GPU usage for training

[kanayer]

Currently, my dataset has around 230 000 sentences and it takes about 30 hours to train the model. Is it possible to use GPU for the training process?

[mdelhoneux]

I don't really have plans to implement this at the moment.
RIght now, UUParser does not do any kind of batching and therefore the use of GPUs would not provide any speedup.
If you want to use it on GPU, you would first need to use batching. Dynet has autobatching (https://dynet.readthedocs.io/en/latest/tutorials_notebooks/Autobatching.html) which I tried using for UUParser a while ago but accuracy was just too low and I gave up on this.