thibo73800 / starcraft-reinforcement-learning

Reinforcement learning on starcraft using pysc2.

starcraft-reinforcement-learning

Reinforcement learning on starcraft using pysc2.

First level: Move to Beacon

I used vanissa policy gradient to solve this environments/

Train the model

python beacon.py --training 1

The model will be saved every 100 epochs into the "logger" folder. The model name is "simple_save".

Load the trained model

python beacon.py --model simple_save --replay 1

--replay 1 is used to save a replay of the game every 10 run.

Load and train the agent

python beacon.py --model simple_save --training 1

Watch a replay

python -m pysc2.bin.play --replay <absolute-path>