```
np.uint8(0).dtype.kind in np.typecodes['AllInteger'] == False
np.uint8(0).dtype.char in np.typecodes['AllInteger'] == True
```
```
dtype.kind A character code (one of ‘biufcmMOSUV’) identifying the general kind of data.
dtype.char A unique character code for each of the 21 different built-in types.
```
* Add automated dev setup using Gitpod
* Updating README to make instuctions more clear.
* Updating gitpod config to avoid directory change
* Moved Gitpod environment files to external repository
They're now in https://github.com/gitpod-io/definitely-gp/tree/master/gym
* Respect the order of keys in a Dict's observation space when flattening
Prior to this change, the order of the key/values in the observation was used instead of the order in the Dict's observation space. unflatten already uses the order specified by the Dict's observation space.
* add tests for FlattenObservation
* Force explicit Box dtype
Having the `dtype` argument defaut to `np.float32` without warning can generate some unexpected issues.
For example:
```python
import gym
import numpy as np
num = np.array(3.1415926035897932)
gym.spaces.Box(-num, num).contains(num) # returns False
```
* Warn on lowered precision
See also space.Box.contains(), it already has this change
Reproduce Error with following lines:
from gym import spaces
import numpy as np
space_Multidiscrete = spaces.MultiDiscrete(np.ones((10,3)))
sample_test = np.ones((9,3))
space_Multidiscrete.contains(sample_test) #errors: ValueError: operands could not be broadcast together with shapes (9,3) (10,3)
* Delete dict.py
* Update __init__.py
* Update nested_dict_test.py
* update nested_dict_test
* oh huh revert commented out part of the test, now the test fails due to the lack of proper error :(
* remove incorrect argument test - now all combinations of spaces and dtypes are possible to flatten (eveything is reduced to float32, discrete are one-hot encoded)
* make daemon=True an option of async_vector_env
* custom worker in async_vector_env
* add compatibility methods to SyncVectorEnv
* fix name in sync_vector_env
* vectorenv api cleanup
* add docstrings for daemon and worker options in AsyncVectorEnv
* added frame scaling in atari preprocessing
- default value is False
* added frame scaling in atari preprocessing
- default value is False
* corrected logic for selection of observation_space params
* initial commit for scale test
* added test-case for observation scaling in atari
* added message related to framestack in docstring.
* refactored scaled obs test for atari preprocessing
