search

openai / baselines

OpenAI Baselines: high-quality implementations of reinforcement learning algorithms
MIT License
15.84k stars 4.88k forks source link

Each mpi worker holds an individual replay buffer in HER? #1211

Open NoListen opened 1 year ago

NoListen commented 1 year ago

For example, if I want to run the script python baselines/her/experiment/train.py --num_cpu 19 --env_name HandManipulateBlock-v0 --n_epochs 200 --replay_strategy future

Will each mpi worker hold an individual replay buffer?

Within each mpi worker, I calculated the sum of the actions in the replay buffer and found them to be different.