Yonatan Belinkov
Yonatan Belinkov
The charCNN models were trained with [seq2seq-attn](https://github.com/harvardnlp/seq2seq-attn), please see the instructions there. Once a model has been trained, you can use [this script](https://github.com/ybisk/charNMT-noise/blob/master/dump_charcnn_weights.lua) to inspect the weights. The meanChar models...
I think these models don't have the charCNN part. You'll need to train a model with the character options from seq2seq-attn: `-use_chars_enc 1` and `-cudnn 1` (cudnn is not a...
We could possibly make some pre-trained models available if people find that useful. @ybisk , what do you think?
Hi there, For each original sentence, there will be one noisy sentence. The percentage perturbed is something we vary, and I think it's at the level of sentences iirc. @ybisk...
Did you end up finding a solution? We're having a similar problem on Windows, but the same text runs smoothly without encoding problems on Mac/Unix, so we're guessing it's an...
Yes, that works - thank you!