Hi, 1. easy way- use on line training here "http://bioinf.uni-greifswald.de/webaugustus/training/create" and use the paramaeter for you species.
you can prepare training data set from the goat refseq files from NCBI (ftp://ftp.ncbi.nlm.nih.gov/genomes/refseq/vertebrate_mammalian/Capra_hircus/latest_assembly_versions/GCF_001704415.1_ARS1/).
- You can follow the instructions given in the following links for making custom files and training AUGUSTUS. "https://vcru.wisc.edu/simonlab/bioinformatics/programs/augustus/docs/tutorial2015/training.html" "http://avrilomics.blogspot.in/2013/04/training-augustus-gene-finding-software.html" you can use refseq RNA and protein files from the NCBI link for getting RNA and PROTEIN files.
Notes: I used MAKER to annotate genome initially using RNA and Protein sequences from the closest species and the GFF output from this was used to get 1) multi-exonic 2) non-overlaping 3) relatively non-close sequences (<70% identity) transcripts. I have converted this GFF file (after filtration) to gb file format using "./scripts/gff2gbSmallDNA.pl" available in AUGUSTUS and used further according to the instructions given in above mentioned links (to be specific first link).