up to use FloydHub.

FloydHub provide web interface to interact with the project. Once the code run on FloydHub, we can manage easily from the provided web interface. Very comfortable.

And since I run Jupyter environment, the notebook will be available to use from the URL given by FloydHub.

That’s it.

If you need a GPU machine to learn ML / DL then you can utilize FloydHub services to begin with, and once you’re ready to run the code on production system, you can move to another infrastructure such as Bare Metal Machine, Hetzner, AWS, or Google Cloud.

Note: It seems I finished the free credit for FloydHub and received notification to add payment information on https://www.floydhub.com/plans, but when I write this article the link is “404”. Hmm…🤔🤔🤔