search

zhejz / carla-roach

Roach: End-to-End Urban Driving by Imitating a Reinforcement Learning Coach. ICCV 2021.
https://zhejz.github.io/roach
Other
274 stars 50 forks source link

Setting the argument of n_episodes #3

Closed yixiao1 closed 2 years ago

yixiao1 commented 2 years ago

Hi,

Very impressive work! One simple question, when we set up the argument of n_episodes, if we need to take into account of the number of towns? For example, if I want to collect 1 episode per towns (Town 1, 3, 4, 6). Shall I set up the n_episodes to 1, or 4?

And for each episode, is the route (the start point and the end point) always set to the same?

Cheers, Yi

zhejz commented 2 years ago

Hi, thanks for liking our work! The n_episodes is the total episode collected, so in your example it should be set to 4 such that 1 episode will be collected for each town. The start/end point depends on the gym_env specified. For benchmark envs (NoCrash, CoRL2017, Leaderboard), the start/end point is specified by the xml file. For data collection and RL training there is an EndlessEnv, which has randomly selected start/end point.

yixiao1 commented 2 years ago

Alright, I see. So I understand that the number is the amount of the total episodes among all towns. So if we set up the n_episodes to a number, let's say 60, this number will be automatically divided by 4, which means, 15 episodes per town.

Thanks a lot, Yi