Make it clear when the observation doesn't match the defined space #460
Description
I've been implementing my own env wrapper and run into the problem with this traceback:
Traceback (most recent call last):
File "/Users/fedor/anaconda3/envs/vpg-env/lib/python3.6/multiprocessing/process.py", line 258, in _bootstrap
self.run()
File "/Users/fedor/anaconda3/envs/vpg-env/lib/python3.6/multiprocessing/process.py", line 93, in run
self._target(*self._args, **self._kwargs)
File "/Users/fedor/anaconda3/envs/vpg-env/lib/python3.6/site-packages/catalyst/rl/scripts/run_samplers.py", line 154, in run_sampler
sampler.run()
File "/Users/fedor/anaconda3/envs/vpg-env/lib/python3.6/site-packages/catalyst/rl/core/sampler.py", line 317, in run
self._start_sample_loop()
File "/Users/fedor/anaconda3/envs/vpg-env/lib/python3.6/site-packages/catalyst/rl/core/sampler.py", line 313, in _start_sample_loop
self._run_sample_loop()
File "/Users/fedor/anaconda3/envs/vpg-env/lib/python3.6/site-packages/catalyst/rl/core/sampler.py", line 291, in _run_sample_loop
trajectory, trajectory_info = self._run_trajectory_loop()
File "/Users/fedor/anaconda3/envs/vpg-env/lib/python3.6/site-packages/torch/autograd/grad_mode.py", line 49, in decorate_no_grad
return func(*args, **kwargs)
File "/Users/fedor/anaconda3/envs/vpg-env/lib/python3.6/site-packages/catalyst/rl/core/sampler.py", line 264, in _run_trajectory_loop
self.trajectory_sampler.reset(exploration_strategy)
File "/Users/fedor/anaconda3/envs/vpg-env/lib/python3.6/site-packages/torch/autograd/grad_mode.py", line 49, in decorate_no_grad
return func(*args, **kwargs)
File "/Users/fedor/anaconda3/envs/vpg-env/lib/python3.6/site-packages/catalyst/rl/core/trajectory_sampler.py", line 139, in reset
self._init_with_observation(observation)
File "/Users/fedor/anaconda3/envs/vpg-env/lib/python3.6/site-packages/catalyst/rl/core/trajectory_sampler.py", line 77, in _init_with_observation
self.observations.append(observation)
File "/Users/fedor/anaconda3/envs/vpg-env/lib/python3.6/site-packages/catalyst/utils/dynamic_array.py", line 154, in append
value = self._as_dtype(value)
File "/Users/fedor/anaconda3/envs/vpg-env/lib/python3.6/site-packages/catalyst/utils/dynamic_array.py", line 140, in _as_dtype
for key in self._dtype.fields.keys():
AttributeError: 'NoneType' object has no attribute 'keys'
Which tells nothing about what actually happened. It happened that I forgot to instantiate my observation space as a gym.spaces.Dict(). So I'm proposing to add a little self-explanatory exception to help if people who will run into the same problem.