Examples

We provide examples to train the following algorithms:

To run these examples, make sure you have installed hydra:

pip install hydra-core

Scripts can be run from the directory of interest using:

python sac.py

or similar. Hyperparameters can be easily changed by providing the arguments to hydra:

python sac.py collector.frames_per_batch=63

Results

Here we can see some results for the SAC and REDQ algorithm. We average the results over 5 different seeds and plot the standard error.

To reproduce a single run:

python sac/sac.py env.name="HalfCheetah-v4" env.task="" env.library="gym"

python redq/redq.py env.name="HalfCheetah-v4" env.library="gymnasium"

To reproduce a single run:

python sac/sac.py env.name="cheetah" env.task="run" env.library="dm_control"

python redq/redq.py env.name="cheetah" env.task="run" env.library="dm_control"