in tiktoken_ext/openai_public.py [0:0]
def r50k_base():
mergeable_ranks = load_tiktoken_bpe(
"https://openaipublic.blob.core.windows.net/encodings/r50k_base.tiktoken",
expected_hash="306cd27f03c1a714eca7108e03d66b7dc042abe8c258b44c199a7ed9838dd930",
)
return {
"name": "r50k_base",
"explicit_n_vocab": 50257,
"pat_str": r50k_pat_str,
"mergeable_ranks": mergeable_ranks,
"special_tokens": {ENDOFTEXT: 50256},
}