TracerWarning

[maingoc1605]

when i using torch.jit.trace to convert model, i have this warning. How can i turn off this warning
TracerWarning: Converting a tensor to a Python boolean might cause the trace to be incorrect. We can't record the data flow of Python values, so this value will be treated as a constant in the future. This means that the trace might not generalize to other inputs!
assert condition, message