FlashML — Machine Learning in Seconds. Dead Simple. PyTorch Flexible.

Hey Hunters, happy new year! We realized that low-level machine learning frameworks (i.e. TensorFlow and PyTorch) are great for their flexibility, but they are complicated to learn and require a lot of time and repetitive code. On the other hand, AutoML tools (i.e. Google AutoML, CreateML) are extremely simple to use and abstract away most of the repetitive coding, but they have very little flexibility and lead to problems with leaky abstractions. When we suggested a middle ground to the PH community, a lot of people were interested. So we combined the best of both worlds in FlashML, a no-code machine learning platform to quickly create and train PyTorch code in three easy steps: 1. Drag-and-drop layers, hyper-parameters, and a dataset in the FlashML platform. 2. Download the equivalent PyTorch code with the press of a button. 3. Train your model locally with a single terminal command. It can be used educationally when you are first learning ML and it can also be used by experienced ML engineers to quickly start-up projects and avoid repetitive code. Anyone can now go from zero to training their PyTorch model on CIFAR10 in less than 3 minutes!. We are constantly adding more features, layers, datasets, and functionality. We also plan to open-source the whole platform in the near future so that everyone can help us expand it. Please let us know your feedback and thoughts so we can keep improving this together. Cheers!