These are chat archives for beniz/deepdetect

8th
Nov 2017
rperdon
@rperdon
Nov 08 2017 14:42 UTC
I still had an error with the new GPU build, I've been on training this week so I will further diagnose friday or next week.
I restarted the process with async off and it started training; nevermind

INFO - 14:45:42 - Iteration 40, lr = 0.001[14:45:43] /opt/deepdetect/src/caffelib.cc:812: smoothed_loss=1.29258

INFO - 14:45:58 - Iteration 80, lr = 0.001[14:45:59] /opt/deepdetect/src/caffelib.cc:812: smoothed_loss=0.485691

INFO - 14:46:14 - Iteration 120, lr = 0.001[14:46:14] /opt/deepdetect/src/caffelib.cc:812: smoothed_loss=0.397993

INFO - 14:46:30 - Iteration 160, lr = 0.001[14:46:30] /opt/deepdetect/src/caffelib.cc:812: smoothed_loss=0.299478

rperdon
@rperdon
Nov 08 2017 15:51 UTC
CPU 60 hrs 4500 iterations; GPU 1 hour 9000 iterations.
Emmanuel Benazera
@beniz
Nov 08 2017 16:37 UTC
sounds good :)