Hello Nathan, https://github.com/knowledgedefinednetworking/Unveiling-the-potential-of-GN… This link contains the models used in our paper: K. Rusek, J. Suárez-Varela, A. Mestres, P. Barlet-Ros, A. Cabellos-Aparicio; "Unveiling the potential of Graph Neural Networks for network modeling and optimization in SDN," in Proceedings of ACM Symposium on SDN Research (SOSR) , pp. 140-151, April 2019. https://github.com/knowledgedefinednetworking/demo-routenet/tree/master/tra… This other link contains the models that we used in our demo paper: J. Suárez-Varela, S. Carol-Bosch, K. Rusek, P. Almasan, M. Arias, P. Barlet-Ros, A. Cabellos-Aparicio "Challenging the generalization capabilities of Graph Neural Networks for network modeling," in ACM SIGCOMM Posters and Demos, August 2019 In this paper, we trained RouteNet on a dataset with samples from the 14-node NSFNET topology and a (larger) 50-node synthetically-generated topology, Then, we evaluated the trained models on samples from these two topologies and a third toplogy (24-node Geant2). Please, find the datasets used in this paper at the following link: https://github.com/knowledgedefinednetworking/NetworkModelingDatasets/tree/… Note that these datasets include variable capacity on links. Also, the RouteNet implementation used for the demo (https://github.com/knowledgedefinednetworking/demo-routenet/tree/master/code) has support to encode the link capacity on the hidden states of links. Regards, José El 22/09/2019 a las 16:57, Nathan Sowatskey escribió: