This repo is based on the original GPS. We modified the repo to perform benchmarking as part of the Model Based Reinforcement Learning Benchmarking Library (MBBL). Please refer to the project page for more information.
For full documentation, see rll.berkeley.edu/gps. First install this library as instructed in rll.berkeley.edu/gps. Then please go to MBBL to install the mbbl package for the environments.
To disable the rendering and run the experiments on headless servers. first run
xvfb-run -s "-screen 0 1400x900x24" bash
Please refer to exp_script/gym_search_2.sh.
An example to run HalfCheetah looks like this:
env_name=gym_cheetah
# python python/gps/gps_main.py gym_cheetah_mdgps_example 2>&1 | tee output.log
for batch_size in 5000; do
for rand_seed in 1234 2345 2314 1234 1235; do
# generate the config files
exp_name=example_${env_name}_batch_${batch_size}_seed_${rand_seed}
cp -r experiments/gym_mdgps_example experiments/${exp_name}_mdgps
# modify the config files
sed -i "s/ENV_NAME/${env_name}/g" experiments/${exp_name}_mdgps/hyperparams.py
sed -i "s/RAND_SEED/${rand_seed}/g" experiments/${exp_name}_mdgps/hyperparams.py
sed -i "s/TIMESTEPS_PER_BATCH/${batch_size}/g" experiments/${exp_name}_mdgps/hyperparams.py
# run the experiments
python python/gps/gps_main.py ${exp_name}_mdgps 2>&1 | tee ./log/${exp_name}_mdgps.log
done
done
The configuration can be changed by modifying the template config file experiments/gym_mdgps_example.