A rule-based Star Craft II bot. Compatible with pysc2.agents.
cd to the folder and run the command:
pip install -e .
pysc2 (Use Tencent AI Lab fork, required!)
pillow
We recommend pip install each Python package.
Run the agent using the scripts from pysc2.bin.
Example:
python -m pysc2.bin.agent \
--map AbyssalReef \
--feature_screen_size 64 \
--agent tstarbot.agents.zerg_agent.ZergAgent \
--agent_race zerg \
--agent2 Bot \
--agent2_race zerg
See more examples here.
Evaluate the agent (e.g., winning rate) using tstarbot.bin.eval_agent.
Example:
python -m tstarbot.bin.eval_agent \
--max_agent_episodes 5 \
--map AbyssalReef \
--norender \
--agent1 tstarbot.agents.zerg_agent.ZergAgent \
--screen_resolution 64 \
--agent1_race Z \
--agent2 Bot \
--agent2_race Z \
--difficulty 3
See more examples here. In particular, see how a well configured agent plays against difficulty-A (cheat_insane) builtin bot here.
Use pysc2.lib.stopwatch to profile the code.
As an example, see tstarbot/agents/micro_defeat_roaches_agent.py and run the following command:
python -m pysc2.bin.agent \
--map DefeatRoaches \
--feature_screen_size 64 \
--max_episodes 2 \
--agent tstarbot.agents.micro_defeat_roaches_agent.MicroDefeatRoachesAgent \
--agent_race terran \
--agent2 Bot \
--agent2_race zerg \
--profile
See examples here for AI-vs-AI and examples here for Human-vs-AI.
Be consistent with that of pysc2.
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent