search

IBM / EvolveGCN

Code for EvolveGCN: Evolving Graph Convolutional Networks for Dynamic Graphs
Apache License 2.0
517 stars 164 forks source link

The parameter "adj_mat_time_window" cannot be set to None. #18

Open maxmustermann123 opened 2 years ago

maxmustermann123 commented 2 years ago

When setting the parameter adj_mat_time_window: None in the parameters_example.yaml file, the experiment crashes with the error:

Traceback (most recent call last): File "/evolveGCN/run_exp.py", line 220, in <module> tasker = build_tasker(args,dataset) File "/evolveGCN/run_exp.py", line 122, in build_tasker return lpt.Link_Pred_Tasker(args,dataset) File "/evolveGCN/link_pred_tasker.py", line 38, in __init__ self.get_node_feats = self.build_get_node_feats(args,dataset) File "/evolveGCN/link_pred_tasker.py", line 98, in build_get_node_feats max_deg,_ = tu.get_max_degs(args,dataset) File "/evolveGCN/taskers_utils.py", line 60, in get_max_degs cur_adj = get_sp_adj(edges = dataset.edges, File "/evolveGCN/taskers_utils.py", line 96, in get_sp_adj subset = subset * (idx[:,ECOLS.time] > (time - time_window)) TypeError: unsupported operand type(s) for -: 'int' and 'str'

The explanation from the same file is: "# Time window to create the adj matrix for each timestep. Use None to use all the history (from 0 to t)". Is there another way to use all the history from 0 to t?

NikhilT-DS commented 1 year ago

Maybe you can comment out line number 96 in taskers_utils.py subset = subset * (idx[:,ECOLS.time] > (time - time_window))