How to use a frequency real input matrix as input for a Latent Dirichlet Allocation (LDA) algorithm?

I am managing an input matrix table made of real values, where each entry correspond to the frequency of a word to be associated with a topic. These are the first rows:

0.000 0.000 0.850 2.825 0.000 0.750

0.000 0.000 0.000 4.933 0.000 0

1.150 0.000 0.000 4.300 0.350 0

0.000 0.000 0.000 0.000 0.600 0

0.000 0.000 0.000 13.900 0.600 0

15.636 4.55 1001.681 57.781 5.809 0.226

I've been studying several LDA packages, such as "topicmodels" ( http://cran.r-project.org/web/packages/topicmodels/index.html ) R package, and they all need an input integer matrix. I cannot normalize my real input matrix to integer values otherwise I will corrupt it, and it will loose its original sense.

How could I use my original real value table matrix as input to an LDA package?