Open Source hardware (OpenPower Foundation)

Christopher Sullivan (blog post) writes:

Many times, these ML/DL frameworks, such as TensorFlow can be very difficult to install with GPU capability for individual users and system administrators. Because technologies such as general purpose GPU (GPGPU) require these frameworks to be compiled correctly to enable the hardware not having easy access to pre-compiled versions limits the use of these technologies. Computational researchers are constantly fighting the need to use new tools and enable them versus using the tools to answer scientific questions.

Open-CE is valuable to researchers because it provides the latest and greatest AI package and framework versions pre-integrated in an easy-to-consume and use Conda environment.

This isn’t directly going to impact the radiation oncologists clinical practice, but it is heartening to know that open source hardware for pre-compiling complex packages are available that would speed up scientific progress elsewhere, especially in resource constrained environment.

I will be monitoring the gradual prgress of this iniatative- because as we move towards the AI in radiation physics and complex computational algorithms at scale, it would require a deeper understanding of these open access structures.