question: some positive result after python -m joeynmt train configs/small.yaml about joeynmt HOT 3 CLOSED

Hi,
The small config is just an example showing what setting Joey has, it does not result in a good translation system.
Try out the reverse task instead (reverse.yaml). Check out scripts/generate_reverse_task.py to generate the data for it.

from joeynmt.

I fact I try use my dataset: it 543k so I split it 540000 train and 35000 dev. BUT it's not translation it's BOT and language used Finnish :( .

But after about 7-8 hours on Nvida Tesla P100 GPU Hypothesis: still empty. It's a little bit strange for me? As config I use small.yaml . Probably I must switch on reverse.yaml ? But I don't see particular difference.

from joeynmt.

The model there is not configured well for your task I'm afraid.
Try using this one

Make sure to change the data set correctly in the config.

from joeynmt.