Node regression dtype error in .backward()

[bhanu1106]

I wanted to do a simple node regresssion problem , so i used coral dataset from pytorch geometric and generated random values for target variable . On this toy dataset, I had tried to do a node regression using torch.nn.MSELoss() loss function . Now I am facing dtype error in .backward() as follows : RuntimeError: Found dtype Double but expected Float . I even tried to change the data type from float64 to float32 . But facing the same error .

[rusty1s]

You should check that out.dtype and data.y.dtype match. Can you confirm?

[bhanu1106]

Thank you very much for the reply . changing data.y.dtype from float64 to float32 solved the problem