MCPcopy Create free account
hub / github.com/THUDM/GLM / __init__

Method __init__

data_utils/sp_tokenizer.py:35–40  ·  view source on GitHub ↗
(self, encoder, bpe_merges)

Source from the content-addressed store, hash-verified

33
34class Encoder:
35 def __init__(self, encoder, bpe_merges):
36 self.encoder = encoder
37 self.decoder = {v: k for k, v in self.encoder.items()}
38 self.bpe_ranks = dict(zip(bpe_merges, range(len(bpe_merges))))
39 self.cache = {}
40 self.max_len = 0
41
42 def bpe(self, token):
43 if token in self.cache:

Callers

nothing calls this directly

Calls

no outgoing calls

Tested by

no test coverage detected