OPUS-MT-eval - benchmarks for evaluating MT models

A collection of machine translation benchmarks. Currently it includes:

• WMT news translation
• Tico19 tanslation benchmark
• Test sets from the Tatoeba translation challenge