I really dont think that you can compute UniFrac distances without a tree. Remember that Unifrac "incorporates information on the relative relatedness of community members by incorporating phylogenetic distances between observed organisms in the computation" (wikipedia). To incorporate phylogenetic distances, you need a phylogenetic tree. Such tree cannot be constructed from the OTU table, and if you do, is not going to be phylogenetic, and therefore the computation of UniFrac is not really UniFrac.

Thank you for your answer. Since you stated that a tree cannot be constructed from the OTU table, i assume i need to transfer my raw sequence data to QIIME and repeat the process? 

Either that, or construct your tree by any other means (RaXML, MEGA, PHYML, PAUP, Mr.Bayes), and then use any of the R packages. I recommend you stick to likelihood-based methods (either frequentist of bayesian), rather than distance measures such NJ.

Best

Though late, Jose is right. UNIFRAC is totally a phylogenetic distance approach for diversity estimation. Since you have your OTU table, you will also have your OTU representative sequences (centroid sequences). Just take the dataset and use QIIME for three steps a) align_seqs.py, b) filter_alignment.py and c) make_phylogeny.py. This should provide you with the phylogenetic tree. Point to note: If you generated the OTU table by  using other pipelines or program then it should be considered that sometimes many of your OTU representative sequences will fail to align with the reference 16S database (greengenes, silva or rdp). Don't forget to remove these failed OTUs from the OTU table as well. Else there will be a difference in the number of OTUs in otu table and OTU representative sequences in fasta file which finally formed the phylogenetic tree.

QIIME has scripts for this purpose as well.

How do I generate OTU table? I actually have my tree generates by RaxML. So to use the skbio.diversity.beta.unweighted_unifrac, I will need to have the u_counts, v_counts.

Is there any software to use on raw data and get the abundances? I'm a bit lost how to compute v_counts and u_counts actually. Thanks!