1 datasets tagged "103"
Policy loss, value error, game length, and MCTS agreement metrics across 103 self-play training iterations.