all passes‘ cost is 0 besides the first pass

[Damon-wyg]

I train a dnn model with cluster paddle，but all pass cost is zero，checked the data which is ok, and local running can get normal cost。As follows：
[RIG_NOTICE] log_time=1498196443 cost=0.006403 job_id=job.38811.instances
[RIG_NOTICE] log_time=1498196793 cost=0.000000 job_id=job.38811.instances
[RIG_NOTICE] log_time=1498197141 cost=0.000000 job_id=job.38811.instances
[RIG_NOTICE] log_time=1498197488 cost=0.000000 job_id=job.38811.instances
[RIG_NOTICE] log_time=1498197834 cost=0.000000 job_id=job.38811.instances
[RIG_NOTICE] log_time=1498198182 cost=0.000000 job_id=job.38811.instances
[RIG_NOTICE] log_time=1498198529 cost=0.000000 job_id=job.38811.instances
[RIG_NOTICE] log_time=1498198878 cost=0.000000 job_id=job.38811.instances
[RIG_NOTICE] log_time=1498199227 cost=0.000000 job_id=job.38811.instances
[RIG_NOTICE] log_time=1498199577 cost=0.000000 job_id=job.38811.instances
[RIG_NOTICE] log_time=1498199925 cost=0.000000 job_id=job.38811.instances
[RIG_NOTICE] log_time=1498200272 cost=0.000000 job_id=job.38811.instances
[RIG_NOTICE] log_time=1498200619 cost=0.000000 job_id=job.38811.instances
[RIG_NOTICE] log_time=1498200967 cost=0.000000 job_id=job.38811.instances
[RIG_NOTICE] log_time=1498201316 cost=0.000000 job_id=job.38811.instances
[RIG_NOTICE] log_time=1498201665 cost=0.000000 job_id=job.38811.instances
[RIG_NOTICE] log_time=1498202013 cost=0.000000 job_id=job.38811.instances
[RIG_NOTICE] log_time=1498202360 cost=0.000000 job_id=job.38811.instances
[RIG_NOTICE] log_time=1498202704 cost=0.000000 job_id=job.38811.instances
[RIG_NOTICE] log_time=1498203052 cost=0.000000 job_id=job.38811.instances
[RIG_NOTICE] log_time=1498203401 cost=0.000000 job_id=job.38811.instances
[RIG_NOTICE] log_time=1498203757 cost=0.000000 job_id=job.38811.instances
[RIG_NOTICE] log_time=1498204118 cost=0.000000 job_id=job.38811.instances
[RIG_NOTICE] log_time=1498204478 cost=0.000000 job_id=job.38811.instances
don't understand why~~

[Damon-wyg]

I've resolved it, which is caused by that wrong train data list path was writed. fixed in the trainer_config will be ok.