WARNING:tensorboardX.x2num:NaN or Inf found in input tensor.

[HX-Jhao]

请问这个警告你在跑的时候也会出现吗？
对模型的训练有影响吗？

[ardian-selmonaj]

This warning is not very crucial, however, it can happen that training may stop at a certain time. I also encountered this warning, which is mainly due to some slight version mismatch of packages. Alternatively, you can tweak learning rate (lr), entropy coefficient (entropy_coeff) or gradient clipping (grad_clip) within .training in the algorithm.