python - Fast sequential lists for tensorflow? -
i have array a
of matrices (or 3-dim tensor) , want following:
denote each matrix number, a
[1,2,3,4,...,]
, , let's have window of length 3, want pass input tensorflow graph 4-dim array [[1,2,3],[2,3,4],[3,4,5],....]
. what's efficient way of doing this? (it's bit convolution constant kernel, without summing on resulting matrices).
at moment i'm doing:
input_nn = [data[t, t + window] t in range(my_range)]
and pass tf placeholder.
shall think of better way of doing in numpy
, pass result placeholder or there fast way of doing in tensorflow passing a
directly?
Comments
Post a Comment