google-deepmind / google-deepmind/dm_control
Error when trying to deepcopy environment from the manipulation module
- Dominant language
- Python
- Stars
- 4.7k
- Forks
- 764
- PR merge metrics
- No merged PRs in 30d
Description
Hi, I'm using Pytorch Lightning to solve a robotic manipulation task (reach_duplo_features) and this library makes a copy of the hyperparameters passed to its main class. When it tries to copy the environment, I get the following error:
NotImplementedError: Cannot copy from elements with attachments
Environments of the suite module work well, but those on the manipulation module crash.
P.S.: While I'm at it, environments on the manipulation domain don't allow flat observations, like those loaded using suite.load(). It would be great to have that option.
Thanks!
Contributor guide
Research direction
Start by reproducing the deepcopy failure with an environment from the manipulation module, then compare its observation handling with environments loaded through suite.load(). The work is done when manipulation environments can be deep-copied and support the requested flat-observation option without breaking existing environments.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- python, pytorch
- Domain
- machine-learning, robotics
- Issue type
- Bug
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Stale
- Clarity
- Mostly clear
- Newbie friendliness
- 35/100