Official codebase of the paper "Model-Based Offline Planning with Trajectory Pruning (MOPP)".
See requirements.txt
.
- Download the benchmark dataset from D4RL in
data/d4rl
. - Run
run_planning.sh
to run the D4RL benchmark tests. - The trained models are placed in
learn_model/
. - Check results of MOPP in
learn_log/
.