solver.prototxt meaning

[rahim entezari]

Hi

My questions are about settings in solver.prototxt

1. # Carry out testing every 500 training iterations.
   

            test_interval: 500   >>> so if my training set if 10 times larger than test set,I should set test_interval=10?

       2. what is appropriate base_lr when fine-tuning? 0.01?

       3.what is momentum? and best value?

       4. the same as 3 for weight_decay?

       5.the same for gamma?

       6. what are the  options for lr_policy other that "step"?