in tiktoken_ext/openai_public.py [0:0]
def p50k_base():
mergeable_ranks = load_tiktoken_bpe(
"https://openaipublic.blob.core.windows.net/encodings/p50k_base.tiktoken",
expected_hash="94b5ca7dff4d00767bc256fdd1b27e5b17361d7b8a5f968547f9f23eb70d2069",
)
return {
"name": "p50k_base",
"explicit_n_vocab": 50281,
"pat_str": r50k_pat_str,
"mergeable_ranks": mergeable_ranks,
"special_tokens": {ENDOFTEXT: 50256},
}