vpruthi/doc-exports
1
0
Fork 0
forked from docs/doc-exports
Code Pull Requests Activity
doc-exports/docs/modelarts/umn/modelarts_21_0083.html
Jiang, Beibei 781e07249c ModelArts 2021430 (GA) UMN 25072022 provided by R&D (third review) 
Reviewed-by: gtema <artem.goncharov@gmail.com>
Co-authored-by: Jiang, Beibei <beibei.jiang@t-systems.com>
Co-committed-by: Jiang, Beibei <beibei.jiang@t-systems.com>
2022-09-06 10:45:57 +00:00

1.4 KiB
Raw Permalink Blame History

Only Three Valid Digits Are Retained in a Training Output Log. Can the Value of loss Be Changed?

In a training job, only three valid digits are retained in a training output log. When the value of loss is too small, the value is displayed as 0.000. Log content is as follows:

INFO:tensorflow:global_step/sec: 0.382191
INFO:tensorflow:step: 81600(global step: 81600) sample/sec: 12.098 loss: 0.000
INFO:tensorflow:global_step/sec: 0.382876
INFO:tensorflow:step: 81700(global step: 81700) sample/sec: 12.298 loss: 0.000

Currently, the value of loss cannot be changed. You can multiply the value of loss by 1000 to avoid this problem.

Parent topic: Training Jobs