Hi there,
first of all congratulations to this great work and thanks a lot for sharing your code!
I have a short question regarding the results from the paper with respect to the posted training logs:
Why do the numbers for modular results (Section 3.2, Table 3,4,5,6,7) differ from the numbers in the stage2 training log ?
Are all results taken from the stage2 training? Especially the planning results are slightly worse in the training log as compared to the paper.
Do you report the numbers for the perception tasks with respect to stage 1 or stage 2? E.g. the tracking performance is significantly higher in the stage1 setting?
May you post the training log of the run used in the paper, or shall we simply refer to the numbers provided in the stage 2 training log, especially for metrics that are not given in the paper?
Hi there, first of all congratulations to this great work and thanks a lot for sharing your code! I have a short question regarding the results from the paper with respect to the posted training logs: Why do the numbers for modular results (Section 3.2, Table 3,4,5,6,7) differ from the numbers in the stage2 training log ?
Are all results taken from the stage2 training? Especially the planning results are slightly worse in the training log as compared to the paper. Do you report the numbers for the perception tasks with respect to stage 1 or stage 2? E.g. the tracking performance is significantly higher in the stage1 setting?
May you post the training log of the run used in the paper, or shall we simply refer to the numbers provided in the stage 2 training log, especially for metrics that are not given in the paper?
Thanks a lot in advance for any help!