* pyglet GL context for SimpleImageViewer case
* typo fxd 'viewer_type' argument in get_window()
* use **kwargs in get_window()
* remove viewer_type from Viewer class
Added description for the mountain_car environment. As a beginner this description would have saved a lot of time for me for understanding the environment.
fixed close() method for Viewer to be consistent with SimpleImageViewer(), otherwise I get the same error:
```
Exception ignored in: <bound method Viewer.__del__ of <gym.envs.classic_control.rendering.Viewer object at 0x112ea2f60>>
Traceback (most recent call last):
File "/Users/hadavid/anaconda/lib/python3.6/site-packages/gym/envs/classic_control/rendering.py", line 162, in __del__
File "/Users/hadavid/anaconda/lib/python3.6/site-packages/gym/envs/classic_control/rendering.py", line 81, in close
File "/Users/hadavid/anaconda/lib/python3.6/site-packages/pyglet/window/cocoa/__init__.py", line 277, in close
File "/Users/hadavid/anaconda/lib/python3.6/site-packages/pyglet/window/__init__.py", line 820, in close
ImportError: sys.meta_path is None, Python is likely shutting down
```
The current implementation of Box doesn't allow passing an array as low or high while also providing a shape.
The current implementation of Box doesn't allow passing an array as low or high and a constant as the other bound.
Co-authored-by: pzhokhov <peterz@openai.com>
* Add tests for gym.spaces.utils.
* Add docstrings to gym.spaces.utils.
* Remove some trailing whitespace.
* Add gym.spaces.utils.flatten_space.
The new function also is reexported as gym.spaces.flatten_space. It
improves the determination of observation_space in
gym.wrappers.FlattenObservation.
* Produce OrderedDict instead of dict in gym.spaces.unflatten().
`gym.spaces.Dict` is very particular about producing its samples as
`OrderedDict` in order preserve the order of its items. Hence,
`unflatten()` should reproduce this behavior.
* In test_utils.compare_nested, also verify order of OrderedDict items.
* Add examples to flatten_space() docstring.
* Document ``flatten(space, space.sample()) in flatten_space(space)``.
Co-authored-by: Nico Madysa <nico.madysa@tu-dresden.de>
* add type of argument
* fix typos
* split lines for formatting
* reformat string, add ellipsis, remove r string
* make docstring stylistically consistent
* make docstrings a little more elaboratet
* reduce by 1 space
* make line wrap 120
* remove unnecessary line
* add returns to docstring
* add docstring, make code more pep8 and delete some unused print functions
* more pep8
* file docstring instead of comments
* delete unused variables, add file docstring and add some pep8 spring cleaning
* add file docstring, fix typos and add some pep8 correections
Co-authored-by: Dan <daniel.timbrell@ing.com>
In the current implementation, *reward* sometimes returns integer values, and other times returns float values.
The example below corresponds to three random episodes and showcases how reward contains -1.0 and -1 values.
[((13, 1, False), 0, -1.0)]
[((18, 10, False), 1, -1)]
[((5, 10, False), 1, 0), ((8, 10, False), 1, 0), ((18, 10, False), 1, 0), ((20, 10, False), 0, -1.0)]
This change tries to fix that, and guarantee that a float value is always returned.
Hope it helps!
Miguel
I have developed a new open source simulator for poly-articulated system called Jiminy. It relies on the powerful rigid body dynamics framework Pinocchio for the physics evaluation, and meshcat for real-time visualization.
I think the project is now mature enough for everyone to use it, despite being partially undocumented. The library has been designed for fast prototyping of control algorithms and machine learning. It is easy to use and several examples of classical control theory and machine learning are available (cartpole, double and simple pendulums).
* Video recorder: distinguish input and output fps
Common video players such as VLC, or Quicktime struggle to play videos
with low framerate, which is problematic for environments where the
policy
frequency is low (e.g. 1 Hz).
See: https://superuser.com/a/601916
This commit sets:
- The input image sequence framerate to the metadata
'video.frames_per_second'
- The output video framerate to at least 30 fps
* Use a configurable output video framerate
This output video framerate is different from the image sequence
framerate:
- when the environment fps is low, the output video fps can be increased
so that players like VLC or Quicktime handle them more gracefully (they
struggle at low fps)
- when the environment fps is very high, the output video can still be
recorded at a lower fps to better compression.
It is configured through the 'video.output_frames_per_second' metadata
and defaults to 'video.frames_per_second' (which itself defaults to 30)
* drop python 2 support
* remove python 2.7 logic from Dockerfile
* Revert "remove python 2.7 logic from Dockerfile"
This reverts commit 43ceabfa8d70ca083661ab3d0d03f2a1adbef5ff.
* remove python 2.7 logic from Dockerfile
* use python 3.7.3 (3.7.6 segfaults with mujoco for some reason?
* include cmake into system level packages
* add swig to system packages
* python3.8 build still segfaults on mujoco tests, disabling for now
* add python 3.8 and 3.9 to the build
* python 3.9.1 -> 3.9.0
* oops i did not realize python 3.9 is still in alpha
Added italics to some important words in sentence to enhance readability(I missed them but those words shouldn't be missed).
Added comma wherever necessary.
* Fix BipedalWalker - now returns closest lidar trace instead of furthest
* additional fix for raycast failing to get nearest point
* bump bipedal walker versions from v2 to v3
