Installation

Follow HW.md and pip install moviepy

Usage

• Training

bash scripts/train_policy.sh {diff_pusht, reg_pusht, diff_calvin}

• Evaluation follows HW.md

Push-T Benchmark

Success rates

	train	test
diffusion	0.996	0.822
regression	0.846	0.823

Evaluation rollout videos

diffusion

100000	100001	100002	100003

regression

100000	100001	100002	100003

Sampled trajectories (seed10001.mp4)

diffusion

seed=8	seed=16

regression (no difference)

seed=8	seed=16

CALVIN Benchmark

Success rates

1/5 : 55.0% | 2/5 : 0.0% | 3/5 : 0.0% | 4/5 : 0.0% | 5/5 : 0.0% ||

Evaluation rollout videos

open succeed	open failed	close succeed 1	close succeed 2

Youtube DEMO

https://youtu.be/dq_oauqjtqg