Skip to content

Latest commit

 

History

History
112 lines (76 loc) · 3.94 KB

experiment.md

File metadata and controls

112 lines (76 loc) · 3.94 KB
Raw

Experiment Command

Safe Finger

python train.py --task=ShadowHandCatchOver2Underarm_Safe_finger --algo=ppol --headless --cost_lim 25.0 --max_iterations=1000000 --rl_device=cuda:0
python train.py --task=ShadowHandOver_Safe_finger --algo=ppol --headless --cost_lim 25.0 --max_iterations=1000000 --rl_device=cuda:0

Safe Joint

python train.py --task=ShadowHandCatchOver2Underarm_Safe_joint --algo=ppol --headless --cost_lim 25.0 --max_iterations=1000000 --rl_device=cuda:0
python train.py --task=ShadowHandOver_Safe_joint --algo=ppol --headless --cost_lim 25.0 --max_iterations=1000000 --rl_device=cuda:0

Die Rotation

python train.py --task=ShadowHandDieRotation --algo=ppol --cost_lim 1.0 --max_iterations=1000000 --rl_device=cuda:0 --headless  --num_envs=2048

python train.py --task=ShadowHandDieRotation --algo=p3o --cost_lim 1.0 --max_iterations=1000000 --rl_device=cuda:1 --headless  --num_envs=2048

python train.py --task=ShadowHandDieRotation --algo=focops --cost_lim 1.0 --max_iterations=1000000 --rl_device=cuda:2 --headless  --num_envs=2048

python train.py --task=ShadowHandDieRotation --algo=cppo_pid --cost_lim 1.0 --max_iterations=1000000 --rl_device=cuda:3  --headless  --num_envs=2048

Grasp

python train.py --task=ShadowHandGrasp --algo=ppol --cost_lim 25.0 --max_iterations=1000000 --rl_device=cuda:0 --debug --num_envs=1

Hand Over Wall & Wall Down

python train.py --task=ShadowHandOverWall --algo=ppol --cost_lim 3 --max_iterations=1000000 --rl_device=cuda:0 --headless
python train.py --task=ShadowHandOverWallDown --algo=ppol --headless --cost_lim 3 --max_iterations=1000000 --rl_device=cuda:0

Hand Over Wall House & Wall Down House

python train.py --task=ShadowHandOverWallHouse --algo=ppol --cost_lim 0.5 --max_iterations=1000000 --rl_device=cuda:0
python train.py --task=ShadowHandOverWallDownHouse --algo=ppol --cost_lim 0.5 --max_iterations=1000000 --rl_device=cuda:0 --num_envs=1

Hand Over Wall PC

python train.py --task=ShadowHandOverWallPC --algo=ppol --headless --cost_lim 0.5 --max_iterations=1000000 --rl_device=cuda:0

Hand Over Wall PC NEW

python train.py --task=ShadowHandOverWallPCNew --algo=ppol --headless --cost_lim 0.5 --max_iterations=1000000 --rl_device=cuda:7 --headless

Catch Under Arm

python train.py --task=ShadowHandCatchUnderarmWall --algo=ppol --cost_lim 0.5 --max_iterations=1000000 --rl_device=cuda:0 --num_envs=2
python train.py --task=ShadowHandCatchUnderarmWallDown --algo=ppol --cost_lim 0.5 --max_iterations=1000000 --rl_device=cuda:0 --num_envs=2

Pick Bottle

python train.py --task=ShadowHandPickBottle --algo=ppol --headless --cost_lim 13 --max_iterations=1000000 --rl_device=cuda:0

Pick Bottle House

python train.py --task=ShadowHandPickBottleHouse --algo=ppol --cost_lim 0.5 --max_iterations=1000000 --rl_device=cuda:0

Jenga

python train.py --task=ShadowHandJenga --algo=focops --cost_lim 0.5 --max_iterations=1000000 --rl_device=cuda:5 --headless

Jenga(shadow hand and allegro hand)

python train.py --task=ShadowHandAllegroHandJenga --algo=ppol --cost_lim 0.5 --max_iterations=1000000 --rl_device=cuda:0 --num_envs=4
## JengaHouse(shadow hand and allegro hand)

```bash
python train.py --task=ShadowHandAllegroHandJengaHouse --algo=ppol --cost_lim 0.5 --max_iterations=10000 --rl_device=cuda:0 --num_envs=2

JengaHousePC(shadow hand and allegro hand)

python train.py --task=ShadowHandAllegroHandJengaHousePC --algo=ppol --cost_lim 0.5 --max_iterations=10000 --rl_device=cuda:0 --num_envs=2 --headless

Jenga (shadow hand+ allegro with arm)

python train.py --task=ShadowHandAllegroHandArmJenga --algo=ppol --cost_lim 0.5 --max_iterations=1000000 --rl_device=cuda:0 --num_envs=4

Clean

python train.py --task=ShadowHandClean --algo=ppol --headless --cost_lim 100 --max_iterations=1000000 --rl_device=cuda:0