2 minute read

Update: Using a virtual machine has not been as smooth as I expected. I’ve ended up with some weird problems after a few days. For example: sometimes my trackpad will stop working properly after closing the VM; I needed to restart the VM today because it didn’t properly connect to the internet for some reason; the scaling is just off - 100% is too small but 125% is too large; the VM doesn’t suspend properly so my laptop drains battery when I close the lid and walk away. WSL2 should role out one week from today, and I’m very excited to try it. I suspect accessing WSL2 from the terminal in VS Code should be easy, and I assume working within a virtual environment (.venv) should be no problem. I just hope file management will be fixed enough

For my Deep Learning class at TUM, everything is written to be used in Jupyter Notebooks in a virtual environment in Unix. Up to this point, I’ve done all my python work in Windows using MiniConda and toying around with WSL (Windows subsystem Linux).

Frustrations with WSL

The most significant problem I have with WSL is its undecipherable file system. I continually ran into problems trying to load files in while trying to call the ubuntu terminal from within Atom. Navigating within WSL was practically impossible. The May 2020 windows update is supposed to address many of these issues, but I felt like I was always trying to force things to work every step of the way. Is it doable, yes certainly. But I decided it wasn’t worth the effort – with the exception of git. Git management through the Ubuntu terminal in WSL worked great.

Dual Boot or Virtual Machine

On my previous laptop, I dual booted Ubuntu 18.10 with Windows and used Ubuntu as my daily driver. I had so many frustrations with Windows on that machine (mainly a trackpad that kept breaking), that I needed a change. Ubuntu was great – my only frustration was the lack of adobe suite. I made nearly all the figures for my thesis using Adobe Illustrator.

This time around, I’ve actually really been enjoying my windows experience. The anti-aliasing is improved (I found it strange how much better text looked on my HP Spectre in Ubuntu than Windows), and the inclusion of Edge Web Apps has really been a game changer. I’m concerned that if I dual boot, I’ll end up never using Windows.

The performance trade-off of running Linux within a VM should actually be pretty minimal considering the improvements intel has continued to make with recent chips. I figure that everything I’d be doing in Linux will be CPU based anyway – I’d run any GPU intensive programming in something like Colab or Kaggle