Taking long to train EMNLP

percyliang / sempre

Semantic Parser with Execution

Other

830 stars 299 forks source link

Taking long to train EMNLP #56

Open SaraMitchell opened 9 years ago

SaraMitchell commented 9 years ago

Hi folks

I have been tried to retrain the EMNLP 2013(in webQuestions db). It took for 4 days to read 453 samples of 3084.(but you said 3 days for complete) I have core i7,16G ram,ubuntu 15.04

Would you please tell me,where my problem is? I think I dont use of all my computer.so How can I?

lxafly commented 9 years ago

Hello, have you figured out the issue? I am using the parser and am experiencing similar issue. I'm wondering if I have done something wrong.

SaraMitchell commented 9 years ago

@lxafly hello.unfortunately....not yet.

amirSharif32dm commented 8 years ago

I think parasempre might be compiled by GPU or google's service.not a PC.... https://cloud.google.com/compute/pricing

but for sempre.........????!!!!!!!!

codeashcode commented 8 years ago

I am running overnight training for geo880 domain. I am using 64GB machine and it is running for more than 24 hours. I am not sure if it will complete soon or not? I am using beamSize: 100.

Is there a way to run it faster to see it working end to end - I am fine with low accuracy. Can higher value of beamSize or depth make it faster?