oat

chore: update docs and python version (#47)

feat: supporting TP for vLLM and multi-node training (#46)

feat: add math rl examples and data (#45)

allow actors dump to save_path

fix broken examples (#44)