The dependencies for this repository is in MAPI.yml
Run DG_train.py to train the DGCRL mdoel
Take Hopper train as an example, run DG_train.py with the parameter as --env hopper -- curve_mode normal --demo_num 50 and --reset_mode a_c
The initial guidance demonstrations are stored in DGCRL/train/rand_demo
The task info is shown in task_info repository.