Correct Annovar input

Question

[storm1907]

Hello, I tried to convert my raw unfiltered Deepvariant vcf samples to Annovar format. First, I runned script with hg19 build, but got this error

WARNING: 1884 exonic SNPs have WRONG reference alleles specified in your input file!
WARNING: An example input line is <chr11        11964514        11964514        T       C       hom     54.5    171     chr11   11964514        .       T       C       54.5    PASS    .       GT:GQ:DP:AD:VAF:PL      1/1:54:171:0,171:1:54,62,0>
WARNING: ANNOVAR can still annotate exonic_variant_function for the mutation correctly!
WARNING: you may have used wrong -buildver, or specified incorrect reference allele, or used outdated mRNA FASTA file!

I switched to hg38 build, so I downloaded additional packages. Then I got this error message

A total of 27174 sequences cannot be found in humandb/hg38_refGeneMrna.fa

here is my command line

./convert2annovar.pl -format vcf4 $file -outfile $outfile -includeinfo -withzyg -withfreq

So which reference genome should I use then? Or there are certain file types I can only use for Annovar?

Thank you!