Compile to a single native binary with AVX! This #Free_Pascal library runs GPT-style transformers, Mixture-of-Experts, RWKV, xLSTM, spiking neurons, and normalizing flows, natively, with no Python nor CUDA lock-in.
https://doi.org/10.5281/zenodo.5810077
https://github.com/joaopauloschuler/neural-api