Deletes README that was sitting inside of visualize folder
When you follow the install instructions, you end up getting an error when installing from requirements.txt because pip can't find carla. I think this is because you automatically exported requirements.txt, even though carla wasn't installed via pip? Instead, you can ask the user to set $CARLAROOT in the README, then in the script where you append to PYTHONPATH also append the PythonAPI from their CARLAROOT.
modified requirements.txt
modified precog_env.sh
Changed scenario_runner.py to take two new args, max_episodes and video_name
I did this so that could automate the test experiments (since each takes >1hr and otherwise has to be manually exited on the right number of episodes)
Also modified scenario_runner.py to clear the images folder before running again (so that it can be run automatically back-to-back)
Added a test script test.py that runs the test experiments (from slack) serially using the new args (max_episodes/video_name)
I'm not sure if we actually want this in the public release, but I think it would be good to have something similar that runs ALL the experiments from the paper (so that we can say "to reproduce the results from the paper run test.sh")
Also, this script makes it clear that a checkpoint is included in the repo itself and shows how to use it
carla. I think this is because you automatically exported requirements.txt, even though carla wasn't installed via pip? Instead, you can ask the user to set $CARLAROOT in the README, then in the script where you append to PYTHONPATH also append the PythonAPI from their CARLAROOT.