Some LLM post-training experiments may be open-sourced

A developer says they’ve been having fun trying a wide range of LLM post-training experiments using KerasHub, JAX and TPUs. They might open-source some of that work.