test bpe-gpt2/encode ... bench: 1,652,368,479 ns/iter (+/- 9,033,859)
test bpe-gpt2/encode-batch ... bench: 952,831,546 ns/iter (+/- 5,006,277)
test bpe-gpt2/encode-no-cache ... bench: 2,321,917,640 ns/iter (+/- 19,957,696)
test bpe-gpt2/encode-batch-no-cache ... bench: 1,144,185,772 ns/iter (+/- 6,991,090)

test llama3/encode ... bench: 1,654,386,307 ns/iter (+/- 7,932,582)
test llama3/encode-batch ... bench: 954,896,308 ns/iter (+/- 3,472,255)
test llama3/encode-fast ... bench: 1,580,607,181 ns/iter (+/- 7,765,018)
test llama3/encode-char-offsets ... bench: 1,151,526,922 ns/iter (+/- 4,410,764)
test llama3/concurrent-4t ... bench:  57,720,363 ns/iter (+/- 822,415)

test serialization/from-file-roberta ... bench:  69,615,893 ns/iter (+/- 1,473,155)
test serialization/from-file-llama3 ... bench: 408,322,490 ns/iter (+/- 6,740,072)
test serialization/from-file-albert ... bench:  50,251,605 ns/iter (+/- 1,124,802)
test serialization/bpe-from-file-gpt2 ... bench:  48,885,968 ns/iter (+/- 1,336,711)
test serialization/save-llama3 ... bench: 108,096,657 ns/iter (+/- 4,695,647)
test serialization/deserialize-roberta ... bench:  80,615,940 ns/iter (+/- 1,639,592)
test serialization/deserialize-llama3 ... bench: 428,055,220 ns/iter (+/- 4,835,669)

test train/bpe-small ... bench:  14,999,874 ns/iter (+/- 435,810)