Closed eenin closed 7 years ago
Did you check that you have no sentences longer than 1024 words? Please see https://github.com/facebookresearch/fairseq/issues/57
Yeah, that was the problem. Thanks!
I have meet the same issue ,but I don't konw how to limit sentence size in the training files. could you give me some help,Thanks! @michaelauli
This script removes sentence pairs whose source or target are longer than a specified number of words: https://github.com/moses-smt/mosesdecoder/blob/master/scripts/training/clean-corpus-n.perl
Hi, I am trying to train a fully conv. model on a en-fr pair, but it keeps crashing ~30mins into the training. Here is the stacktrace of the assertion errors I am getting:
I trained another language pair successfully, but I cannot understand why this is not working