I have a program that produces a tensor, I can print the shape,dtype of this tensor, and it looks good. when I try to print the value of this tensor by print(x)
, there goes the error message:
(Pdb)*** RuntimeError: numel: integer multiplication overflow
(Pdb)*** RuntimeError: numel: integer multiplication overflow
(Pdb)*** RuntimeError: CUDA error: misaligned addressCUDA kernel errors might be asynchronously reported at some other API call,so the stacktrace below might be incorrect
But I do not know when this happed, it will only become a error when this tensor is accessed.
How to figure this out? Thanks a lot!