search

percyliang / sempre

Semantic Parser with Execution
Other
829 stars 300 forks source link

Save training result #142

Closed MinseokCho closed 7 years ago

MinseokCho commented 7 years ago

I use sempre table version, so i run this model such like './run @mode=tables @data=u-1 @feat=all @train=1' But training model takes long time, so i want to save training model and test a new question on pre-trained model. Is it possible? If it is possible, could you tell me how to do it? Thanks.

ppasupat commented 7 years ago

The model is saved in the execution directory (default: state/execs/___.exec/params).

To load a saved parameter in a new run, add -Builder.inParamsPath path/to/params to the command line.