Deepity: a hand-tuned C++/CUDA predictive coding library, beats jpc (JAX) and ngc-learn on MNIST (CPU benchmarks + source inside) [P]

A developer released Deepity, a from-scratch predictive coding engine written in C++20 with a thin Python layer and hand-tuned CUDA kernels, marking the project's first public release. Deepity reports outperforming jpc (JAX) and ngc-learn on MNIST in CPU benchmarks, with source code included in the release. The author said they had not seen many from-scratch predictive coding implementations backed by real kernel-level optimization work.

I've been building a predictive coding PC engine from scratch in C++20, with a thin Python layer on top, and just cut the first public release. Posting here for feedback and because I haven't seen many from-scratch PC implementations with real kernel-level optimization work behind them. What predi