You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Yes, currently we fix the maximum number of neighbors to 100. One way to increase this would be via templating, so that we can create specific kernels depending on the number of k, e.g.
when i use knn to calculate atom neighbour,it's easy count > 100,can i set k > 100?
in your cuda code u set k<=100
https://github.com/rusty1s/pytorch_cluster/blob/master/csrc/cuda/knn_cuda.cu line 98
AT_ASSERTM(k <= 100, "
k
needs to smaller than or equal to 100");The text was updated successfully, but these errors were encountered: