-
Hierarchical multithreading appears to be broken (haven't tested with phrase-based). For the big picture, here is a plot of decoding times with Moses for a large German-English grammar (Europarl + Com…
-
Hi team,
Apart from ``` perplexity``` score is there any other way for measuring model performance within openNMT.
For example do we have a way to calculate BELU score of the model built using open…
-
Aside from the travis checks, we definitely need regular (weekly?) tests of learning performance. These tests should take e.g. 8 hours on a GPU.
Let's use this issue to come up with a plan how to b…
-
I got quite low performance compared to the paper.
So, i did some research, and I found the sizes of wmt_ende_tokens_32k-{dev, train}* are too small as follows.
444K wmt_ende_tokens_32k-dev-00…
-
Does ModernMT use some sort of cross validation to avoid overfit with its translations? If so, what technique is used to avoid overfit?
-
Hello
I am using rather old GPU GeForce GT 640, but it is CUDA compatible.
Nvidia drivers are installed as per https://github.com/ModernMT/MMT/blob/master/INSTALL.md#pre-installation-actions
I am…
-
Would it be possible to perform R2L rescoring with amunmt? How could I integrate amunmt into Rico's r2l translation script?
```
#!/bin/bash
# this sample script translates a test set, including…
-
**For training sparse features (as with `train/config.sparse.yml)`:**
`Line 293` of `train/run_cross.perl`:
The following file does not exist: `$DIR/cross.00/work.err-cor/binmodel.err-cor/moses.…
-
The BLEU implementation has sort of generated quite a lot of "wack-a-mole" situations where the fringe cases of using BLEU cause errors or unexpected output values from the following issues: #1328 #12…
-
Use neuralmonkey to replicate the scores of [Bahdanau et al. (2015)](https://arxiv.org/pdf/1409.0473v7.pdf) on English-to-French, [WMT-14 translation task](http://www.statmt.org/wmt14/translation-task…