Training Error with Numba

[gwoksansan]

Hi,
when I run the training code, there is an error message with numba.errors.ConstantInferenceError. How can I solve this problem?
Should I update the version of CUDA or torch?