Linux container for tuning stockfish params with a modified nevergrad4sf. The objective function was updated to use SPRT LLR calculated from pentanomial game results. This appears to reduce the compute needed to converge on good parameters during TBPSA optimization.
Either start a docker container and attach to its shell or reference ./Dockerfile to set up a local environment.
docker compose up --build -d
docker exec -it tuning-fork bash
./run_nevergrad.sh