The buggy_script.py file uses the SampleCollector class from mini_rllib.py
(all located in this root directory here) to collect observation samples
in parallel from n RL environments or in sequence from a single environment.
Each observation is a simple int for simplicity (discrete observation space). Results from the different episodes are concatenated and returned.
However, using one parallel and one sequential setup of SampleCollector yields
different results (which should not be the case), even when
SampleCollector.sample() is provided with the same initial random seed.
Try to fix the assert at the end of buggy_script.py by finding the bug(s).