One of the key drivers of machine learning’s growth has been improvements in computing power. Graphics processing units (GPUs), originally designed for rendering complex graphics, are suitable for quickly parallelizing the kinds of matrix operations common in machine learning code, making it computationally feasible to create larger and larger models trained on more and more data. With the introduction of tensor processing units (TPUs) designed for machine learning, machines are only getting more capable of handling larger machine learning tasks.

However, such computing resources are expensive and in high demand, making it difficult for personal ML enthusiasts like high schoolers to train their own models and take on personal projects. Here we list several places where you can find computing resources for your ML projects at no cost.

Google Colab

Google’s Colaboratory provides an online platform where you can write Python code in your browser in an interactive, shareable environment. Additionally, Google provides hardware acceleration for free, allowing you to speedup your code with both GPUs and TPUs. However, due to resource constraints, they do set dynamic usage limits and prioritize allocating computational resources based on previous usage history (users that’ve used fewer resources are prioritized).

Kaggle

Kaggle may be most well known as a website that hosts datasets and various data science competitions, but it also offers hardware acceleration for free in its notebooks. Users have two 30 hour quotas for running accelerated notebooks, one for GPU acceleration and one for TPU acceleration.

Need More?

Are you a high schooler with more ML experience working on a personal project that would like additional computational resources? Are the above resources on this list insufficient for your project? If so, please contact us.