New contributor

[Jorge António]

Hello,

First let me present myself. I am currently a data scientist at an US based company. I am familiar with the python tensorflow library, which I used on some projects in reinforcement learning and image classification. Prior to this, I was a postdoc and a PhD beforehand in pure mathematics, in the field of algebraic and differential geometry. You can check some of my academic work in https://sites.google.com/view/jantonio-homepage/home.

It has been a while that I am looking for interesting open source projects, which I can help improve and mature as well as learn as much as possible from experts and more experienced peers.

Recently, I came across with Swift for TensorFlow, and I got extremely enthusiastic about its innovative vision and by being such a young project. Moreover, after spending some time learning the core fundamentals of Swift programming language, I came to deeply appreciate its design, expressivity and efficiency.

For this reason, I would be delighted to join the community and start contributing as much as I can. I already did some research on some of the project code, but it has been difficult to find material that I can contribute to, right away.

Therefore, I am writing this email to not only present myself, but also to kindly ask you if you could give me some hints on starter projects, that I could take on, as well as meaningful learning material.

I will be delighted to have the opportunity to be part of Swift for TensorFlow project.

Thank you in advance for your attention and consideration,

Best wishes and merry christmas,

JA

[Brad Larson]

Welcome! Sorry for the extremely delayed response, we're all slowly getting back to things after the holidays.

Glad to hear that you're excited to dive in. To get started with Swift in general, I tend to recommend pulling up the Swift Programming Language book and starting to write code. The easiest way to do that, without needing to install anything, is to start a blank Swift Colab notebook and interactively explore the language and our machine learning APIs. We have a number of tutorials that you can also open and work with in Colab.

As a next step beyond that, try cloning our Swift models repository locally and running various example models to get a feel for setting up a toolchain (the process for which will be changing, as we talked about this morning) and building models. The models themselves are an attempt to provide best-practices examples of how to use the machine learning APIs in practice, so read through the code for those and see how they're assembled. To look for areas to contribute, we've created and tagged some issues in tensorflow/swift-models with "good first issue" and / or "help wanted" and those are places to start. We haven't done a great job of listing all areas where models could be added to or improved, so there are plenty of opportunities outside that.

When you feel more comfortable with the build process, this is a great entry-level issue in tensorflow/swift-apis that would be a nice improvement and cleanup opportunity. The code already exists for this in tensorflow/swift-models, it just needs to be elevated to tensorflow/swift-apis and to have tests added for correct functionality. Then the code for that in tensorflow/swift-models can be removed in multiple places.

Hopefully, that's enough to get you started, but when you have questions along the way, please don't hesitate to ask in the forums here or offline.