I am looking to extract the normalized FPKM values that DEseq2 uses for it's models. I know that it it likely is not stored. Is there a way to generate them? 

hi,

DESeq2 uses the raw counts in modeling. For internal normalization, it by default accounts only for library size. By "internal", I mean this happens on the right side of equation (1) in the Methods section our paper.

You can generate normalized counts with counts(dds, normalized=TRUE). But this does not divide out the feature size. These are the original counts with each column divided by a factor such that they can be compared across samples.

If you use our newest import pipeline, tximport, then DESeq2 will also account for potential changes in the gene length (due to differential transcript usage) across samples. Then you can also use the fpkm() function to generate normalized values, although keep in mind that these normalized values are not used by DESeq2, which directly models the counts.

Take a look at the current tximport and DESeq2 vignettes for information on this pipeline.