Solving a Nightmare with a Headache
Issue: The widgets in Jupyter Notebook were not all displaying.
Non-solution: Try re-do some installs in the virtualenv. While I am at it, let’s also try out pyenv.
New Issue: Windows does not play well with pyenv. You can get pyenv to work, but then virtualenv is no longer set up on your global environment.
Attempted Solution: Try to clone the pyenv-virtualenv repo into a new plugins directory of the pyenv root folder. This did not work, despite sinking about an hour into this whole mess.
New Issue: It’s time to put Linux on my machine to try and escape some of these Windows nightmares. This is going to be a headache, but it seems like the pay off will be worth-while, given that the Fastbook also doesn’t play well with Windows. I have been frustrated with doing Python things on Windows before so I had Linux available already. I will note that it took be a while to remember I already had a Linux partition and in a cascading wave of failure, I forgot my linux password, so was not able to run a needed sudo command. But, I got lucky again (and this is a yikes security-wise) and it turns out you can easily reset the password for Ubuntu.
From here I followed Real Python’s primer on pyenv. Setting this up was not bad and putting some lines in .bashrc virtualenv becomes much nicer to use. It allows you connect a particular environment to a folder. I created a virtualenv named fab (FastAI Book, short environment names are good when you’re used to activating an environment every time you want to work on something) by again following Real Python and now it is automatically activated when I am in the fastbook directory. A quick pip install -r requirements.txt -v got the environment running smoothly with the notebook for Chapter 1, and things that did not work 100% in Windows worked in my Linux boot nicely (at least after a sudo apt install graphviz): * I could run the cells without having to change num_workers to 0. * My GPU was visible to Torch (torch.cuda.get_device_name(0) returned the name of my GPU). * The text example, which did not work at all on Windows, ran. * And, the whole reason I started this endeavor: when I created the FileUploader widget, it appeared in the notebook.