add continuation indicator layer for intel caffe

[lddsdu]

The continuation_indicator_layer is used to produce one blob as the bottom input for lstm layer, and this layer used the parameter below message ContinuationIndicatorParameter { optional uint32 time_step = 1 [default = 0]; optional uint32 batch_size = 2 [default = 0]; } The shape of this blob is (time_step, shape), this may useful when using lstm layer.

[ftian1]

thanks @lddsdu . we usually do same thing by having a python layer. it's mainly because user may not know the time step before data preprocessing. so it's hard to assign a max value in proto file. but anyway your path looks good for me.

[lddsdu]

@ftian1 Thank you! by the way, I want to know if you would like to add 3d convolution layers in intel caffe, when I using 2d convolution in a 40 cores cpu computer server, it is about 1.5 times as fast as the tesla p4. It's really amazing,I need to use 3d convolution when dealing with videos.