Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

RLBench returning incorrectly shaped gripper_matrix and wrist_camera_matrix? #124

Closed
ericrosenbrown opened this issue Apr 12, 2021 · 1 comment

Comments

@ericrosenbrown
Copy link

ericrosenbrown commented Apr 12, 2021

I pulled the current version of RLBench from master, and now whenever I tried to get an observation from the environment I get the following error

Traceback (most recent call last):
File "", line 1, in
File "/home/eric/venvs/new_bench/lib/python3.7/site-packages/gym/envs/registration.py", line 145, in make
return registry.make(id, **kwargs)
File "/home/eric/venvs/new_bench/lib/python3.7/site-packages/gym/envs/registration.py", line 90, in make
env = spec.make(kwargs)
File "/home/eric/venvs/new_bench/lib/python3.7/site-packages/gym/envs/registration.py", line 60, in make
env = cls(
_kwargs)
File "/home/eric/new_Github/RLBench/rlbench/gym/rlbench_env.py", line 38, in init
_, obs = self.task.reset()
File "/home/eric/new_Github/RLBench/rlbench/task_environment.py", line 91, in reset
return desc, self._scene.get_observation()
File "/home/eric/new_Github/RLBench/rlbench/backend/scene.py", line 261, in get_observation
if self._obs_config.gripper_matrix else None),
File "<array_function internals>", line 6, in reshape
File "/home/eric/venvs/new_bench/lib/python3.7/site-packages/numpy/core/fromnumeric.py", line 299, in reshape
return _wrapfunc(a, 'reshape', newshape, order=order)
File "/home/eric/venvs/new_bench/lib/python3.7/site-packages/numpy/core/fromnumeric.py", line 58, in _wrapfunc
return bound(*args, **kwds)
ValueError: cannot reshape array of size 16 into shape (3,4)

I went into the scene.py file, and found that tip.get_matrix().shape was returning a [4,4] matrix, not a [3,4] as the one specified in the file. I changed it (And wrist_camera_matrix) to 4,4 instead and now everything appears to work. Is there something I did incorrectly or is this a bug?

@ericrosenbrown ericrosenbrown changed the title RLBench returning incorrectly shaped gripper_matrix? RLBench returning incorrectly shaped gripper_matrix and wrist_camera_matrix? Apr 12, 2021
@stepjam
Copy link
Owner

stepjam commented Apr 12, 2021

Thanks for pointing out. There was a recent update to PyRep. Will fix shortly.

stepjam added a commit that referenced this issue May 31, 2021
* [skip ci] Update matrix reshape following PyRep update. Resolves #124.

* fixed gripper not releasing object

Co-authored-by: stephen <slj12@ic.ac.uk>
stepjam added a commit that referenced this issue May 31, 2021
* Bug fix and visual change to open_door.
Previously the agent could open without needing to turn the handle.

* Visual change to close_door.

* New overhead camera. Moved over-shoulder cameras.

* Hanger tasks visual updates.

* Added UR5. Resolves #100. Resolves #90.

* [ci skip] Update README.

* Can specify what episodes to load when loading demos from disk.

* Bump PyRep version.

* Fix spawning dummys after each task swap. Resolves #113.

* Workflow fix.

* Update unit test assets.

* Workflow fix.

* Workflow fix.

* Fix typo. Resolves #102.

* Update TV tasks.

* Update clock, box, and shoe tasks.

* Update door tasks.

* Update fridge tasks.

* Jenga and books on bookshelf improvements.

* Experimental action mode. Improved task checks.

* Collision check fix when getting demos.

* Bug fixes in ABS_EE_POSE_PLAN_WORLD_FRAME

* Allow noise to be added to stored demos.

* RGB returned as uint8 rather than float. Point cloud obs added. Ability to get camera info.

* Update tests accounting for RGB in range 0-255.

* Added option for collision checking in ABS_EE_POSE_PLAN_WORLD_FRAME.

* Swapped planner to RRTConnect.

* Path action modes are no longer interrupted mid path.

* Phone on base waypoint update.

* Collision checking swapped to use respondables rather than visuals.

* Improve plan action space when already colliding with object.

* Made stack_wine more difficult by requiring bottle to be ungrasped in rack.

* Fixed gripper not releasing object (#128)

* [skip ci] Update matrix reshape following PyRep update. Resolves #124.

* fixed gripper not releasing object

Co-authored-by: stephen <slj12@ic.ac.uk>

* Update complex_task.md

Convex is better for simulation.

* Revert: Plan action mode can break early.
markusgrotz pushed a commit to markusgrotz/RLBench that referenced this issue Jul 12, 2024
markusgrotz pushed a commit to markusgrotz/RLBench that referenced this issue Jul 12, 2024
* Bug fix and visual change to open_door.
Previously the agent could open without needing to turn the handle.

* Visual change to close_door.

* New overhead camera. Moved over-shoulder cameras.

* Hanger tasks visual updates.

* Added UR5. Resolves stepjam#100. Resolves stepjam#90.

* [ci skip] Update README.

* Can specify what episodes to load when loading demos from disk.

* Bump PyRep version.

* Fix spawning dummys after each task swap. Resolves stepjam#113.

* Workflow fix.

* Update unit test assets.

* Workflow fix.

* Workflow fix.

* Fix typo. Resolves stepjam#102.

* Update TV tasks.

* Update clock, box, and shoe tasks.

* Update door tasks.

* Update fridge tasks.

* Jenga and books on bookshelf improvements.

* Experimental action mode. Improved task checks.

* Collision check fix when getting demos.

* Bug fixes in ABS_EE_POSE_PLAN_WORLD_FRAME

* Allow noise to be added to stored demos.

* RGB returned as uint8 rather than float. Point cloud obs added. Ability to get camera info.

* Update tests accounting for RGB in range 0-255.

* Added option for collision checking in ABS_EE_POSE_PLAN_WORLD_FRAME.

* Swapped planner to RRTConnect.

* Path action modes are no longer interrupted mid path.

* Phone on base waypoint update.

* Collision checking swapped to use respondables rather than visuals.

* Improve plan action space when already colliding with object.

* Made stack_wine more difficult by requiring bottle to be ungrasped in rack.

* Fixed gripper not releasing object (stepjam#128)

* [skip ci] Update matrix reshape following PyRep update. Resolves stepjam#124.

* fixed gripper not releasing object

Co-authored-by: stephen <slj12@ic.ac.uk>

* Update complex_task.md

Convex is better for simulation.

* Revert: Plan action mode can break early.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

2 participants