The full test vectors generated by the reference code are rather large
(approximately 28 MiB) and can not be comfortably included in the git
Instead, a compact representation containing the SHA256 digest of each
test vector set (compactVectors.json) will be used by default.
The tests will load and use the full test vector output if available.
The files can be generated from the reference implementation by something
$ ./testvectors512 > KEM-Kyber-512.full
$ ./testvectors768 > KEM-Kyber-768.full
$ ./testvectors1024 > KEM-Kyber-1024.full
[Copy the `.full` files to `testdata/`.]