Investigating env divergence
Submission that use import random
seem to diverge. will test to where the inference with the env happens.
Traceback (most recent call last):
File "./run.py", line 732, in <module>
observation, all_rewards, done, info = remote_client.env_step(action)
File "/srv/conda/envs/notebook/lib/python3.6/site-packages/flatland/evaluators/client.py", line 246, in env_step
raise Exception("local and remote `reward` are diverging")
Exception: local and remote `reward` are diverging