Tensorflow: Retraining inception v3 shows 0 % GPU usage -
i trying out tensorflow's inception model , while use gpu bottlenecks generation, doesn't seem on training-wise.
- gpu usage monitored through
nvidia-smi
, stays @ 0% - log_device_placement returns gpu:0 80% of operation seems fine.
- at beginning detect titan x usual
/job:localhost/replica:0/task:0/gpu:0 -> device: 0, name: titan x (pascal), pci bus id: 0000:01:00.0
- the memory on gpu allocated
- cuda installed , cudnn both last version
- the temperature rise (about 20 celsius 40->60) cpu heating whole case?
another way check whether gpu used nvprof. can tell sure.
for many models, check argument specify gpu count. example: num_gpus. few examples default not use gpu unless specified.
Comments
Post a Comment