I was running a demo with our standard sample data (60K pairs) that 
creates three models. We've been using the demo and sample data for 
years. The host is stable and there are no changes to our core system 
(Moses 0.91 & DoMY). Train/tun/eval on the first two models completed 
without incident. On the third model, mert-moses.pl unexpectedly failed 
during the third run. The failure was a moses binary segfault. This is 
most unusual. I reviewed all previous logs and nothing stands out.

I cleared the output, reset this third demo and started the tuning 
again. If it completes, I'll assume this was an edge-case. In the mean 
time, can anyone think of why moses would fail during the third run of 
mert-moses.pl?
_______________________________________________
Moses-support mailing list
[email protected]
http://mailman.mit.edu/mailman/listinfo/moses-support

Reply via email to