Hi,
Thank you for open-sourcing the code. I have a couple of questions regarding the implementation and reproducibility of the reported results.
From my reading of the current code, it seems that the implementation distills a decentralized one-step flow policy from a decentralized multi-step flow policy, which appears different from the joint multi-step flow policy described in the paper. Please correct me if I have misunderstood the implementation.
-
Would the authors consider releasing the implementation used to produce the results reported in the paper?
-
Would the authors also consider releasing the hyperparameters/configurations used to reproduce the paper results, particularly for MAC-Flow? It seems that some of these hyperparameters may need to vary across environments and tasks.
Thank you!
Hi,
Thank you for open-sourcing the code. I have a couple of questions regarding the implementation and reproducibility of the reported results.
From my reading of the current code, it seems that the implementation distills a decentralized one-step flow policy from a decentralized multi-step flow policy, which appears different from the joint multi-step flow policy described in the paper. Please correct me if I have misunderstood the implementation.
Would the authors consider releasing the implementation used to produce the results reported in the paper?
Would the authors also consider releasing the hyperparameters/configurations used to reproduce the paper results, particularly for MAC-Flow? It seems that some of these hyperparameters may need to vary across environments and tasks.
Thank you!