Embodied Multi-Modal Artificial Intelligence
python train_light.py --env hopper --dataset medium-replay --model dt
python train_light.py --env hopper --dataset medium-replay --model bc
- action mse loss (not work)
python train_light.py --env hopper --dataset medium-replay --model ofa
- action bce loss
sh run_scripts/rl/rl_ofa_try.sh
- GPU
sh run_scripts/pretraining/pretrain_ofa_base.sh
- CPU
sh run_scripts/pretraining/pretrain_ofa_trial.sh
cd metamorph
python tools/train_ppo.py --cfg ./configs/ft.yaml