More disk space or RAM (sent 2017-04-30 16:38 CET)

Dear shared task participants,

some of you have found out that their trained models are too large for the default parameters of the TIRA virtual machine. It is actually possible to get more resources (disk space, RAM, …) if your system needs it.

If this applies to you, estimate how much extra space you need and then ask Martin Potthast to upgrade your VM.

Remember that during the test run you will be processing one test file at a time. You don’t have to keep the models for all the languages in memory during the entire run—unless you have one huge model that covers all languages, of course. To give you an estimate of how much data you will have to process (although it is not directly related to the size of your models): we expect to have probably 84 test files with something over 1.5 million words. That includes the extra test data from the new parallel treebank, and the surprise languages.

Best regards, Dan Zeman on behalf of the costocom :) (connl shared task organizing committee) http://universaldependencies.org/conll17/