TeamHG-Memex / tensorboard_logger

Log TensorBoard events without touching TensorFlow
MIT License
630 stars 50 forks source link

Check performance - maybe there is room for improvement? #5

Open lopuhin opened 8 years ago

meownoid commented 7 years ago

I believe that library should be executed in the separate process since the interprocess communication is mush faster than disk I/O.

lopuhin commented 7 years ago

@meownoid yes, this is also a valid approach, thanks for suggestion! Do you have any use-cases where performance is a concern?

meownoid commented 7 years ago

@lopuhin learning networks with pytorch for example