Closed jindrahelcl closed 7 years ago
Where is this happening?
I am trying to run this now, maybe I've got it wrong. I've run the training from commit cc1ec7d. I use the same ini script that I used for training (see attachment qtleap-alpha.ini
), and I use the val_dataset.ini
as the dataset config.
It seems the dataset arguments are not ignored.
I am also getting errors probably caused by the dataset init helper:
Error while loading 'val_data': '_sre.SRE_Match' object is not subscriptable
(...)
File "/home/jihe01/mmmt/bin/neuralmonkey/config/utils.py", line 46, in <dictcomp>
if SERIES_OUTPUT.match(key)}
val_datasets.ini.txt qtleap-alpha.ini.txt
NOTE the out_...out is just a thing I tried.. before this change there was the same error with s..._out keys..
It seems I've figured out the error. There was a leftover Python 2 matching of the regex.
Is this solved?
It shoudl be solved in the experiment object branch. (#73) (berliner refactor)
I don't think it happens any more, though I don't when it got fixed.
AFAIK it's not fixed.. I use two separate configuration files for this.
2016-11-30 13:19 GMT+00:00 Jindřich Libovický notifications@github.com:
I don't think it happens any more, though I don't when it got fixed.
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/ufal/neuralmonkey/issues/67#issuecomment-263871325, or mute the thread https://github.com/notifications/unsubscribe-auth/ABwcs1DK6m8Wuoej7sEZPl9gT75Bow2Lks5rDXfngaJpZM4JLO10 .
According to the tests cases logs it does not happen any more.
:D :D cože? :-)
:-D už je to dobře
škoda.. :-)
Isn't it somehow connected with #210 ?
Not at all. This was about handling configuration in run.py
script.
Runner should not load the training dataset again.