Method do of PigChaseEnvironment returns inconsistent type

[philipjhj]

the env.do(...) method produces both tuple and numpy.ndarray for the returned state.

Add print(type(obs)) after line 125 in evaluation.py to reproduce the following output:

Episode 6 (0.00)%
<type 'tuple'>
<type 'numpy.ndarray'>
<type 'tuple'>
<type 'numpy.ndarray'>
<type 'tuple'>
<type 'numpy.ndarray'>
<type 'tuple'>

[mfuntowicz]

Hi @philipjhj,

At first sight I suspect your plotting the state for 2 different agents.

For example:

• the PigChaseTopDownStateBuilder returns an numpy.ndarray (the board as a 2D tensor)
• the PigChaseSymbolicStateBuilder returns a tuple, with the board, and the entities information

As I told you before, this is might initial feelings, but I might be wrong, and it might be an issue, we would require more context to be fixed :)

Morgan