We’re releasing 3 extra test cases for your help. These are just for your analysis, you don’t need to submit them during the actual submission.
I am getting an inference error upon submission. Please guide.
My policies exactly match the 3 test cases target policies, my value functions after convergence with tolerance 1e-8 match till 1st decimal point for all 3 test cases. Still AIcrowd score is coming as 0. Do we need to exactly match the value functions, or else to what decimal point should they match?
Do you still get the error? If yes share the submission id.
The tolerance on value function match has been relaxed significantly, very approximate matches should give correct score. Correctly implementing the algorithm should give you at least a match around 2-3 decimal places.
Yeah, got it. Thanks.