published on December 9th, 2019
I had my last day at my old job this week. It took me a little to wind down and I didn't feel like jumping full force into the learning project right away - that's why this post is comming out a day late. During the last few days at my old job I didn't do any learning and I also did only limited learning last Thursday and Friday.
Not my usual self, I decided to give me an extra day to complete my goals for the last week and jumped in and worked on getting the simple neural network up and running today. In the process, I discovered the fantastic new colab service from google which enabled me to run a GPU-powered notebook much quicker and more easily than when fidling with the setup in paperspace or via a cloud provider. One of the drawbacks however was that I'm running the fast.ai library version 1.0 there and not 0.7 as in the lectures. No matter, I thought, why go with the past, I'll figure out how this new library works. That took longer than expected, but at least in the process I discovered that (what was shown of) the 0.7 version of the fast.ai library was a very thin wrapper around pytorch and that these parts were easily worked around. I think this was time well spent, because while the "batteries included" style of fast.ai's library is great to get best practices packaged, I do like to look a little bit under the surface.
I also worked on reviewing the fast.ai courses so far and produced some summary charts and notes. I'm not quite done with this yet. I also didn't complete the review of the course by Andrew Ng. To be honest I am a bit frustrated with the pace of my progress. I really think that I should be making progress much faster. My time is limited, I need to make the most of it and I don't think I've been doing this too well ;(
Looking back at my goals for week 7, here are the results:
This is one of the few weeks which I can truely dedicate full time to learning data science and now I have four days left in it. These are my goals for this week: