First phase of GPU project:

• Investigate best profiling framework for testing GPU speed-up.
• Set up automated profiling on GPU (either remote e.g. Google Collab, Kaggle Kernels) or local (e.g. on NeuroSpin machine).
• Prepare data metaclass for choosing appropriate device for computations (CPU or GPU).
• Prepare first implementation using Forward-Backward algorithm and simplest possible set-up (i.e. no gradient, identity operator, etc.)