Contributing

In this repository we welcome contributions. If you found a problem, or want to add a feature, please create a Pull-Request or open an Issue.

We need your help!

There are a few problems we haven't managed to solve on ourselves. You might be the one to solve them!

In NaivePlanner, calling plan_best does not clear memory. Instead, if called in a loop will kill the process with an out-of-memory error.
In OpenNMTModelRunner, calling train returns all checkpoints, which then need to be evaluated. While this gives a lot of flexibility, evaluating every checkpoint is very slow, and we're looking for a trick to evaluate BLEU in training time.
In OpenNMTModelRunner, calling train hides the stdout as intended, but it is hard to gauge the time it takes to train. It would be nice to parse the stdout and run a tqdm progress bar.