We have now fresh Ubuntu’s with all the machine learning libraries and GPU support handily configured by Docker. Tensorflow docker images come also with Jupyter notebook, so I played with it around a little bit. It seems like a great tool for collaboration and teaching, but could it work with Git? And in this project?
Combining Jupyter notebooks inside a container and Git turned out to be not so straightforward. There are bound to be more changes in the files due to how Jupyter works, and I probably would have to do some tweaking with the container itself.
PyCharm works like a charm, with Docker too!
PyCharm has been my personal favorite IDE for Python, and it plays nice with Git already. It turned out PyCharm can control Docker containers! I had the official tensorflow with cuda/gpu acceleration running, and PyCharm had no trouble finding it. All I had to do was to install the relevant plugin for PyCharm.
https://www.jetbrains.com/help/pycharm/docker.html#
Hitting the paywall
All this was great and looked promising, but there was a catch: the ‘professional’ -tag on the guide. In order to configure a remote Python interpreter (in the Tensorflow container) I would need the pro version. 30-day trial for the rescue, and personal licenses are not _that_ expensive (8,90€/month).
Breaking the paywall
Wanting badly to test this, I installed the professional version anyway, and while shuffling through the JetBrains webpages, I saw that as a student I might be eligible for a free license!
https://www.jetbrains.com/student/
I registered an account with all the email-hassle and started the pro-version, it asked for my credentials:
Sweet. Now the PyCharm-Docker-Git combination has a chance 🙂
Detecting log anomalies with machine learning 