Learning to Manipulate Deformable Objects without Demonstrations

This is the code that corresponds to the paper

rlpyt Usage

See the original library for more information on the design of the library

Install mujoco-py with mujoco200

Install a custom version of dm_control

Install a custom version of dm_env

Install the original rlpyt environment

All launch scripts are in rlpyt/experiments/scripts/dm_control/qpg/sac/launch

For Cloth (State), see launch_dm_control_sac_state_cloth_point.py

For Cloth (Pixel), see launch_dm_control_sac_pixels_cloth_point.py

For Cloth-Simplified (State), see launch_dm_control_sac_state_cloth_corner.py

For Rope (State), see launch_dm_control_state_rope.py

For Rope (Pixel), see launch_dm_control_pixels_rope.py

Name		Name	Last commit message	Last commit date
Latest commit History 626 Commits
data		data
examples		examples
images		images
rlpyt		rlpyt
scratch		scratch
.gitignore		.gitignore
LICENSE		LICENSE
README.md		README.md
linux_cpu.yml		linux_cpu.yml
linux_cuda10.yml		linux_cuda10.yml
linux_cuda9.yml		linux_cuda9.yml
macos_cpu.yml		macos_cpu.yml
max_q_eval_policy.py		max_q_eval_policy.py
random_policy.py		random_policy.py
sample_random.py		sample_random.py
setup.py		setup.py
simulate_policy.py		simulate_policy.py