I want to train a model for text summarization and I run into a problem when predicting the summary.
My model uses @pltrdy’s implementation of “A Deep Reinforced Model for Abstractive Summarization”.
The errors can be found here but they include lines like
/pytorch/aten/src/THC/THCTensorIndex.cu:308: void indexSelectSmallIndex(TensorInfo<T, IndexType>, TensorInfo<T, IndexType>, TensorInfo<long, IndexType>, int, int, IndexType, long) [with T = float, IndexType = unsigned int, DstDim = 2, SrcDim = 2, IdxDim = -2]: block: [0,0,0], thread: [127,0,0] AssertionsrcIndex < srcSelectDimSize
a lot. The whole error can be found here.
I guess the reason is that I have very long input sequences (up to 90000 tokens) and the model is not able to cope with that, although I am not sure about this.
I would really appreciate if you could give me a hint how to solve this issue.