opendilab/DI-star

Health v2 · maintenance only

Flags: no_releases

Adoption not part of the score

1393 stars · 127 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

DI-star is a large-scale distributed training platform for building StarCraft II game AI, including supervised and reinforcement learning training code and pre-trained grand-master-level Zerg agents. It lets researchers train agents via self-play and play against the released models.

Use cases

• train a StarCraft II AI agent with reinforcement learning
• run large-scale self-play distributed training for game AI
• play against a pre-trained grand-master StarCraft II bot
• reproduce AlphaStar-style league training on a budget
• do supervised learning from StarCraft II replays
• research multi-agent reinforcement learning in a real-time strategy game

When to choose

• you want to train or evaluate deep RL agents specifically for StarCraft II
• you need a reference implementation of AlphaStar-style league/self-play training
• you want to play against a strong pre-trained SC2 agent

When to avoid

• you need a general-purpose RL library not tied to StarCraft II
• you want agents for other StarCraft II races than Zerg vs Zerg
• you lack a GPU or a StarCraft II installation
• you need actively developed features or newer game patch support

Facets

framework · maturity maintenance

reinforcement-learning deep-learning machine-learning agent-framework llm-training reinforcement-learning artificial-intelligence deep-learning gaming-tools windows python starcraft2 self-play distributed-training game-ai league-training pytorch game-development linux macos gpu

Member repositories

For agents

markdown · JSON · MCP: product_card(name="opendilab/DI-star")