Created by: water1981

现在我们的seq2seq中的机器翻译例子是产生式的（is_generating=True），我想把这个任务改成判别式的（就是只是给出源语言翻译到给定目标语言的得分），训练的时候遵循https://github.com/PaddlePaddle/Paddle/blob/develop/doc/tutorials/text_generation/index_en.md 里面的步骤训练出一个翻译模型， 在test阶段在gen.conf 中设置 is_generating=False ， 结果出core了，错误信息如下： I /home/img/baidu/idl/paddle/paddle/trainer/Trainer.cpp:148] trainer: in testing mode I /home/img/baidu/idl/paddle/paddle/trainer/Trainer.cpp:155] trainer mode: Testing I /home/img/baidu/idl/paddle/paddle/gserver/dataproviders/PyDataProvider2.cpp:247] loading dataprovider dataprovider::process [INFO 2016-12-15 13:57:33,367 dataprovider.py:27] src dict len : 30000 [INFO 2016-12-15 13:57:33,367 dataprovider.py:37] trg dict len : 30000 I /home/img/baidu/idl/paddle/paddle/gserver/dataproviders/PyDataProvider2.cpp:247] loading dataprovider dataprovider::process [INFO 2016-12-15 13:57:33,416 dataprovider.py:27] src dict len : 30000 [INFO 2016-12-15 13:57:33,416 dataprovider.py:37] trg dict len : 30000 F /home/img/baidu/idl/paddle/paddle/trainer/Tester.cpp:286] Check failed: (paramUtil_->loadParameters(passId, true , true )) == (true) /home/img/liushui/bin/paddle: line 81: 4510 Aborted (core dumped) ${DEBUGGER} $MYDIR/../opt/paddle/bin/paddle_trainer ${@:2}