When I ran pre-train, pc said that the data of the graph was not placed in the gpu. Is it related to the dgl version? Or is there a problem with the code you wrote

[NeuJzh]

dgl._ffi.base.DGLError: Cannot assign node feature "h" on device cuda:0 to a graph on device cpu. Call DGLGraph.to() to copy the graph to the same device.

[NeuJzh]

[NeuJzh]

I use debug code，this place might have some errors。

[cai-yuxiang]

yes , i also meet the problem. Do you deal with it?

[lemony-sea]

copy the graph to the device : cuda0 use the following code : for key in graph_dict: graph_dict[key] = graph_dict[key].to(torch.device("cuda" if use_cuda else "cpu"))

[suojifeng]

thanks