Contents

% c_dtype = dtype_to_c_type(dtype)
 // same dimension add floating point op
 __kernel void apply_adadelta_<%= dtype %>(const int M, const int N,
                                       __global const <%= c_dtype %> *lr,
                                       __global const <%= c_dtype %> *rho,
                                       __global const <%= c_dtype %> *epsilon,
                                       __global const <%= c_dtype %> *grad,
                                       __global <%= c_dtype %> *output,
                                       __global <%= c_dtype %> *acc,
                                       __global <%= c_dtype %> *acc_update
                                       ) {
    // Get the index of the current element to be processed
    const int globalRow = get_global_id(0); // Row ID of C (0..M)
    const int globalCol = get_global_id(1); // Col ID of C (0..N)
    const int index = globalRow * N + globalCol;

    acc[index] = acc[index] * rho[0] + (grad[index] * grad[index]) * ((<%= c_dtype %>)1 - rho[0]);
    const <%= c_dtype %> update = sqrt(acc_update[index] + epsilon[0]) * rsqrt(acc[index] + epsilon[0]) * grad[index];
    output[index] -= update * lr[0];
    acc_update[index] = acc_update[index] * rho[0] + update * update * ((<%= c_dtype %>)1 - rho[0]);
}

Version data entries

2 entries across 2 versions & 1 rubygems

Version	Path
tensor_stream-opencl-0.1.3	lib/tensor_stream/opencl/kernels/apply_adadelta.cl
tensor_stream-opencl-0.1.2	lib/tensor_stream/opencl/kernels/apply_adadelta.cl