The code for the SAC algorithm was built on top of this curl repo This code can be adapated for pixels, but is currently focused on states
Check standard deviation, it had been changed **Uses weight scheme **
conda
environment (curl
) taken from theconda_env.yml
filetorch
has been downgradedconda install pytorch==1.2.0 torchvision==0.4.0 cudatoolkit=10.0 -c pytorch
, andtensorboard
has been removed
git clone https://github.com/alec-tschantz/mbmf.git
cd mbmf
conda activate curl
python scripts/sac_script.py
sac_script.py
- train a SAC agentmpc_script.py
- train an MPC agenthybrid_script.py
- train a hybrid agenttest_script.py
- test hybrid agent with trained SAC & ensemble model