Any plan to distributed DIGITS?

[dccho]

I'm really happy to use DIGITS in multi-GPU system. It reduce training time almost 20 times.
However, it still needs almost one day to train little bit big network.
Do you have any plan to release distributed DIGITS?
If a dataset is big enough (such as Imagenet or Places DB), data parallelism is suitable solution for DIGITS.
Hope to release it ASAP.

[lukeyeager]

Do you want one training job on multiple nodes? See BVLC/caffe#3252

Do you want to schedule single-node training jobs on multiple nodes? See #108.