65% CPU & 15% Memory with Tensorflow on Google Cloud ML -
i'm running google cloud ml job using "scaletier": "basic_gpu"
, , following chart outlines details on utilisation:
i'm executing experiment using learn_runner.run(...)
on custom estimator , feed input using pipeline based approach using file name queue read data.
is using pipeline based approach main reason low memory utilisation , there should consider optimize training utilisation?
Comments
Post a Comment