Intrinsic phase timeout

We’ve got an error “Timeout : Evaluation took too long.” for round 2 submission,
even our agent do nothing. (return np.zeros(action_space.shape[0]))

We also noticed that the number of intrinsic steps was given as 10000000,
that would be take the time to simulate for more than 1 week.

Is this the intended number of intrinsic steps?

Hi @kim135797531,

Please update your real_robots package to latest one (0.1.16) and create a new submission. We have made some stability fixes which will be important to include in your Round 2 submission.


On the amount of timesteps of the intrinsic phase: yes, 10M is the correct figure.
We gave quite some time to the agents to try things and learn :wink:
An agent which does nothing is evaluated at about 50 steps/s, so it takes 2 days and 8 hours just to run the environment for the intrinsic phase.
Note that the environment uses only 1 CPU and basically no GPU, so if one makes computations in parallel (e.g. updating/training neural networks) there’s a lot of room (7 CPU and the GPU) to spare.
If training is not done in parallel, yes, it could take a week.

Hi, I am using the latest real_robots package (0.1.16) and found out that my evaluation is stuck around 2M steps for a few hours. I made no change to the policy in the starter kit (RandomPolicy) just to see how long it takes to submit no-learning agent. I assume there is still something that slows down the evaluation in the environment.
Could you investigate a little more?