How to access solver parameters from a (loss-type) layer in caffe

[Vimal Thilak]

Hi, 

Does caffe have a mechanism to signal to the layers that the framework is transitioning from training the network to testing (or running validation) while running training? I understand that "test_interval" provides this information to the solver. I'd like to have access to this piece of information in one of my layers. Is there a way to do this in an elegant manner within the current framework? 

Thanks!

-Vimal