lucidrains / memorizing-transformers-pytorch

Implementation of Memorizing Transformers (ICLR 2022), attention net augmented with indexing and retrieval of memories using approximate nearest neighbors, in Pytorch
MIT License
620 stars 46 forks source link

index out of #7

Open chxiag opened 1 year ago

chxiag commented 1 year ago

when I run train.py, error like this ,"index out of range: Tried to access index 10218 out of table with 255 rows. at /pytorch/aten/src/TH/generic/THTensorEvenMoreMath.cpp:418"happens