RuntimeError:one of the variables needed for gradient computation has been modified by an inplace operation
原因：0.4.0把Varible和Tensor融合为一个Tensor，inplace操作，之前对Varible能用，但现在对Tensor，就会出错了，所以找到模型中所有的inplace操作，换成非inplace的写法就行
参考：https://blog.csdn.net/DuinoDu/article/details/80435127
RuntimeError: cuda runtime error (59) : device-side assert triggered at /pytorch/aten/src/THC/THCTensorCopy.cu:102
/pytorch/torch/lib/THCUNN/ClassNLLCriterion.cu:101: void cunn_ClassNLLCriterion_updateOutput_kernel(Dtype *, Dtype *, Dtype *, long *, Dtype *, int, int, int, int, long) [with Dtype = float, Acctype = float]: block: [0,0,0], thread: [3,0,0] Assertion t >= 0 && t < n_classes failed.
原因：有可能是最后一层的layer的参数num_labels不对，导致GPU跑的时候出错
参考：https://github.com/pytorch/pytorch/issues/10303