Created by: zhouyuan0828

I have about 90w samples needed to generate text by seqToseq model. when generate locally, and trainer_count set as 1, it is quite slow. On the other hand, when i set trainner_count as 5, it will gen core as below. SO wonder to know generate text on cluster is ok? and how to do it , please help share a link? thanks yuan F1216 11:31:51.538795 18198 MemoryHandle.cpp:52] Check failed: size != 0 allocate 0 bytes

* Check failure stack trace: *

@ 0x13b2d38 google::LogMessage::Fail() F1216 11:31:51.540397 18199 MemoryHandle.cpp:52] Check failed: size != 0 allocate 0 bytes

* Check failure stack trace: *

@ 0x13b2c90 google::LogMessage::SendToLog()