As I dive into my machine learning research, I have to answer the classic question, "what deep learning framework should I use?" Below is how I categorized the two options before reading about the current status of each framework.

TensorFlow: the default choice, largely used in commercial / production environments, mature options for edge deployments

PyTorch: the new kid on the block, largely used in the academic / research community, less support for edge deployments

After spending some time with good ol' Google, I came to pretty much the same conclusion. So which one will I be using for my research? PyTorch. Why? Mainly because I'm doing academic research, a lot of the open-source toolkits I'm interested in are PyTorch based, I want to learn something new, and I'm curious about ExecuTorch - PyTorch's new solution for running inference on edge devices.

The pre-release of ExecuTorch only came out in October of last year and it's not yet recommended for production. That's alright though, I'm not going to production.