8

Here offered the way how to profile tensorflow code. In my case I launch tf.run in several threads in parallel. How I can use this technique to profile multithreading architecture? When I use global metadata and options they log only single thread.

Thanks!

Community
  • 1
  • 1
Andrey Atanov
  • 243
  • 2
  • 12
  • 2
    Each run metadata profiles a single run call, if you have several.run calls in parallel, you could combine metadata info manually – Yaroslav Bulatov Oct 06 '16 at 16:35
  • If you are using tf.Estimator, ProfilerHook is the way to go. See https://www.tensorflow.org/api_docs/python/tf/train/ProfilerHook – Urs Dec 16 '17 at 00:14

0 Answers0