If you haven’t heard of PyTorch, then you might need to read more blogs and articles in Deep Learning, or really even just your local tech RSS feed because the world of Data Science and Machine Learning has been abuzz about PyTorch since it came out last year in public GA releases. PyTorch is arguably the most flexible deep learning platform available today. It’s open source and backed by some really smart contributors. The community that supports the open source platform is a strong group of engineers from all over the world.
Neural Networks and Machine Learning algorithms have a broad application scope. But because at the end of the day Deep Learning is about computing numbers (over simplified I know) many times its best to just visualize the power of these platforms. So here’s an amazingly popular gif from one of the community contributions overlaying pose in near perfect sync a zebra and a horse using Cycle-GAN and pix2pix.
Theirs now even a TensorBoard-PyTorch plugin available for excellent visualization of key machine learning metrics, model graphs, profiling, etc. Hopefully we can cover the fundamentals of TensorBoard in a subsequent article.
Obviously one of the goals is to get closer to Numpy and many of its scientific convention and functionality differentiators. Since Numpy is nearing 15 years since it was first introduced, perhaps PyTorch will move towards supplanting it. We’ll wait to see if that is even part of the equation or not.
We’re looking for more excellent things from PyTorch. And keeping up on the latest techniques and use cases. We’d be interested to know your use cases for PyTorch or other Deep Learning initiatives you are looking to begin.