MachineLabs is best described as the CodePen for Machine Learning. We aim to lower the barrier of entry to Machine Learning and embrace knowledge sharing.
Machine Learning is a field that seems rather exclusive to people with academic background (professors, Phd’s, …). It’s often not trivial to follow even beginner guides since the field comes with a heavy focus on mathematical theories. On top of that, setting things up to just get started is also very challenging, especially if one isn’t a data scientist who is familiar with all required tools.
We try to lower the barrier of entry as well as making the tooling scape more pleasant, so more people are able to start learning in this field, by implementing the following features:
Online Code Editor - Everyone should be able to simply open up a new browser tab and get started. No setup hassle, no prerequisites required. It’s possible to literally copy and paste Machine Learning code and execute it right in the browser.
Ready to use environments for main ML frameworks - Apart from Tensorflow and Theano, there’s other major libraries like PyTorch or Caffe. MachineLabs provides execution environments for all of these including common libraries needed to perform Machine Learning tasks. Setting up and using a particular environment is as simple as changing one line of configuration code.
Access to blazingly fast GPU hardware - Training a neural net can take hours, or even several days to finish. That’s why we want to execute our code on super fast GPU optimized machines. MachineLabs offers GPU accelerated machines as part of an advanced pricing plan.
Keep track of work - All executions performed stay available for us so we can always keep track of our experiments. Each execution offers information about the duration of training, environment and execution status.
Expose generated assets via API - Outputs and generated assets can be stored on MachineLabs, which are then exposed via a REST API, enabling us to request trained models for browser based front-end apps.
Our main priority is to make the Machine Learning field accessible to everyone. This not only means lowering the barrier, but also making it easy to share and explore experiments. Coming from the web development landscape, tools like JSFiddle, Plunker and CodePen are the status quo. We want to enable the same for the Machine Learning field by providing features like:
- Sharing public labs - Labs are public by default and can be shared with anyone, anywhere anytime. Whether it’s a freshly trained model or a training that is running in this very moment, so it can be observed by friends.
- Forking labs - Obviously, it should be easy for friends to get started as well. We can take existing labs and simply fork them, making the forked lab our own. The new lab can now be changed and tweaked to a our needs.
- Embedding Labs - MachineLabs provides an embedded editor which can be used to embed labs in websites and blogs. This enables a great experience for readers as they can see the code and their execution from right within the article they are reading.