search

openai / baselines

OpenAI Baselines: high-quality implementations of reinforcement learning algorithms
MIT License
15.78k stars 4.88k forks source link

Support for MultiDiscrete and MultiContinuous action spaces in A2C #449

Open hari-sikchi opened 6 years ago

hari-sikchi commented 6 years ago

Hi, Support for MultiDiscrete and MultiContinuous action spaces is not built for A2C implementation. Does anyone know which algorithm can help me with large discrete action spaces?

Thanks!

kirk86 commented 6 years ago

I believe a3c although you could also use ddpg which is for continuous space with a couple of modifications.