1. Train CVAE: 
cd s/
python data_from_ori.py  #convert D4RL data to local
python traj_vae_loop_otr.py --env antmaze --seed=6 --dataset=medium-play  #line233: need local dataset; line 302: save cvae model

2. Add scores:
python add_score_vae_otr.py --num_iters=1 --env antmaze --dataset medium-play  # line52: local dataset; line159: saved model; line205: local dataset for relabel; line 206: relabeled dataset saving

3. train IQL:
cd o/
python ant_1.py
