Parameter synchronization mode

yahoo / CaffeOnSpark

Distributed deep learning on Hadoop and Spark clusters.

Apache License 2.0

1.27k stars 357 forks source link

Parameter synchronization mode #295

Closed bjlgcxc closed 6 years ago

bjlgcxc commented 6 years ago

As I can see, CaffeOnSpark do the synchronization of parameters per iteration. Is there some way to let CaffeOnSpark do that in round, which contains certain number of iterations.For example, do the parameter synchronization ervery 10 iterations.(This idea is similar to the mode that SparkNet adopt) Looking forward to your reply! 3Q

junshi15 commented 6 years ago

It is not supported directly. To achieve that you have to change the on_start() and on_gradients_ready() function in this file.

bjlgcxc commented 6 years ago

ok, thank you very much! I will try it later