jwjeonn/

smacv2-offline

Offline reinforcement-learning datasets collected on the SMACv2 (StarCraft Multi-Agent Challenge v2) benchmark. Trajectories were generated by QMIX policies and stored as episode batches in HDF5 format, suitable for offline MARL research.

Three races × three team sizes:

Race Team sizes

Protoss 3v3, 5v5, 10v10

Terran 3v3, 5v5, 10v10

Zerg 3v3, 5v5, 10v10

For each scenario two dataset qualities are… See the full description on the dataset page: https://huggingface.co/datasets/jwjeonn/smacv2-offline.