Monday May 17th
Week One RLOS
What is RLOS Fest?
- RLOS is short for Reinforcement Learning Open Source Fest. It’s a yearly programme run by Microsoft Research (NY), specifically the Real World Reinforcement Learning Team, in which students contribute to open source projects specifically involving reinforcement learning. These projects are used by researchers and engineers alike, so some of the projects are not clearly defined, which is also interesting. To take part, 12 of us had to submit a proposal and a code sample. I’m very thankful for the opportunity.
- It was not long ago that I was a recently laid-off worker in the technical side of filmmaking from a manufacturing shop, taking courses part-time, hearing that I had been accepted for GSoC, in 2018 (and Mozilla’s Rust Reach). I honestly don’t know how I would have made it through those months if I did not obtain that opportunity. It ended up changing my trajectory, giving me access to opportunities, and basically helped give me a foot in the door to a career in “tech”. The following year, I also reviewed the proposal for another student who was able to participate, and also got an internship in the Bay area shortly thereafter, among other great opportunities. So open source is very near and dear to my heart.
- Anyways, at the end of this opportunity, we have to present our projects to some people at MSR, which I don’t really want to think about right now because it will give me cold feet, but I’m pretty sure that our mentors will prepare us well for this part of the project, too.
This is my first Official week
- This is my first official week of RLOS.
- I’ve been set up with an Azure box, and am currently working on building a distributed pipeline, by first running an Azure script remotely testing several nodes.
- So far, so good. Or not. I started by installing the
azureml-sdkbut realized that it was taking all day to install, and getting stuck, because it couldn’t resolve my various flavours of packages on my laptop. This was fixed by just using a
conda envinstead. Phew. For a hot minute I thought that I wouldn’t be able to get anything done because installations were taking a while, and I thought back to my little 32 bit laptop that was taking all day to build a project in 2018. Alas, this was not the case!
- My contribution will eventually meet with the estimators library that is being worked on by my fellow contributor. Within a month, we will meet in the middle so that our tasks will start to interact, but for now, I’m learning and getting ramped up on Azure. But right now, I’m working on a different repo, and in particular just learning Azure as best I can.
- The mentorship is very good! I’ve really enjoyed meeting our MSR mentors, and they’ve been really patient and very helpful. They also included some really good resources for me to get started. They have said already that they’re willing to help us “any time”, although we meet regularly on Tuesdays and Thursdays in the morning, as my fellow contributor is based in India (she’s awesome and very talented!). We also anticipate that in the earlier part of onboarding, we will be meeting more regularly, most likely daily. My classes have ended, so I’ve been working on research with my advisor and this project solely, with some events randomly sprinkled in between (usually involving R or cryptography related-events).
Taking what I learned from my last Open Source Project Mentorships
- One of the things that was immediately useful was that in GSoC, my mentors told me it was a good idea to write things down. So I’ve already started to make a detailed document for each day of my progress, taking snapshots and noting where I get stuck. This is also useful for writing documentation later on.
- Am I nervous? Of course I am! Knowing I will get stuck, feel super silly, and of course, always feel like I am behind is really humbling. But if I learn something (I already have learned quite a bit!) I’ll be happy. And I’ll meet some nice people along the way, too. I’m also really excited! I really feel humbled by the opportunity to continue being mentored to contribute in Open Source once again. I’ve loved every minute of this opportunity!
I will be writing more throughout the week
- Prepare for progress weekly about my failures, what I learned, insecurities, and all of that. Some of it initially may not be viewable because I’m dealing with a lot of config stuff that is specific to my box, and I don’t want to have the sensitive information for that readily available (for obvious reasons).
- But certainly as I continue to work on this and the project takes shape, I’ll be writing more. It’s also going to be super interesting working on this and an internship at Microsoft, and doing a blockchain mentorship (which I will talk about in terms of the speakers and meetings as I can).
- I also met another lovely research scientist at MSR, who is on the programme committee for ICLR, as it turns out I’ll be a DEI co-chair for 2022. That was awesome, and I’ve left feeling very highly of all the people I’ve met at both Microsoft and MSR so far. But this is just the beginning!
And that’s it.
Written on May 17, 2021