Error While training the model for ger-eng

Support
1

Hi,

I am presently building a model for GER-ENG Translation. The pre-process was successful. While I
run train command
“python train.py -data data/demo -save_model demo-model -world_size 1 -gpu_ranks 0”,
I get the following error messages :

  1. RuntimeError: merge_sort: failed to synchronize: an illegal memory access was encountered
  2. RuntimeError: CUDA error: an illegal memory access was encountered

Could someone please assist me in resolving this error ?

Thank You,
Kishor.

2

Hi,

Make sure you are using the latest version, i.e. OpenNMT-py 0.9.2 with PyTorch 1.2.