I think DESeq2 expects counts, so passing it some normalized values (like TPM, RPKM, FPKM) is probably not optimal.
Do you have RPKM, or RPM (Reads-per-million)? If you have the latter, you can convert them to RPKM by dividing each RPM by the length of the gene in kilobases. If you want to get counts, you could multiply your RPM by (Number of reads in Sample / 1 000 000) and that should give you counts, which you should be able to use with DESeq. Make sure you know whether you have RPMs or RPKMs.
See also: https://support.bioconductor.org/p/56275/#56299, which suggests converting back to count preferably, or if you're stuck with RPKM then use
limma instead of DESeq2.
Also, handy reference on RPKM, FPKM, TPM measures: https://www.rna-seqblog.com/rpkm-fpkm-and-tpm-clearly-explained/