So I am a startup enthusiast who worked on a variety of different projects, researched BOINC's mechanisms and contributed to a series of things such as Lonero OS, BitToken, EtherStone, etc. I wanted to learn more about PyTorch's decentralized tutorial and how it can be related to distributed computing protocols.

What it does

So I was running into trouble trying to install the decentralized tutorial samples for PyTorch as seen in my GIT repository, but basically I was trying to make it compatible with grid computing and not require cuda support for running multiple instances of PyTorch untop of each other, and do things such as speed up data parsing.

How I built it

I started by installing PyTorch which was easy after using the Anaconda installer and installing from the repo instead of through conda. After that, I tried running the code in the demos and had problems starting from there, so I just started planning out what updates I can do for future steps.

Challenges I ran into

I had only 30 minutes saved up for this Hackathon due to a recent increase in workload but still wanted to attempt to submit something.

Accomplishments that I'm proud of

Submitting something and having an action plan for what I am going to do later

What I learned

Got introduced to the basics of PyTorch

What's next for DecentTorch

I think I am going to update the sample code, and then maybe make a Docker package or something for what I want to do. Then I will allow for people to run distributed python applications quicker with it untop of some decentralized, distributed and grid computing protocols.

Built With

  • pytoch
Share this project: